# AWS Summit Sydney — Free Knowledge Base — Full Content Source: https://aws-summit-2026-kb.pages.dev Last updated: 2026-05-09 This document contains the full text of every topic page and every session summary on the site, structured for LLM ingestion. Headings use markdown. Every block links back to a canonical URL. --- # PART 1 — TOPICS ## Agentic AI URL: https://aws-summit-2026-kb.pages.dev/topics/agentic-ai Tagline: Autonomous AI systems that plan, reason, and act on your behalf. ### Overview Agentic AI refers to systems where large language models go beyond answering prompts — they plan multi-step tasks, call tools and APIs, browse the web, execute code, and coordinate with other agents to achieve goals autonomously. Unlike a chatbot, an AI agent maintains state, reasons about what to do next, and takes action in the real world. AWS provides Amazon Bedrock AgentCore as the production runtime for these agents, with built-in identity, memory, observability, gateways, and a code interpreter sandbox. ### Key concepts - Reasoning loops (ReAct, Plan-and-Execute, Reflection patterns) - Tool use and function calling — letting an LLM invoke external APIs - Multi-agent orchestration — supervisor agents that delegate to specialist agents - Memory: short-term scratch memory and long-term persistent memory - Guardrails, evaluation, and observability for non-deterministic systems - Identity and authorization — agents acting on behalf of users ### Key AWS services - Amazon Bedrock AgentCore - Amazon Bedrock Agents - Strands Agents SDK - AWS Step Functions - Amazon Q ### Curated external resources - [Amazon Bedrock AgentCore — Official AWS page](https://aws.amazon.com/bedrock/agentcore/) - [Build Effective Agents — Anthropic engineering blog](https://www.anthropic.com/engineering/building-effective-agents) - [Strands Agents — Open source SDK by AWS](https://strandsagents.com/) - [AWS Bedrock Agents Workshop](https://catalog.workshops.aws/agents-for-amazon-bedrock) - [AgentCore deep dive — AWS Machine Learning Blog](https://aws.amazon.com/blogs/machine-learning/category/artificial-intelligence/amazon-bedrock-agents/) - [ReAct: Synergizing Reasoning and Acting in LLMs (paper)](https://arxiv.org/abs/2210.03629) ### Live monitored sources (Parallel AI) - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Artificial Intelligence May 2026 - arXiv.org](https://arxiv.org/list/cs.AI/current) — arxiv.org (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi ### Sessions on this topic (75) - PRT202-S — 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT202-S) - AIM201 — From demo to deployment: solving agentic AI's toughest challenges (https://aws-summit-2026-kb.pages.dev/sessions/AIM201) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - ARC301 — Build an AI-ready data foundation (https://aws-summit-2026-kb.pages.dev/sessions/ARC301) - MAM302 — Agentic AI for VMware migrations with AWS Transform for VMware (https://aws-summit-2026-kb.pages.dev/sessions/MAM302) - MAM306 — Adding Agentic AI to legacy apps with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/MAM306) - ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/ISV302) - AIM302 — Agentic AI Meets Responsible AI - Science, Strategy and Practice (https://aws-summit-2026-kb.pages.dev/sessions/AIM302) - ARC305 — Transforming from SaaS to multi-tenant agentic SaaS (https://aws-summit-2026-kb.pages.dev/sessions/ARC305) - COP301 — Elevating your Agentic AI Observability (https://aws-summit-2026-kb.pages.dev/sessions/COP301) - DVT201 — Building Software like never before with agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/DVT201) - MAM301 — From tech debt to competitive advantage: Migrate & modernize with AWS (https://aws-summit-2026-kb.pages.dev/sessions/MAM301) - MAM303 — Digital transformation excellence using agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/MAM303) - DEV205 — Securing Amazon Bedrock AgentCore: A Practical Framework (https://aws-summit-2026-kb.pages.dev/sessions/DEV205) - GHJ301 — R1 — AWS Game Day : Secret Agent Unicorns (https://aws-summit-2026-kb.pages.dev/sessions/GHJ301) - DEV313 — From Timeout to Throughput: Scaling Resilient Agentic Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV313) - STP210 — TeamForm's Generative Dashboards with Strands & Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/STP210) - PRT102-S — Efficiency to Innovation: How Agentic AI Unlocks New Business Models (https://aws-summit-2026-kb.pages.dev/sessions/PRT102-S) - PRT201-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT201-S) - PRT216-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT216-S) - PRT217-S — Your Agents Should Be Durable (https://aws-summit-2026-kb.pages.dev/sessions/PRT217-S) - AIM204 — Get to know Amazon Quick, your new agentic teammate (https://aws-summit-2026-kb.pages.dev/sessions/AIM204) - DEV211 — Evolution of Automation: Orchestration to Intent-Based Supervision (https://aws-summit-2026-kb.pages.dev/sessions/DEV211) - DEV301 — Evolution of Automation: Orchestration to Intent-Based Supervision (https://aws-summit-2026-kb.pages.dev/sessions/DEV301) - DEV304 — Building Agentic AI: Amazon Nova Act and Strands Agents in Practice (https://aws-summit-2026-kb.pages.dev/sessions/DEV304) - MAM307 — Modernise legacy code using fine-tuned Gen AI models (https://aws-summit-2026-kb.pages.dev/sessions/MAM307) - PRT101-S — Accelerating Innovation with GitLab DAP Powered by Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/PRT101-S) - PRT203-S — Design, Deploy, and Govern AI Agents with Boomis Agentstudio 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT203-S) - SEC305 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC305) - TNC202 — Accelerate Your Cloud Journey with AWS Transform (https://aws-summit-2026-kb.pages.dev/sessions/TNC202) - DEV312 — Strands Agents on Lambda: Observability With Powertools & X-Ray (https://aws-summit-2026-kb.pages.dev/sessions/DEV312) - STP211 — Authenticating AI Agents: How Kinde Navigates Agentic Identity (https://aws-summit-2026-kb.pages.dev/sessions/STP211) - AIM203 — Prompt Engineering to Learning Systems: Woodside's Agentic Ecosystem (https://aws-summit-2026-kb.pages.dev/sessions/AIM203) - ARC304 — Demystifying Agent Identity (https://aws-summit-2026-kb.pages.dev/sessions/ARC304) - DEV305 — Agents in the enterprise: Best practices with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/DEV305) - DEV401 — Build Intelligent Memory Systems for AI Agents (https://aws-summit-2026-kb.pages.dev/sessions/DEV401) - GHJ301 — R2 — AWS Game Day : Secret Agent Unicorns (https://aws-summit-2026-kb.pages.dev/sessions/GHJ301) - TNC201 — Explore the Agentic Capabilities of Amazon Quick Suite (https://aws-summit-2026-kb.pages.dev/sessions/TNC201) - DEV209 — CI/CD Guardrails for Agentic Coding Workflows (https://aws-summit-2026-kb.pages.dev/sessions/DEV209) - ISV304 — Managing AI Agents with Confidence and Control using Kasada & AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV304) - INO201 — Build and scale AI: from reliable agents to transformative systems (https://aws-summit-2026-kb.pages.dev/sessions/INO201) - SEC401 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC401) - DAT301 — Powering your Agentic AI experience with AWS Streaming and Messaging (https://aws-summit-2026-kb.pages.dev/sessions/DAT301) - MAM304 — Modernize SQL Server & .NET Together with AWS Transform's New AI Agent (https://aws-summit-2026-kb.pages.dev/sessions/MAM304) - MAM305 — Legacy App modernization and reverse engineering using Kiro (https://aws-summit-2026-kb.pages.dev/sessions/MAM305) - AIM101 — AI League Championship | 14-May | 08:00 - 16:00 (https://aws-summit-2026-kb.pages.dev/sessions/AIM101) - AIM403 — AI League (https://aws-summit-2026-kb.pages.dev/sessions/AIM403) - FSI206 — Agentic AI Transforming Quality at Cloud Speed (https://aws-summit-2026-kb.pages.dev/sessions/FSI206) - TNC203 — Structured Approach to AI coding with Spec-Driven Development on Kiro (https://aws-summit-2026-kb.pages.dev/sessions/TNC203) - FSI207 — From enterprise data mesh to AI with Amazon SageMaker Unified Studio (https://aws-summit-2026-kb.pages.dev/sessions/FSI207) - DEV306 — Taming Legacy Code: Multi-Agent AI in Brownfield Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV306) - ISV211 — Scaling Conversation Intelligence with Agentic AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV211) - ISV209 — From dev tools to customer value: BGL's agentic AI journey (https://aws-summit-2026-kb.pages.dev/sessions/ISV209) - STP209 — How Cartesian Turns AI Agents from SaaS Killer to SaaS Moat (https://aws-summit-2026-kb.pages.dev/sessions/STP209) - FSI204 — Agentic AI in Financial Services: Architectural Patterns That Work (https://aws-summit-2026-kb.pages.dev/sessions/FSI204) - IND204 — How Transurban Transformed Customer Experience with AI Agents on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND204) - TNC301 — Using Tools and Agents in Generative AI applications (https://aws-summit-2026-kb.pages.dev/sessions/TNC301) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) - STP212 — How Apate AI uses Amazon Bedrock and voice AI to catch scammers (https://aws-summit-2026-kb.pages.dev/sessions/STP212) - IND301 — Stockland Empowers People with a GenAI Assistant Built on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND301) - MAE205 — AI at Speed of News: Unlocking Value from Media with Generative AI (https://aws-summit-2026-kb.pages.dev/sessions/MAE205) - STP207 — How RedOwl Built Real-Time Financial Governance and Control on AWS (https://aws-summit-2026-kb.pages.dev/sessions/STP207) - IND201 — Transforming software license efficiency - Human-centered AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND201) - INO101 — From Zero to 270 AI Agents: how Lendi built Guardian (https://aws-summit-2026-kb.pages.dev/sessions/INO101) - ISV104 — hipages Journey Towards an Agentic Engineering Organisation (https://aws-summit-2026-kb.pages.dev/sessions/ISV104) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) - ISV214 — Grounding AI Agents: How to give your AI real-world data with MCP (https://aws-summit-2026-kb.pages.dev/sessions/ISV214) - WPS302 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS302) - STP202 — Stop Vibing, Start Shipping: How Startups Build with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/STP202) - INO202 — Build and scale AI: from reliable agents to transformative systems (https://aws-summit-2026-kb.pages.dev/sessions/INO202) - SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR (https://aws-summit-2026-kb.pages.dev/sessions/SMB203) - AIM304 — Agentic AI Meets Responsible AI - Science, Strategy and Practice (https://aws-summit-2026-kb.pages.dev/sessions/AIM304) - AIM402 — Agentic AI Meets responsible AI: Strategy and best practices (https://aws-summit-2026-kb.pages.dev/sessions/AIM402) - WPS202 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS202) - STP401 — How WhiteHorse AI Deploys Openclaw Agents on AWS with Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/STP401) --- ## Generative AI & Foundation Models URL: https://aws-summit-2026-kb.pages.dev/topics/generative-ai Tagline: Foundation models that generate text, images, code, and more. ### Overview Generative AI uses foundation models (FMs) — large neural networks trained on broad data — to generate new content. Amazon Bedrock is AWS's fully managed service that gives you access to leading FMs from Anthropic (Claude), Amazon (Titan, Nova), Meta (Llama), Mistral, Cohere, and AI21 through a single API, with built-in privacy, guardrails, fine-tuning, and knowledge bases. You don't need to manage infrastructure, GPUs, or model weights — just call the API. ### Key concepts - Foundation models vs. fine-tuned models vs. distilled models - Prompt engineering — system prompts, few-shot, chain-of-thought - Context windows, tokens, and inference cost optimization - Guardrails — content filtering, PII redaction, topic blocking - Model evaluation — automatic and human-in-the-loop - Provisioned throughput vs. on-demand inference ### Key AWS services - Amazon Bedrock - Amazon Nova - Amazon Titan - SageMaker JumpStart - Bedrock Guardrails ### Curated external resources - [Amazon Bedrock — Official AWS page](https://aws.amazon.com/bedrock/) - [Amazon Nova foundation models](https://aws.amazon.com/ai/generative-ai/nova/) - [Anthropic Claude on Bedrock](https://www.anthropic.com/claude) - [Generative AI on AWS — Learning Path](https://aws.amazon.com/training/learn-about/generative-ai/) - [Prompt Engineering Guide](https://www.promptingguide.ai/) - [Bedrock Workshop — hands-on labs](https://catalog.workshops.aws/amazon-bedrock) ### Live monitored sources (Parallel AI) - [Meta Acquires Assured Robot Intelligence for Humanoid AI](https://aitoolly.com/ai-news/article/2026-05-02-meta-acquires-humanoid-startup-assured-robot-intelligence-to-advance-ai-models-for-robotics) — aitoolly.com (2026-05-02): Meta has acquired Assured Robot Intelligence, a startup specializing in AI models for robots, to advance its humanoid robot technology. - [2026 - TechCrunch](https://techcrunch.com/2026/) — techcrunch.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [aws.amazon.com](https://aws.amazon.com/about-aws/whats-new/2026/04/bedrock-openai-models-codex-managed-agents/) — aws.amazon.com (2026-04-29): Amazon Bedrock (AWS) now offers OpenAI models, Codex, and Managed Agents (Limited Preview) — announced 2026-04-28. What changed: OpenAI models and Managed Agents are available inside AWS Bedrock limited preview, letting AWS customers run OpenAI models and managed agent capabiliti - [Top Tech News Today, May 1, 2026 - Tech Startups](https://techstartups.com/2026/05/01/top-tech-news-today-may-1-2026/) — techstartups.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [aitoolly.com](https://aitoolly.com/ai-news/2026-04-28) — aitoolly.com (2026-04-28): A new open-source project 'free-claude-code' was launched on GitHub, allowing developers to use Claude Code functionalities without Anthropic API keys. ### Sessions on this topic (50) - PRT104-S — Building Resilience for AI Data Foundations and Cloud-Native Apps 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT104-S) - PRT202-S — 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT202-S) - PRT204-S — Optimising GenAI at Runtime with Experimentation and Guardrails 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT204-S) - AIM401 — Beyond API Dependency: Fine-tuning Cost-Effective Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/AIM401) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - ARC301 — Build an AI-ready data foundation (https://aws-summit-2026-kb.pages.dev/sessions/ARC301) - MAM306 — Adding Agentic AI to legacy apps with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/MAM306) - ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/ISV302) - DEV205 — Securing Amazon Bedrock AgentCore: A Practical Framework (https://aws-summit-2026-kb.pages.dev/sessions/DEV205) - DEV313 — From Timeout to Throughput: Scaling Resilient Agentic Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV313) - PRT103-S — Cloud Anywhere: Architectural Freedom for Unified Data and AI 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT103-S) - DEV202 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV202) - DEV314 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV314) - MAM307 — Modernise legacy code using fine-tuned Gen AI models (https://aws-summit-2026-kb.pages.dev/sessions/MAM307) - PRT101-S — Accelerating Innovation with GitLab DAP Powered by Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/PRT101-S) - PRT203-S — Design, Deploy, and Govern AI Agents with Boomis Agentstudio 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT203-S) - COP302 — Applying AI for FinOps and FinOps for AI (https://aws-summit-2026-kb.pages.dev/sessions/COP302) - DEV305 — Agents in the enterprise: Best practices with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/DEV305) - DEV401 — Build Intelligent Memory Systems for AI Agents (https://aws-summit-2026-kb.pages.dev/sessions/DEV401) - STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI (https://aws-summit-2026-kb.pages.dev/sessions/STP208) - TNC201 — Explore the Agentic Capabilities of Amazon Quick Suite (https://aws-summit-2026-kb.pages.dev/sessions/TNC201) - INO201 — Build and scale AI: from reliable agents to transformative systems (https://aws-summit-2026-kb.pages.dev/sessions/INO201) - ARC303 — Unlock GenAI inference anywhere with Amazon EKS Hybrid Nodes (https://aws-summit-2026-kb.pages.dev/sessions/ARC303) - DAT301 — Powering your Agentic AI experience with AWS Streaming and Messaging (https://aws-summit-2026-kb.pages.dev/sessions/DAT301) - DEV311 — Serverless Developer Experience: Day in a life of builder (https://aws-summit-2026-kb.pages.dev/sessions/DEV311) - ARC307 — AI Powered Resilience Lifecycle (https://aws-summit-2026-kb.pages.dev/sessions/ARC307) - AIM101 — AI League Championship | 14-May | 08:00 - 16:00 (https://aws-summit-2026-kb.pages.dev/sessions/AIM101) - AIM403 — AI League (https://aws-summit-2026-kb.pages.dev/sessions/AIM403) - IDE301 — Diversity In Tech - AI Literacy Skills - Rapid prototyping with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/IDE301) - STP204 — How Heidi Health Fine-Tunes Speech-to-Text Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/STP204) - ISV209 — From dev tools to customer value: BGL's agentic AI journey (https://aws-summit-2026-kb.pages.dev/sessions/ISV209) - IND202 — How Zuru Uses AI to Analyze TikTok Trends for Rapid Content Creation (https://aws-summit-2026-kb.pages.dev/sessions/IND202) - TNC301 — Using Tools and Agents in Generative AI applications (https://aws-summit-2026-kb.pages.dev/sessions/TNC301) - STP301 — AI-Native Remediation with Pleri: Your Security Engineer That Ships (https://aws-summit-2026-kb.pages.dev/sessions/STP301) - DEV308 — AI Blast-Radius Reviews for AWS Changes Using Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/DEV308) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) - STP212 — How Apate AI uses Amazon Bedrock and voice AI to catch scammers (https://aws-summit-2026-kb.pages.dev/sessions/STP212) - IND301 — Stockland Empowers People with a GenAI Assistant Built on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND301) - INO203 — Behind the curtain: How Amazons AI innovations are powered by AWS (https://aws-summit-2026-kb.pages.dev/sessions/INO203) - MAE205 — AI at Speed of News: Unlocking Value from Media with Generative AI (https://aws-summit-2026-kb.pages.dev/sessions/MAE205) - STP101 — Driving Profitable Growth with Generative AI: From Prompt to Product (https://aws-summit-2026-kb.pages.dev/sessions/STP101) - ISV101 — How AI is Transforming Pharmacy Care with Amazon Nova:MedAdvisor Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV101) - DEV309 — AI Outputs: Amazon Bedrock Structured Output in Production (https://aws-summit-2026-kb.pages.dev/sessions/DEV309) - ISV104 — hipages Journey Towards an Agentic Engineering Organisation (https://aws-summit-2026-kb.pages.dev/sessions/ISV104) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) - ISV213 — From GRC Platform to AI-Native Risk Intelligence on AWS:Protecht Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV213) - INO202 — Build and scale AI: from reliable agents to transformative systems (https://aws-summit-2026-kb.pages.dev/sessions/INO202) - MAE204 — How Amazon Ads Creative Agent uses AWS to democratize ad creation (https://aws-summit-2026-kb.pages.dev/sessions/MAE204) - SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR (https://aws-summit-2026-kb.pages.dev/sessions/SMB203) - STP401 — How WhiteHorse AI Deploys Openclaw Agents on AWS with Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/STP401) --- ## Retrieval Augmented Generation (RAG) URL: https://aws-summit-2026-kb.pages.dev/topics/rag Tagline: Ground LLMs in your own data without retraining. ### Overview RAG combines a retrieval system (usually a vector database) with a generative model so the model can answer questions about private or up-to-date data it was never trained on. Documents are chunked, embedded into vectors, and stored. At query time, the most similar chunks are retrieved and inserted into the model's context. AWS offers Bedrock Knowledge Bases as a managed RAG pipeline, with Amazon OpenSearch Serverless, Aurora pgvector, MemoryDB, and S3 Vectors as vector store options. ### Key concepts - Embeddings and vector similarity (cosine, dot product) - Chunking strategies — fixed, semantic, hierarchical - Hybrid search — combining keyword (BM25) and vector search - Reranking with cross-encoders for precision - Graph RAG — using knowledge graphs alongside vectors - Evaluation: faithfulness, answer relevance, context recall ### Key AWS services - Bedrock Knowledge Bases - Amazon OpenSearch Serverless - Aurora PostgreSQL (pgvector) - Amazon MemoryDB - Amazon S3 Vectors ### Curated external resources - [Knowledge Bases for Amazon Bedrock](https://aws.amazon.com/bedrock/knowledge-bases/) - [What is RAG? — AWS explainer](https://aws.amazon.com/what-is/retrieval-augmented-generation/) - [Original RAG paper (Lewis et al., 2020)](https://arxiv.org/abs/2005.11401) - [pgvector — open source extension](https://github.com/pgvector/pgvector) - [Advanced RAG techniques — LangChain blog](https://blog.langchain.dev/) - [RAGAS evaluation framework](https://docs.ragas.io/) ### Live monitored sources (Parallel AI) - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv - [Agentic RAG: The 2026 Production Guide | MarsDevs](https://www.marsdevs.com/guides/agentic-rag-2026-guide) — marsdevs.com (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv - [Mem0 - The Memory Layer for your AI Apps](https://mem0.ai/) — mem0.ai (2026-05-09): Mem0 introduced 'Memory Decay,' a technical approach to long-term memory management that mimics human forgetting. The system implements a ranking score for memories that is reinforced upon each single retrieval and gently decayed over time if the memory remains untouched. This pr ### Sessions on this topic (51) - AIM201 — From demo to deployment: solving agentic AI's toughest challenges (https://aws-summit-2026-kb.pages.dev/sessions/AIM201) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - DAT304 — AI-Native by Design: How Deputy Rewired Its Operating Model on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DAT304) - MAM306 — Adding Agentic AI to legacy apps with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/MAM306) - ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/ISV302) - STP205 — How Dovetail powers Multi-Tenant Agents with Vector Indexing at Scale (https://aws-summit-2026-kb.pages.dev/sessions/STP205) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - AIM303 — AWS Security Agent: Proactive AppSec from Design to Deployment (https://aws-summit-2026-kb.pages.dev/sessions/AIM303) - ARC305 — Transforming from SaaS to multi-tenant agentic SaaS (https://aws-summit-2026-kb.pages.dev/sessions/ARC305) - MAM303 — Digital transformation excellence using agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/MAM303) - PRT201-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT201-S) - PRT207-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform (https://aws-summit-2026-kb.pages.dev/sessions/PRT207-S) - PRT209-S — How Auto & General leverage observability foundations for AI (https://aws-summit-2026-kb.pages.dev/sessions/PRT209-S) - PRT213-S — How NAB is Conquering Multi-Cloud to Secure the Enterprise (https://aws-summit-2026-kb.pages.dev/sessions/PRT213-S) - PRT216-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT216-S) - PRT301-S — Unite Teams, Tools, and AI to Drive Transformation at Scale (https://aws-summit-2026-kb.pages.dev/sessions/PRT301-S) - DEV202 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV202) - DEV314 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV314) - SEC305 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC305) - TNC202 — Accelerate Your Cloud Journey with AWS Transform (https://aws-summit-2026-kb.pages.dev/sessions/TNC202) - ISV301 — Rolling to Scale: Roller's Multi-Tenant SaaS platform on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV301) - DEV201 — How Flybuys Built AI Governance to Accelerate Adoption at Scale (https://aws-summit-2026-kb.pages.dev/sessions/DEV201) - STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI (https://aws-summit-2026-kb.pages.dev/sessions/STP208) - SEC401 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC401) - DEV311 — Serverless Developer Experience: Day in a life of builder (https://aws-summit-2026-kb.pages.dev/sessions/DEV311) - ISV206 — Scaling RAG to Millions of Vectors: The Squiz Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV206) - ARC302 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC302) - ARC307 — AI Powered Resilience Lifecycle (https://aws-summit-2026-kb.pages.dev/sessions/ARC307) - ARC403 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC403) - IDE301 — Diversity In Tech - AI Literacy Skills - Rapid prototyping with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/IDE301) - ISV201 — MCP on EKS: Xero's AI-Driven Developer Experience (https://aws-summit-2026-kb.pages.dev/sessions/ISV201) - PRT210-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform (https://aws-summit-2026-kb.pages.dev/sessions/PRT210-S) - SMB202 — PMY Delivers Realtime Crowd Analytics at the F1 Australian Grand Prix (https://aws-summit-2026-kb.pages.dev/sessions/SMB202) - TNC203 — Structured Approach to AI coding with Spec-Driven Development on Kiro (https://aws-summit-2026-kb.pages.dev/sessions/TNC203) - PRT106-S — The AI Challenge You Don't Yet Know About - Software Supply Chain (https://aws-summit-2026-kb.pages.dev/sessions/PRT106-S) - ISV202 — Architecting for growth and resilience: Cell based design deep dive (https://aws-summit-2026-kb.pages.dev/sessions/ISV202) - MAE202 — Seven's AWS Journey: Streaming Premium Content at the Speed of Innovation (https://aws-summit-2026-kb.pages.dev/sessions/MAE202) - STP213 — AI-Powered Farming: How Halter's ML Models Transform Dairy Operations (https://aws-summit-2026-kb.pages.dev/sessions/STP213) - TNC301 — Using Tools and Agents in Generative AI applications (https://aws-summit-2026-kb.pages.dev/sessions/TNC301) - IDE101 — From principles to practice: Scaling AI responsibly (https://aws-summit-2026-kb.pages.dev/sessions/IDE101) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) - IDE102 — Power of Possibility: Leading Through Innovation and Connection (https://aws-summit-2026-kb.pages.dev/sessions/IDE102) - INO203 — Behind the curtain: How Amazons AI innovations are powered by AWS (https://aws-summit-2026-kb.pages.dev/sessions/INO203) - IND201 — Transforming software license efficiency - Human-centered AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND201) - ISV104 — hipages Journey Towards an Agentic Engineering Organisation (https://aws-summit-2026-kb.pages.dev/sessions/ISV104) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) - IND101 — Test, Learn, Iterate: Amazon Connect Success (https://aws-summit-2026-kb.pages.dev/sessions/IND101) - IND206 — How scalable data foundations helped TGE unlock the power of AI (https://aws-summit-2026-kb.pages.dev/sessions/IND206) - ISV213 — From GRC Platform to AI-Native Risk Intelligence on AWS:Protecht Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV213) - ISV207 — How Canva Scales and Optimizes AI Workloads with Karpenter (https://aws-summit-2026-kb.pages.dev/sessions/ISV207) - FSI203 — How HBF Transformed Claims Processing From Two Weeks to Two Minutes (https://aws-summit-2026-kb.pages.dev/sessions/FSI203) --- ## Model Context Protocol (MCP) URL: https://aws-summit-2026-kb.pages.dev/topics/mcp Tagline: The open standard that lets AI assistants plug into any tool or data source. ### Overview Model Context Protocol (MCP) is an open standard, originally introduced by Anthropic, that defines how LLM-based applications connect to external data sources and tools. Think of it as "USB-C for AI." Once a service exposes an MCP server, any MCP-compatible client (Claude Desktop, Kiro, Cursor, Cline, Q CLI) can use its tools, resources, and prompts without custom integration. AWS has published MCP servers for EKS, Aurora, Knowledge Bases, CloudWatch, Documentation, and more. ### Key concepts - MCP servers expose three primitives: tools, resources, and prompts - Transport: stdio for local, SSE/HTTP for remote - Capability negotiation between client and server - Authentication and scoping — what the agent is allowed to do - Composing multiple MCP servers in one assistant ### Key AWS services - AWS Documentation MCP Server - EKS MCP Server - Aurora MCP Server - Bedrock AgentCore Gateway ### Curated external resources - [Model Context Protocol — official site](https://modelcontextprotocol.io/) - [AWS MCP Servers on GitHub](https://github.com/awslabs/mcp) - [Introducing MCP — Anthropic announcement](https://www.anthropic.com/news/model-context-protocol) - [MCP server registry](https://github.com/modelcontextprotocol/servers) - [MCP specification](https://spec.modelcontextprotocol.io/) ### Live monitored sources (Parallel AI) - [The MCP Ecosystem in 2026: How the Model Context Protocol ...](https://chatforest.com/guides/mcp-ecosystem-2026-state-of-the-standard) — chatforest.com (2026-05-08): Azure.Mcp.Server released version 3.0.0-beta.10, which improves tool validation to ensure the complete registered tool set is recognized at runtime and enhances MSAL error handling with PII-safe telemetry and more accurate exception mapping. - [AI MCP OAuth2 - Plugin - Kong Docs](http://developer.konghq.com/plugins/ai-mcp-oauth2) — developer.konghq.com (2026-05-08): Azure.Mcp.Server released version 3.0.0-beta.10, which improves tool validation to ensure the complete registered tool set is recognized at runtime and enhances MSAL error handling with PII-safe telemetry and more accurate exception mapping. - [AI News — 2026-05-05 | SkillsLLM](https://skillsllm.com/news/ai-news-2026-05-05) — skillsllm.com (2026-05-08): Azure.Mcp.Server released version 3.0.0-beta.10, which improves tool validation to ensure the complete registered tool set is recognized at runtime and enhances MSAL error handling with PII-safe telemetry and more accurate exception mapping. - [Releases · microsoft/agent-framework · GitHub](https://github.com/microsoft/agent-framework/releases) — github.com (2026-05-08): Microsoft Agent Framework released version dotnet-1.5.0, introducing a significant shift to standardize on the Model Context Protocol (MCP) for toolbox consumption. The update also adds hosted agent observability samples and implements message filtering for non-portable content t - [github.com](https://github.com/mcpjungle/MCPJungle) — github.com (2026-04-28): MCPJungle was introduced as a new open-source resource/tool within the Model Context Protocol ecosystem to extend agent capabilities. ### Sessions on this topic (11) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - ARC305 — Transforming from SaaS to multi-tenant agentic SaaS (https://aws-summit-2026-kb.pages.dev/sessions/ARC305) - PRT201-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT201-S) - PRT216-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT216-S) - DEV202 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV202) - DEV314 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV314) - DEV305 — Agents in the enterprise: Best practices with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/DEV305) - ISV201 — MCP on EKS: Xero's AI-Driven Developer Experience (https://aws-summit-2026-kb.pages.dev/sessions/ISV201) - TNC301 — Using Tools and Agents in Generative AI applications (https://aws-summit-2026-kb.pages.dev/sessions/TNC301) - DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops (https://aws-summit-2026-kb.pages.dev/sessions/DEV206) - ISV214 — Grounding AI Agents: How to give your AI real-world data with MCP (https://aws-summit-2026-kb.pages.dev/sessions/ISV214) --- ## Kiro & Spec-Driven Development URL: https://aws-summit-2026-kb.pages.dev/topics/kiro Tagline: Move beyond vibe-coding with AWS's agentic IDE. ### Overview Kiro is AWS's agentic IDE that introduces spec-driven development: instead of prompting an AI to "build me an app," you collaborate with the agent to write a structured spec (requirements, design, tasks), then the agent implements it step by step. This produces production-quality code with traceability — every line of code links back to a requirement. Kiro ships as a desktop IDE and as Kiro CLI for terminal-based workflows. ### Key concepts - Spec-driven workflow: Requirements → Design → Tasks → Code - Hooks: file-save and event-triggered automation - Steering files: persistent project conventions and context - Agentic file edits with diff preview and approval - Multi-step planning with self-correction ### Key AWS services - Kiro IDE - Kiro CLI - Amazon Q Developer ### Curated external resources - [Kiro — official site](https://kiro.dev/) - [Kiro documentation](https://kiro.dev/docs/) - [Kiro launch announcement](https://aws.amazon.com/blogs/aws/introducing-kiro-an-agentic-ide-that-thinks-with-you/) - [Spec-driven development explained](https://kiro.dev/docs/specs/) - [Amazon Q Developer](https://aws.amazon.com/q/developer/) ### Live monitored sources (Parallel AI) - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [Fetched web page](https://windsurf.com/changelog) — indsurf.com (2026-05-07): Windsurf released version 2.2.17, granting all IDE users access to Devin Review and Quick Review. The update also includes improvements to the Agent Command Center (list display for inbox, better session sorting/filtering) and a fix for Windows update issues. - [github.blog](https://github.blog/changelog/2026-04-24-gpt-5-5-is-generally-available-for-github-copilot) — github.blog (2026-04-26): GPT-5.5 is now generally available for GitHub Copilot, delivering stronger performance on complex, multi-step agentic coding tasks. - [github.blog](https://github.blog/changelog/2026-05-05-secret-scanning-with-github-mcp-server-is-now-generally-available) — github.blog (2026-05-06): GitHub released secret scanning in the GitHub MCP (Model Context Protocol) server as generally available, allowing MCP-compatible AI coding agents and IDEs (like GitHub Copilot CLI or VS Code) to scan for exposed secrets before they are committed or part of a pull request. - [macrocode.ai](https://www.macrocode.ai/blog/introducing-the-agentic-sdlc-framework/) — macrocode.ai (2026-04-11): macrocode.ai published "Introducing the Agentic SDLC Framework" (Agentic Flow Framework) — a new governed multi-agent SDLC approach that formalizes spec-driven, auditable agent development. The post (2026-04-11) reports measured data from a 14-day production run (metrics include ### Sessions on this topic (17) - DVT201 — Building Software like never before with agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/DVT201) - PRT201-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT201-S) - PRT216-S — Postman and the Future of AI-Driven API Development in 2026 (https://aws-summit-2026-kb.pages.dev/sessions/PRT216-S) - DEV202 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV202) - DEV314 — AI Native Development: Strategies and Impact across Amazon and AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV314) - COP302 — Applying AI for FinOps and FinOps for AI (https://aws-summit-2026-kb.pages.dev/sessions/COP302) - DEV201 — How Flybuys Built AI Governance to Accelerate Adoption at Scale (https://aws-summit-2026-kb.pages.dev/sessions/DEV201) - MAM305 — Legacy App modernization and reverse engineering using Kiro (https://aws-summit-2026-kb.pages.dev/sessions/MAM305) - IDE301 — Diversity In Tech - AI Literacy Skills - Rapid prototyping with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/IDE301) - ISV201 — MCP on EKS: Xero's AI-Driven Developer Experience (https://aws-summit-2026-kb.pages.dev/sessions/ISV201) - TNC203 — Structured Approach to AI coding with Spec-Driven Development on Kiro (https://aws-summit-2026-kb.pages.dev/sessions/TNC203) - DEV306 — Taming Legacy Code: Multi-Agent AI in Brownfield Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV306) - SMB201 — The AI-Driven Development Lifecycle: How Skyjed Shipped in 48 hours (https://aws-summit-2026-kb.pages.dev/sessions/SMB201) - DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops (https://aws-summit-2026-kb.pages.dev/sessions/DEV206) - SMB205 — How Blackmores accelerated SAP RISE connectivity with an EBA and Kiro (https://aws-summit-2026-kb.pages.dev/sessions/SMB205) - STP202 — Stop Vibing, Start Shipping: How Startups Build with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/STP202) - FSI202 — Accelerating Payment Innovation: Spec-Driven Development with AWS Kiro (https://aws-summit-2026-kb.pages.dev/sessions/FSI202) --- ## Amazon Q & AI Assistants URL: https://aws-summit-2026-kb.pages.dev/topics/amazon-q Tagline: Generative AI assistants for developers, business users, and contact centers. ### Overview Amazon Q is AWS's family of generative AI assistants. Q Developer accelerates coding with inline suggestions, transformations (Java, .NET, mainframe), and security scans. Q Business is an enterprise assistant that connects to 40+ data sources (SharePoint, Salesforce, Google Drive, Confluence) and answers questions while respecting access controls. Q in Connect helps contact center agents in real-time with knowledge and next-best actions. ### Key concepts - Code suggestions, refactoring, and modernization - Document Q&A with row-level security - Transformation agents — Java upgrades, .NET porting, mainframe - Plugins and custom data connectors - Q Apps — citizen-developer no-code AI apps ### Key AWS services - Amazon Q Developer - Amazon Q Business - Amazon Q in Connect - Amazon Q in QuickSight ### Curated external resources - [Amazon Q overview](https://aws.amazon.com/q/) - [Amazon Q Developer](https://aws.amazon.com/q/developer/) - [Amazon Q Business](https://aws.amazon.com/q/business/) - [Q Developer plugins for IDEs](https://docs.aws.amazon.com/amazonq/latest/qdeveloper-ug/q-in-IDE.html) - [Q Developer Free Tier](https://aws.amazon.com/q/developer/pricing/) ### Live monitored sources (Parallel AI) - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [Devin AI Guide 2026: Features, Pricing, How to Use & Complete ...](https://aitoolsdevpro.com/ai-tools/devin-guide/) — aitoolsdevpro.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [Cursor AI 2026: Complete Guide to New Features, Tips ...](https://anycap.ai/page/en-US/blog/cursor-ai-2026-new-features-guide) — anycap.ai (2026-05-09): Reports indicate that GitHub Copilot paused new sign-ups for its Pro, Pro+, and Student tiers and removed Claude Opus models from the Pro tier in May 2026. - [Fetched web page](https://windsurf.com/changelog) — indsurf.com (2026-05-07): Windsurf released version 2.2.17, granting all IDE users access to Devin Review and Quick Review. The update also includes improvements to the Agent Command Center (list display for inbox, better session sorting/filtering) and a fix for Windows update issues. - [github.blog](https://github.blog/changelog/2026-04-24-inline-agent-mode-in-preview-and-more-in-github-copilot-for-jetbrains-ides) — github.blog (2026-04-26): GitHub Copilot for JetBrains IDEs introduced an inline agent mode in preview, alongside enhancements to Next Edit Suggestions and global auto-approve. ### Sessions on this topic (2) - AIM204 — Get to know Amazon Quick, your new agentic teammate (https://aws-summit-2026-kb.pages.dev/sessions/AIM204) - TNC201 — Explore the Agentic Capabilities of Amazon Quick Suite (https://aws-summit-2026-kb.pages.dev/sessions/TNC201) --- ## Machine Learning & SageMaker URL: https://aws-summit-2026-kb.pages.dev/topics/sagemaker Tagline: The end-to-end platform for building, training, and deploying ML models. ### Overview Amazon SageMaker AI is the managed service for the full machine-learning lifecycle: data labeling, notebooks, training jobs (including distributed training on thousands of GPUs/Trainium chips), hyperparameter tuning, model registry, deployment, and monitoring. SageMaker Unified Studio brings together SageMaker, Bedrock, Glue, EMR, Redshift, and QuickSight in one workspace so data engineers, data scientists, and analysts collaborate on the same data. ### Key concepts - Training: SageMaker Training Jobs, distributed training, spot instances - Fine-tuning and distillation for cost-effective specialization - Inference: real-time, serverless, asynchronous, batch transform - Model Registry, Pipelines, and MLOps automation - Feature Store for reusable features across teams - AWS Trainium and Inferentia for cost-optimized ML ### Key AWS services - Amazon SageMaker AI - SageMaker Unified Studio - AWS Trainium - AWS Inferentia - SageMaker JumpStart ### Curated external resources - [Amazon SageMaker AI](https://aws.amazon.com/sagemaker-ai/) - [SageMaker Unified Studio](https://aws.amazon.com/sagemaker/unified-studio/) - [AWS Machine Learning Blog](https://aws.amazon.com/blogs/machine-learning/) - [SageMaker Examples on GitHub](https://github.com/aws/amazon-sagemaker-examples) - [AWS Trainium & Inferentia](https://aws.amazon.com/ai/machine-learning/trainium/) ### Live monitored sources (Parallel AI) - [arxiv.org](https://arxiv.org/abs/2604.06111) — arxiv.org (2026-04-08): ACE-Bench: Agent Configurable Evaluation with Scalable Horizons and Controllable Difficulty. Methodology: unified grid-based planning tasks where agents fill hidden slots with orthogonal controls for Scalable Horizons (H) and Controllable Difficulty (decoy budget B); tools resolv - [arxiv.org](https://arxiv.org/abs/2604.06696) — arxiv.org (2026-04-09): AgentGate: A Lightweight Structured Routing Engine for the Internet of Agents. Methodology: two-stage structured routing framework (action decision + structural grounding) that formulates routing as a constrained decision problem, plus a routing-oriented fine-tuning scheme with c - [arxiv.org](https://arxiv.org/abs/2604.05523) — arxiv.org (2026-04-08): Market-Bench: Benchmarking Large Language Models on Economic and Trade Competition. Methodology: configurable multi-agent supply-chain economic model where LLMs act as retailer agents in procurement (bids/auctions) and retail (pricing, marketing) stages; logs full trajectories (b - [benchlm.ai](https://benchlm.ai/llm-agent-benchmarks) — benchlm.ai (2026-04-23): BenchLM.ai updated its AI Agent & Tool-Use Leaderboard (Apr 23, 2026). Methodology: An 'Agentic Score' is calculated as a weighted average of Terminal-Bench 2.0 (40%), OSWorld-Verified (35%), and BrowseComp (25%). It tracks 24 agentic benchmarks including MCP Atlas and Toolathlon - [Agentic Benchmarks 2026: Tool Use, Browsing, Computer Use | BenchLM.ai](https://benchlm.ai/agentic) — benchlm.ai (2026-05-12): BenchLM.ai updated its Agentic Benchmarks leaderboard on 2026-05-11. The update introduced two new benchmarks: 1) OpenHands Index, a holistic coding-agent benchmark covering issue resolution, frontend work, greenfield development, testing, and information gathering; and 2) SWE-At ### Sessions on this topic (17) - AIM401 — Beyond API Dependency: Fine-tuning Cost-Effective Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/AIM401) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - MAM307 — Modernise legacy code using fine-tuned Gen AI models (https://aws-summit-2026-kb.pages.dev/sessions/MAM307) - COP302 — Applying AI for FinOps and FinOps for AI (https://aws-summit-2026-kb.pages.dev/sessions/COP302) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - DEV201 — How Flybuys Built AI Governance to Accelerate Adoption at Scale (https://aws-summit-2026-kb.pages.dev/sessions/DEV201) - DAT303 — Explore whats new in data and AI governance with SageMaker Catalog (https://aws-summit-2026-kb.pages.dev/sessions/DAT303) - WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health (https://aws-summit-2026-kb.pages.dev/sessions/WPS203) - FSI207 — From enterprise data mesh to AI with Amazon SageMaker Unified Studio (https://aws-summit-2026-kb.pages.dev/sessions/FSI207) - STP213 — AI-Powered Farming: How Halter's ML Models Transform Dairy Operations (https://aws-summit-2026-kb.pages.dev/sessions/STP213) - STP204 — How Heidi Health Fine-Tunes Speech-to-Text Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/STP204) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) - STP212 — How Apate AI uses Amazon Bedrock and voice AI to catch scammers (https://aws-summit-2026-kb.pages.dev/sessions/STP212) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) - IND101 — Test, Learn, Iterate: Amazon Connect Success (https://aws-summit-2026-kb.pages.dev/sessions/IND101) - FSI202 — Accelerating Payment Innovation: Spec-Driven Development with AWS Kiro (https://aws-summit-2026-kb.pages.dev/sessions/FSI202) - MAE204 — How Amazon Ads Creative Agent uses AWS to democratize ad creation (https://aws-summit-2026-kb.pages.dev/sessions/MAE204) --- ## Data Lakes, Lakehouse & AI-Ready Data URL: https://aws-summit-2026-kb.pages.dev/topics/data-foundation Tagline: Build the data foundation that powers analytics and AI. ### Overview A modern data foundation unifies operational, analytical, and AI workloads on open formats so data is queryable from any engine without copies. AWS's strategy centers on Amazon S3 Tables (managed Apache Iceberg), Amazon SageMaker Lakehouse (unified access across S3, Redshift, federated sources), AWS Glue for catalog and ETL, and Amazon DataZone for governance. With Iceberg as the open table format, you get ACID transactions, time travel, and schema evolution on cheap S3 storage. ### Key concepts - Apache Iceberg — open table format with ACID semantics - Catalog standardization (AWS Glue Data Catalog, AWS Lake Formation) - Zero-ETL integrations between operational stores and analytics - Data quality, lineage, and active metadata - Fine-grained access control with Lake Formation ### Key AWS services - Amazon S3 Tables - AWS Glue - AWS Lake Formation - Amazon DataZone - SageMaker Lakehouse ### Curated external resources - [Amazon S3 Tables](https://aws.amazon.com/s3/features/tables/) - [Apache Iceberg — official site](https://iceberg.apache.org/) - [AWS Lake Formation](https://aws.amazon.com/lake-formation/) - [Amazon DataZone](https://aws.amazon.com/datazone/) - [Building data lakes on AWS — whitepaper](https://docs.aws.amazon.com/whitepapers/latest/building-data-lakes/building-data-lake-aws.html) ### Live monitored sources (Parallel AI) - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Event Sourcing for Agents: Log-Based Architecture for ...](https://understandingdata.com/posts/event-sourcing-agents/) — understandingdata.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [businesswire.com](https://www.businesswire.com/news/home/20260422902027/en/Bedrock-Datas-ArgusAI-Now-Governs-AI-Agents-Built-on-Google-Vertex-AI-and-Dialogflow) — businesswire.com (2026-04-22): Bedrock Data announced that ArgusAI now provides governance and agent-aware access control for AI agents built on Google Vertex AI Search and Dialogflow. The platform implements a 'Data Bill of Materials (DBOM)' to automatically discover and map the data stores accessed by agents ### Sessions on this topic (4) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - ISV303 — From hours to minutes: SafetyCulture's journey to 90% faster analytics (https://aws-summit-2026-kb.pages.dev/sessions/ISV303) - DAT401 — Real-Time DataLakes with Apache Iceberg, Amazon MSK, and Amazon S3 (https://aws-summit-2026-kb.pages.dev/sessions/DAT401) - DAT201 — Scaling Data Analytics: Easygo's Modern Lakehouse Journey on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DAT201) --- ## Analytics, Redshift & Generative BI URL: https://aws-summit-2026-kb.pages.dev/topics/analytics Tagline: Turn raw data into insights — and let AI write your queries. ### Overview AWS's analytics stack spans Amazon Redshift (cloud data warehouse with serverless and zero-ETL options), Amazon Athena (serverless SQL on S3), Amazon EMR (managed Spark / Trino / Flink), AWS Glue (ETL), and Amazon QuickSight (BI with the Q natural-language assistant). Generative BI lets non-SQL users ask questions in plain English and get charts, summaries, and stories — democratizing data access across the business. ### Key concepts - Cloud data warehouses vs. data lakes vs. lakehouses - Columnar storage and MPP query execution - Federated queries — joining data across sources without ETL - Materialized views and result caching for cost control - Generative BI: NL2SQL, narrative summaries, auto-charts ### Key AWS services - Amazon Redshift - Amazon Athena - Amazon EMR - Amazon QuickSight - AWS Glue ### Curated external resources - [Amazon Redshift](https://aws.amazon.com/redshift/) - [Amazon QuickSight + Amazon Q](https://aws.amazon.com/quicksight/q/) - [Amazon Athena](https://aws.amazon.com/athena/) - [AWS Big Data Blog](https://aws.amazon.com/blogs/big-data/) - [AWS Analytics learning path](https://aws.amazon.com/training/learn-about/analytics/) ### Sessions on this topic (16) - ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/ISV302) - COP301 — Elevating your Agentic AI Observability (https://aws-summit-2026-kb.pages.dev/sessions/COP301) - ISV303 — From hours to minutes: SafetyCulture's journey to 90% faster analytics (https://aws-summit-2026-kb.pages.dev/sessions/ISV303) - STP210 — TeamForm's Generative Dashboards with Strands & Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/STP210) - AIM204 — Get to know Amazon Quick, your new agentic teammate (https://aws-summit-2026-kb.pages.dev/sessions/AIM204) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - TNC201 — Explore the Agentic Capabilities of Amazon Quick Suite (https://aws-summit-2026-kb.pages.dev/sessions/TNC201) - STP302 — Unleash Live: Cloud-Powered Vision for Infrastructure (https://aws-summit-2026-kb.pages.dev/sessions/STP302) - DAT401 — Real-Time DataLakes with Apache Iceberg, Amazon MSK, and Amazon S3 (https://aws-summit-2026-kb.pages.dev/sessions/DAT401) - DEV210 — AI-Driven Incident Triage: From Slack Alert to Root Cause (https://aws-summit-2026-kb.pages.dev/sessions/DEV210) - DAT201 — Scaling Data Analytics: Easygo's Modern Lakehouse Journey on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DAT201) - DAT303 — Explore whats new in data and AI governance with SageMaker Catalog (https://aws-summit-2026-kb.pages.dev/sessions/DAT303) - SMB202 — PMY Delivers Realtime Crowd Analytics at the F1 Australian Grand Prix (https://aws-summit-2026-kb.pages.dev/sessions/SMB202) - PRT111-S — From Risk to Resilience - How Mimecast Works with AWS (https://aws-summit-2026-kb.pages.dev/sessions/PRT111-S) - WPS301 — AWS for healthcare analytics: accelerating time to insights (https://aws-summit-2026-kb.pages.dev/sessions/WPS301) - ISV211 — Scaling Conversation Intelligence with Agentic AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV211) --- ## Streaming & Real-Time Data URL: https://aws-summit-2026-kb.pages.dev/topics/streaming Tagline: Process millions of events per second as they happen. ### Overview Real-time data systems on AWS use Amazon Kinesis (Data Streams, Firehose) and Amazon MSK (managed Apache Kafka) to ingest event streams, then process with Apache Flink (Amazon Managed Service for Apache Flink), Lambda, or EMR. Common patterns include change-data-capture (CDC), clickstream analytics, IoT telemetry, fraud detection, and real-time personalization. Kinesis Data Streams now integrates with Amazon Bedrock for streaming LLM inference. ### Key concepts - Event-driven architectures vs. batch ETL - Apache Kafka topics, partitions, consumer groups - Stream processing: stateful operators, windows, exactly-once - Change Data Capture (CDC) with Debezium / DMS - Backpressure, scaling, and replay for resilience ### Key AWS services - Amazon Kinesis - Amazon MSK - Managed Service for Apache Flink - Amazon EventBridge ### Curated external resources - [Amazon Kinesis](https://aws.amazon.com/kinesis/) - [Amazon MSK (Managed Kafka)](https://aws.amazon.com/msk/) - [Managed Service for Apache Flink](https://aws.amazon.com/managed-service-apache-flink/) - [Apache Kafka — official](https://kafka.apache.org/) - [Apache Flink — official](https://flink.apache.org/) ### Sessions on this topic (9) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - ARC303 — Unlock GenAI inference anywhere with Amazon EKS Hybrid Nodes (https://aws-summit-2026-kb.pages.dev/sessions/ARC303) - DAT301 — Powering your Agentic AI experience with AWS Streaming and Messaging (https://aws-summit-2026-kb.pages.dev/sessions/DAT301) - DAT401 — Real-Time DataLakes with Apache Iceberg, Amazon MSK, and Amazon S3 (https://aws-summit-2026-kb.pages.dev/sessions/DAT401) - DAT201 — Scaling Data Analytics: Easygo's Modern Lakehouse Journey on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DAT201) - MAE202 — Seven's AWS Journey: Streaming Premium Content at the Speed of Innovation (https://aws-summit-2026-kb.pages.dev/sessions/MAE202) - WPS301 — AWS for healthcare analytics: accelerating time to insights (https://aws-summit-2026-kb.pages.dev/sessions/WPS301) - STP212 — How Apate AI uses Amazon Bedrock and voice AI to catch scammers (https://aws-summit-2026-kb.pages.dev/sessions/STP212) --- ## Databases & Aurora URL: https://aws-summit-2026-kb.pages.dev/topics/databases Tagline: Purpose-built databases for every workload. ### Overview AWS believes in purpose-built databases — the right tool for the job. Amazon Aurora (PostgreSQL/MySQL compatible) handles relational workloads at cloud scale, with Aurora DSQL providing active-active multi-region with strong consistency. DynamoDB serves single-digit-millisecond key-value/document workloads. ElastiCache and MemoryDB cover caching and in-memory data. Neptune handles graphs, Timestream handles time-series, and DocumentDB is MongoDB-compatible. All integrate with Bedrock for vector storage and AI workloads. ### Key concepts - Relational vs. NoSQL vs. graph vs. time-series — choosing the right model - Aurora storage architecture: 6-way replication, log-structured - DynamoDB: partition keys, GSIs, on-demand vs. provisioned - Vector storage: pgvector in Aurora, vector search in OpenSearch - Zero-ETL between Aurora/DynamoDB and Redshift ### Key AWS services - Amazon Aurora - Amazon Aurora DSQL - Amazon DynamoDB - Amazon Neptune - Amazon DocumentDB - Amazon Timestream ### Curated external resources - [Amazon Aurora](https://aws.amazon.com/rds/aurora/) - [Aurora DSQL](https://aws.amazon.com/rds/aurora/dsql/) - [Amazon DynamoDB](https://aws.amazon.com/dynamodb/) - [AWS Database Blog](https://aws.amazon.com/blogs/database/) - [Choosing the right AWS database — guide](https://docs.aws.amazon.com/decision-guides/latest/databases-on-aws-how-to-choose/databases-on-aws-how-to-choose.html) ### Live monitored sources (Parallel AI) - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [prnewswire.com](https://www.prnewswire.com/news-releases/constructive-open-sources-agentic-db-the-postgres-memory-layer-for-ai-agents-302755269.html) — prnewswire.com (2026-04-29): Constructive announced and open-sourced "agentic-db": a purpose-built Postgres "memory layer" for AI agents providing long-term episodic memory, conversation and tool-call event logs, token accounting, a versioned skills/tools registry, rules/behavioral policies for governance, t - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent ### Sessions on this topic (21) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - COP301 — Elevating your Agentic AI Observability (https://aws-summit-2026-kb.pages.dev/sessions/COP301) - ISV210 — Boost performance and reduce costs with Aurora: Canva's story (https://aws-summit-2026-kb.pages.dev/sessions/ISV210) - STP210 — TeamForm's Generative Dashboards with Strands & Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/STP210) - PRT112-S — Empower Data with Oracle AI Database and Native AI Services on AWS (https://aws-summit-2026-kb.pages.dev/sessions/PRT112-S) - AIM204 — Get to know Amazon Quick, your new agentic teammate (https://aws-summit-2026-kb.pages.dev/sessions/AIM204) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - DEV201 — How Flybuys Built AI Governance to Accelerate Adoption at Scale (https://aws-summit-2026-kb.pages.dev/sessions/DEV201) - ARC402 — DynamoDB: Resilience & lessons from the Oct 2025 service disruption (https://aws-summit-2026-kb.pages.dev/sessions/ARC402) - DAT201 — Scaling Data Analytics: Easygo's Modern Lakehouse Journey on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DAT201) - DAT303 — Explore whats new in data and AI governance with SageMaker Catalog (https://aws-summit-2026-kb.pages.dev/sessions/DAT303) - AIM403 — AI League (https://aws-summit-2026-kb.pages.dev/sessions/AIM403) - ISV211 — Scaling Conversation Intelligence with Agentic AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV211) - DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV203) - MAE205 — AI at Speed of News: Unlocking Value from Media with Generative AI (https://aws-summit-2026-kb.pages.dev/sessions/MAE205) - INO101 — From Zero to 270 AI Agents: how Lendi built Guardian (https://aws-summit-2026-kb.pages.dev/sessions/INO101) - ISV104 — hipages Journey Towards an Agentic Engineering Organisation (https://aws-summit-2026-kb.pages.dev/sessions/ISV104) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) - FSI202 — Accelerating Payment Innovation: Spec-Driven Development with AWS Kiro (https://aws-summit-2026-kb.pages.dev/sessions/FSI202) - SEC501 — Where Big Ideas Live: How to Actually Read Research Papers (https://aws-summit-2026-kb.pages.dev/sessions/SEC501) - WPS202 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS202) --- ## OpenSearch & Vector Search URL: https://aws-summit-2026-kb.pages.dev/topics/opensearch Tagline: Open source search and vector databases at any scale. ### Overview Amazon OpenSearch Service is a managed search and analytics engine forked from Elasticsearch. It powers full-text search, log analytics (the ELK pattern), and — importantly for modern AI — vector search for semantic retrieval and RAG. OpenSearch Serverless eliminates capacity management, while OpenSearch Ingestion provides a managed data pipeline. The OR1 storage option delivers up to 11x better indexing throughput. ### Key concepts - Inverted index (lexical) vs. HNSW (vector) search - Hybrid search — combining BM25 and vector scores - Index lifecycle management and tiered storage - k-NN plugin and quantization for memory efficiency - Observability: ingest pipelines, dashboards, alerting ### Key AWS services - Amazon OpenSearch Service - OpenSearch Serverless - OpenSearch Ingestion ### Curated external resources - [Amazon OpenSearch Service](https://aws.amazon.com/opensearch-service/) - [OpenSearch project (open source)](https://opensearch.org/) - [Vector search in OpenSearch](https://opensearch.org/platform/search/vector-database.html) - [OpenSearch documentation](https://opensearch.org/docs/latest/) ### Live monitored sources (Parallel AI) - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Agentic RAG Explained: AI Agents + RAG in 2026](https://freeacademy.ai/blog/agentic-rag-ai-agents-supercharge-retrieval-2026) — freeacademy.ai (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [The State of AI Agent Memory in 2026: What the Research ...](https://dev.to/vektor_memory_43f51a32376/the-state-of-ai-agent-memory-in-2026-what-the-research-actually-shows-3aja) — dev.to (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [AI Agent Memory Systems Cut Costs 60% with Long-Term Context 2026](https://iterathon.tech/blog/ai-agent-memory-systems-implementation-guide-2026) — iterathon.tech (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys ### Sessions on this topic (5) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - STP205 — How Dovetail powers Multi-Tenant Agents with Vector Indexing at Scale (https://aws-summit-2026-kb.pages.dev/sessions/STP205) - PRT112-S — Empower Data with Oracle AI Database and Native AI Services on AWS (https://aws-summit-2026-kb.pages.dev/sessions/PRT112-S) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - MAE205 — AI at Speed of News: Unlocking Value from Media with Generative AI (https://aws-summit-2026-kb.pages.dev/sessions/MAE205) --- ## Serverless: Lambda & Step Functions URL: https://aws-summit-2026-kb.pages.dev/topics/serverless Tagline: Run code without managing servers — pay only for what you use. ### Overview Serverless computing lets you build apps as small functions that run on demand. AWS Lambda is the original — over 2 million customers use it. Pair Lambda with API Gateway (HTTP), EventBridge (events), SQS/SNS (messaging), and Step Functions (workflows) and you have an entire backend that scales from zero to millions of requests with no servers to patch. SnapStart cuts cold starts to milliseconds for Java, Python, and .NET. ### Key concepts - Event-driven architecture and choreography vs. orchestration - Cold start mitigation: provisioned concurrency, SnapStart - Idempotency, retries, and dead-letter queues - Step Functions: standard vs. express, distributed map - Lambda Powertools — observability and best practices library ### Key AWS services - AWS Lambda - AWS Step Functions - Amazon EventBridge - Amazon API Gateway - Amazon SQS ### Curated external resources - [AWS Lambda](https://aws.amazon.com/lambda/) - [AWS Step Functions](https://aws.amazon.com/step-functions/) - [Serverless Land — patterns and tutorials](https://serverlessland.com/) - [AWS Lambda Powertools](https://docs.powertools.aws.dev/lambda/) - [AWS Compute Blog (serverless)](https://aws.amazon.com/blogs/compute/category/compute/aws-lambda/) ### Live monitored sources (Parallel AI) - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [API7.ai, Original Creator of Apache APISIX | LinkedIn](http://linkedin.com/company/api7-ai) — linkedin.com (2026-05-10): Arcjet introduced 'Guards,' a runtime security service for AI agent workflows that enables enforcement of per-user token budgets and spend limits inside agent loops and can detect prompt injection in tool results. - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [AI Agent Token Budget Enforcement [2026]](https://www.waxell.ai/blog/ai-agent-token-budget-enforcement) — axell.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios ### Sessions on this topic (11) - MAM306 — Adding Agentic AI to legacy apps with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/MAM306) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - DEV312 — Strands Agents on Lambda: Observability With Powertools & X-Ray (https://aws-summit-2026-kb.pages.dev/sessions/DEV312) - DAT301 — Powering your Agentic AI experience with AWS Streaming and Messaging (https://aws-summit-2026-kb.pages.dev/sessions/DAT301) - DEV311 — Serverless Developer Experience: Day in a life of builder (https://aws-summit-2026-kb.pages.dev/sessions/DEV311) - ISV206 — Scaling RAG to Millions of Vectors: The Squiz Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV206) - ARC302 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC302) - ARC403 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC403) - DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV203) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) - INO102 — Partnering for Scale & Innovation (https://aws-summit-2026-kb.pages.dev/sessions/INO102) --- ## Containers: EKS, ECS & Fargate URL: https://aws-summit-2026-kb.pages.dev/topics/containers Tagline: Run containers at scale on Kubernetes or ECS. ### Overview AWS gives you two managed container orchestrators: Amazon EKS (managed Kubernetes, conformant with upstream) and Amazon ECS (AWS-native, simpler). Both run on either EC2 (you manage the nodes) or AWS Fargate (serverless containers — AWS manages the nodes). EKS Auto Mode further abstracts node lifecycle. For images, use Amazon ECR; for service mesh, Amazon VPC Lattice or App Mesh; for scaling, Karpenter (open source, AWS-built). ### Key concepts - Kubernetes core objects: pods, deployments, services, ingress - EKS Auto Mode and managed node groups vs. self-managed - Fargate vs. EC2 — cost, isolation, and operational trade-offs - Karpenter — fast, just-in-time node provisioning - Service mesh and zero-trust networking with VPC Lattice ### Key AWS services - Amazon EKS - Amazon ECS - AWS Fargate - Amazon ECR - Karpenter ### Curated external resources - [Amazon EKS](https://aws.amazon.com/eks/) - [Amazon ECS](https://aws.amazon.com/ecs/) - [AWS Fargate](https://aws.amazon.com/fargate/) - [Karpenter — open source autoscaler](https://karpenter.sh/) - [EKS Best Practices Guide](https://aws.github.io/aws-eks-best-practices/) - [AWS Containers Blog](https://aws.amazon.com/blogs/containers/) ### Live monitored sources (Parallel AI) - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. ### Sessions on this topic (19) - DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/DEV204) - STP210 — TeamForm's Generative Dashboards with Strands & Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/STP210) - PRT217-S — Your Agents Should Be Durable (https://aws-summit-2026-kb.pages.dev/sessions/PRT217-S) - DEV209 — CI/CD Guardrails for Agentic Coding Workflows (https://aws-summit-2026-kb.pages.dev/sessions/DEV209) - ARC303 — Unlock GenAI inference anywhere with Amazon EKS Hybrid Nodes (https://aws-summit-2026-kb.pages.dev/sessions/ARC303) - DEV210 — AI-Driven Incident Triage: From Slack Alert to Root Cause (https://aws-summit-2026-kb.pages.dev/sessions/DEV210) - STP203 — Build, Evaluate and Scale Production ready Agents with AWS Containers (https://aws-summit-2026-kb.pages.dev/sessions/STP203) - ISV201 — MCP on EKS: Xero's AI-Driven Developer Experience (https://aws-summit-2026-kb.pages.dev/sessions/ISV201) - DEV307 — Active-Active Global Architecture with CloudFront and Route 53 (https://aws-summit-2026-kb.pages.dev/sessions/DEV307) - IND202 — How Zuru Uses AI to Analyze TikTok Trends for Rapid Content Creation (https://aws-summit-2026-kb.pages.dev/sessions/IND202) - DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV203) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) - INO103 — Adopting AI-DLC at Scale: How SEEK Is Transforming Product Delivery (https://aws-summit-2026-kb.pages.dev/sessions/INO103) - ISV213 — From GRC Platform to AI-Native Risk Intelligence on AWS:Protecht Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV213) - FSI202 — Accelerating Payment Innovation: Spec-Driven Development with AWS Kiro (https://aws-summit-2026-kb.pages.dev/sessions/FSI202) - ISV207 — How Canva Scales and Optimizes AI Workloads with Karpenter (https://aws-summit-2026-kb.pages.dev/sessions/ISV207) - MAE204 — How Amazon Ads Creative Agent uses AWS to democratize ad creation (https://aws-summit-2026-kb.pages.dev/sessions/MAE204) - FSI203 — How HBF Transformed Claims Processing From Two Weeks to Two Minutes (https://aws-summit-2026-kb.pages.dev/sessions/FSI203) --- ## Compute: EC2, Graviton & Nitro URL: https://aws-summit-2026-kb.pages.dev/topics/compute Tagline: The world's broadest selection of compute — including AWS-designed silicon. ### Overview Amazon EC2 offers 850+ instance types across general purpose, compute-optimized, memory-optimized, storage-optimized, and accelerated computing (GPU, Trainium, Inferentia). The AWS Nitro System is the underlying hardware/software platform that delivers near-bare-metal performance with hardware-accelerated networking and storage. AWS Graviton (Arm-based) processors deliver up to 40% better price-performance than comparable x86 instances and now power most new EC2 capacity. ### Key concepts - Instance families and right-sizing for workload - AWS Graviton — Arm advantage for web tier, databases, ML inference - Nitro System: Nitro Cards, Nitro Security Chip, Nitro Hypervisor - Spot, Reserved, Savings Plans, On-Demand pricing - EC2 Capacity Blocks for ML — reserve GPU capacity in advance ### Key AWS services - Amazon EC2 - AWS Graviton - AWS Nitro System - EC2 Auto Scaling - AWS Outposts ### Curated external resources - [Amazon EC2](https://aws.amazon.com/ec2/) - [AWS Graviton processors](https://aws.amazon.com/ec2/graviton/) - [AWS Nitro System](https://aws.amazon.com/ec2/nitro/) - [EC2 instance type explorer](https://aws.amazon.com/ec2/instance-types/) - [AWS Compute Blog](https://aws.amazon.com/blogs/compute/) ### Sessions on this topic (14) - MAM302 — Agentic AI for VMware migrations with AWS Transform for VMware (https://aws-summit-2026-kb.pages.dev/sessions/MAM302) - TNC202 — Accelerate Your Cloud Journey with AWS Transform (https://aws-summit-2026-kb.pages.dev/sessions/TNC202) - COP302 — Applying AI for FinOps and FinOps for AI (https://aws-summit-2026-kb.pages.dev/sessions/COP302) - ISV205 — AWS Graviton: The best price performance for your AWS workloads (https://aws-summit-2026-kb.pages.dev/sessions/ISV205) - STP302 — Unleash Live: Cloud-Powered Vision for Infrastructure (https://aws-summit-2026-kb.pages.dev/sessions/STP302) - CMP501 — Nitro Isolation Engine: Formally Verifying Confidentiality (https://aws-summit-2026-kb.pages.dev/sessions/CMP501) - ARC302 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC302) - ARC403 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC403) - DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV203) - ISV203 — AI Monetization and Pricing Strategies (https://aws-summit-2026-kb.pages.dev/sessions/ISV203) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) - ISV207 — How Canva Scales and Optimizes AI Workloads with Karpenter (https://aws-summit-2026-kb.pages.dev/sessions/ISV207) - SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR (https://aws-summit-2026-kb.pages.dev/sessions/SMB203) --- ## Storage: S3, EBS & EFS URL: https://aws-summit-2026-kb.pages.dev/topics/storage Tagline: Industry-leading object, block, and file storage. ### Overview Amazon S3 is the foundational object store for the cloud — 11 nines of durability, virtually unlimited scale, and the basis for data lakes, backups, and content delivery. S3 Express One Zone delivers single-digit-millisecond latency for high-performance workloads. Amazon EBS provides block storage for EC2 (gp3 is the modern default). Amazon EFS provides shared NFS, and Amazon FSx covers Windows, Lustre (HPC/ML), NetApp ONTAP, and OpenZFS file systems. ### Key concepts - S3 storage classes: Standard, Intelligent-Tiering, Glacier tiers - S3 Tables — Iceberg-native bucket type for analytics - S3 Vectors — native vector storage for AI workloads - EBS volume types: gp3, io2 Block Express, st1, sc1 - Backup, replication, and lifecycle policies ### Key AWS services - Amazon S3 - Amazon EBS - Amazon EFS - Amazon FSx - AWS Backup ### Curated external resources - [Amazon S3](https://aws.amazon.com/s3/) - [S3 Tables](https://aws.amazon.com/s3/features/tables/) - [Amazon EBS](https://aws.amazon.com/ebs/) - [AWS Storage Blog](https://aws.amazon.com/blogs/storage/) - [AWS Storage learning path](https://aws.amazon.com/training/learn-about/storage/) ### Sessions on this topic (5) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI (https://aws-summit-2026-kb.pages.dev/sessions/STP208) - DAT401 — Real-Time DataLakes with Apache Iceberg, Amazon MSK, and Amazon S3 (https://aws-summit-2026-kb.pages.dev/sessions/DAT401) - ISV206 — Scaling RAG to Millions of Vectors: The Squiz Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV206) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) --- ## Networking & Edge URL: https://aws-summit-2026-kb.pages.dev/topics/networking Tagline: Build the network: VPC, CloudFront, Route 53, Transit Gateway. ### Overview AWS networking starts with Amazon VPC — your own virtual network with subnets, route tables, security groups, and gateways. AWS Transit Gateway connects thousands of VPCs and on-prem networks. AWS Cloud WAN gives a global, policy-driven backbone. At the edge, Amazon CloudFront (450+ POPs) delivers content with CloudFront Functions and Lambda@Edge for compute. Amazon VPC Lattice provides application-layer service-to-service connectivity across accounts and VPCs without complex peering. ### Key concepts - VPC design: subnetting, public vs. private, NAT, gateway types - Hybrid: Direct Connect, Site-to-Site VPN, Cloud WAN - CloudFront, AWS Global Accelerator, Route 53 routing policies - Service connectivity: PrivateLink, VPC Lattice, App Mesh - AWS WAF, Shield, and Network Firewall for security at the edge ### Key AWS services - Amazon VPC - Amazon CloudFront - AWS Transit Gateway - AWS Cloud WAN - Amazon VPC Lattice - AWS PrivateLink ### Curated external resources - [Amazon VPC](https://aws.amazon.com/vpc/) - [Amazon CloudFront](https://aws.amazon.com/cloudfront/) - [AWS Networking & Content Delivery Blog](https://aws.amazon.com/blogs/networking-and-content-delivery/) - [AWS Networking Workshop](https://catalog.workshops.aws/networking/) ### Sessions on this topic (5) - DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/DEV204) - ARC401 — The Art of Managing Trade-Offs for your Network Design with Megaport (https://aws-summit-2026-kb.pages.dev/sessions/ARC401) - DEV307 — Active-Active Global Architecture with CloudFront and Route 53 (https://aws-summit-2026-kb.pages.dev/sessions/DEV307) - IDE102 — Power of Possibility: Leading Through Innovation and Connection (https://aws-summit-2026-kb.pages.dev/sessions/IDE102) - ISV204 — AWS Networking Fundamentals: Connect, secure and scale (https://aws-summit-2026-kb.pages.dev/sessions/ISV204) --- ## Security, Identity & Compliance URL: https://aws-summit-2026-kb.pages.dev/topics/security Tagline: Security is job zero — shared responsibility, defense in depth. ### Overview AWS provides a comprehensive security toolkit: IAM and IAM Identity Center for identity and access; AWS KMS and CloudHSM for encryption; AWS WAF, Shield, and Network Firewall for network protection; Amazon GuardDuty and Inspector for threat detection; AWS Security Hub for posture management; AWS Config and CloudTrail for audit. The AWS Security Reference Architecture (SRA) gives you a prescriptive multi-account blueprint, and the AWS Well-Architected Security Pillar codifies best practices. ### Key concepts - Shared responsibility model: AWS secures the cloud; you secure in the cloud - Least privilege with IAM policies, SCPs, and permissions boundaries - Encryption: in transit (TLS), at rest (KMS, BYOK), envelope encryption - Threat detection: GuardDuty, Detective, Macie - Zero-trust patterns and AI security (red-teaming, prompt injection defense) ### Key AWS services - AWS IAM - AWS IAM Identity Center - AWS KMS - Amazon GuardDuty - AWS Security Hub - AWS Config ### Curated external resources - [AWS Cloud Security](https://aws.amazon.com/security/) - [AWS Security Reference Architecture](https://docs.aws.amazon.com/prescriptive-guidance/latest/security-reference-architecture/welcome.html) - [AWS Security Blog](https://aws.amazon.com/blogs/security/) - [OWASP Top 10 for LLM Applications](https://owasp.org/www-project-top-10-for-large-language-model-applications/) - [AWS Well-Architected Security Pillar](https://docs.aws.amazon.com/wellarchitected/latest/security-pillar/welcome.html) ### Live monitored sources (Parallel AI) - [What’s New in Agent 365: May 2026 | Microsoft Community Hub](https://techcommunity.microsoft.com/blog/agent-365-blog/what%E2%80%99s-new-in-agent-365-may-2026/4516340) — techcommunity.microsoft.com (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [MCP Governance (2026): Policy Gates for MCP Servers](https://cordum.io/blog/mcp-governance-servers) — cordum.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [MCP Security Gateway - Agent Governance Toolkit](https://microsoft.github.io/agent-governance-toolkit/tutorials/07-mcp-security-gateway/) — microsoft.github.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [gdplabs/gl-iam-cookbook · GitHub - GitHub](http://github.com/gdplabs/gl-iam-cookbook) — github.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho ### Sessions on this topic (65) - PRT109-S — Hello Future, Meet Reality: Enterprise AI Lessons (https://aws-summit-2026-kb.pages.dev/sessions/PRT109-S) - PRT215-S — The Visibility Gap: Turning Observability into DevSecOps Signals (https://aws-summit-2026-kb.pages.dev/sessions/PRT215-S) - PRT104-S — Building Resilience for AI Data Foundations and Cloud-Native Apps 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT104-S) - PRT202-S — 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT202-S) - PRT204-S — Optimising GenAI at Runtime with Experimentation and Guardrails 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT204-S) - AIM201 — From demo to deployment: solving agentic AI's toughest challenges (https://aws-summit-2026-kb.pages.dev/sessions/AIM201) - MAM306 — Adding Agentic AI to legacy apps with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/MAM306) - DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/DEV204) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - ISV208 — From One Month to Two Days: How Xero Transformed Their DLC with AI (https://aws-summit-2026-kb.pages.dev/sessions/ISV208) - AIM303 — AWS Security Agent: Proactive AppSec from Design to Deployment (https://aws-summit-2026-kb.pages.dev/sessions/AIM303) - DEV205 — Securing Amazon Bedrock AgentCore: A Practical Framework (https://aws-summit-2026-kb.pages.dev/sessions/DEV205) - STP201 — Scaling Security at Startup Speed: Hnry's AI-Powered Approach (https://aws-summit-2026-kb.pages.dev/sessions/STP201) - PRT213-S — How NAB is Conquering Multi-Cloud to Secure the Enterprise (https://aws-summit-2026-kb.pages.dev/sessions/PRT213-S) - PRT301-S — Unite Teams, Tools, and AI to Drive Transformation at Scale (https://aws-summit-2026-kb.pages.dev/sessions/PRT301-S) - PRT103-S — Cloud Anywhere: Architectural Freedom for Unified Data and AI 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT103-S) - AIM204 — Get to know Amazon Quick, your new agentic teammate (https://aws-summit-2026-kb.pages.dev/sessions/AIM204) - PRT101-S — Accelerating Innovation with GitLab DAP Powered by Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/PRT101-S) - PRT203-S — Design, Deploy, and Govern AI Agents with Boomis Agentstudio 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT203-S) - SEC305 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC305) - ISV301 — Rolling to Scale: Roller's Multi-Tenant SaaS platform on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV301) - STP211 — Authenticating AI Agents: How Kinde Navigates Agentic Identity (https://aws-summit-2026-kb.pages.dev/sessions/STP211) - AIM203 — Prompt Engineering to Learning Systems: Woodside's Agentic Ecosystem (https://aws-summit-2026-kb.pages.dev/sessions/AIM203) - ARC304 — Demystifying Agent Identity (https://aws-summit-2026-kb.pages.dev/sessions/ARC304) - DEV201 — How Flybuys Built AI Governance to Accelerate Adoption at Scale (https://aws-summit-2026-kb.pages.dev/sessions/DEV201) - DEV305 — Agents in the enterprise: Best practices with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/DEV305) - ISV205 — AWS Graviton: The best price performance for your AWS workloads (https://aws-summit-2026-kb.pages.dev/sessions/ISV205) - STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI (https://aws-summit-2026-kb.pages.dev/sessions/STP208) - STP302 — Unleash Live: Cloud-Powered Vision for Infrastructure (https://aws-summit-2026-kb.pages.dev/sessions/STP302) - SEC401 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC401) - ARC401 — The Art of Managing Trade-Offs for your Network Design with Megaport (https://aws-summit-2026-kb.pages.dev/sessions/ARC401) - ARC302 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC302) - ARC307 — AI Powered Resilience Lifecycle (https://aws-summit-2026-kb.pages.dev/sessions/ARC307) - ARC402 — DynamoDB: Resilience & lessons from the Oct 2025 service disruption (https://aws-summit-2026-kb.pages.dev/sessions/ARC402) - ARC403 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive (https://aws-summit-2026-kb.pages.dev/sessions/ARC403) - DAT303 — Explore whats new in data and AI governance with SageMaker Catalog (https://aws-summit-2026-kb.pages.dev/sessions/DAT303) - FSI206 — Agentic AI Transforming Quality at Cloud Speed (https://aws-summit-2026-kb.pages.dev/sessions/FSI206) - PRT106-S — The AI Challenge You Don't Yet Know About - Software Supply Chain (https://aws-summit-2026-kb.pages.dev/sessions/PRT106-S) - PRT111-S — From Risk to Resilience - How Mimecast Works with AWS (https://aws-summit-2026-kb.pages.dev/sessions/PRT111-S) - ISV202 — Architecting for growth and resilience: Cell based design deep dive (https://aws-summit-2026-kb.pages.dev/sessions/ISV202) - ISV211 — Scaling Conversation Intelligence with Agentic AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV211) - STP204 — How Heidi Health Fine-Tunes Speech-to-Text Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/STP204) - ISV209 — From dev tools to customer value: BGL's agentic AI journey (https://aws-summit-2026-kb.pages.dev/sessions/ISV209) - FSI204 — Agentic AI in Financial Services: Architectural Patterns That Work (https://aws-summit-2026-kb.pages.dev/sessions/FSI204) - IND204 — How Transurban Transformed Customer Experience with AI Agents on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND204) - DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV203) - ISV103 — Working With AI: Lessons They Don't Put in the Demo (https://aws-summit-2026-kb.pages.dev/sessions/ISV103) - STP301 — AI-Native Remediation with Pleri: Your Security Engineer That Ships (https://aws-summit-2026-kb.pages.dev/sessions/STP301) - IDE101 — From principles to practice: Scaling AI responsibly (https://aws-summit-2026-kb.pages.dev/sessions/IDE101) - DEV308 — AI Blast-Radius Reviews for AWS Changes Using Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/DEV308) - IDE102 — Power of Possibility: Leading Through Innovation and Connection (https://aws-summit-2026-kb.pages.dev/sessions/IDE102) - IND301 — Stockland Empowers People with a GenAI Assistant Built on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND301) - TNC204 — Exam Prep: AWS Solutions Architect Associate (https://aws-summit-2026-kb.pages.dev/sessions/TNC204) - DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops (https://aws-summit-2026-kb.pages.dev/sessions/DEV206) - BIZ201 — AI-Everywhere: Transform Customer Interactions into Memorable Moments (https://aws-summit-2026-kb.pages.dev/sessions/BIZ201) - INO101 — From Zero to 270 AI Agents: how Lendi built Guardian (https://aws-summit-2026-kb.pages.dev/sessions/INO101) - DEV309 — AI Outputs: Amazon Bedrock Structured Output in Production (https://aws-summit-2026-kb.pages.dev/sessions/DEV309) - IND206 — How scalable data foundations helped TGE unlock the power of AI (https://aws-summit-2026-kb.pages.dev/sessions/IND206) - WPS302 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS302) - DEV208 — Production-Grade Platforms: Real-World IaC Practices on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV208) - ISV213 — From GRC Platform to AI-Native Risk Intelligence on AWS:Protecht Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV213) - FSI202 — Accelerating Payment Innovation: Spec-Driven Development with AWS Kiro (https://aws-summit-2026-kb.pages.dev/sessions/FSI202) - SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR (https://aws-summit-2026-kb.pages.dev/sessions/SMB203) - SEC501 — Where Big Ideas Live: How to Actually Read Research Papers (https://aws-summit-2026-kb.pages.dev/sessions/SEC501) - WPS202 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS202) --- ## DevOps, CI/CD & DevSecOps URL: https://aws-summit-2026-kb.pages.dev/topics/devops Tagline: Ship faster, safer, with automation end to end. ### Overview DevOps on AWS combines source control (CodeCommit / GitHub), build (CodeBuild), deploy (CodeDeploy, CodePipeline, AWS Proton), infrastructure as code (CloudFormation, AWS CDK, Terraform), and observability (CloudWatch, X-Ray). DevSecOps shifts security left: code scanning with Amazon Inspector or CodeGuru Security, IaC scanning, secret detection, and runtime protection. Modern teams treat policies as code with cfn-guard, OPA, or Checkov. ### Key concepts - Trunk-based development and continuous delivery - Infrastructure as Code: CloudFormation, AWS CDK, Terraform, Pulumi - Blue/green and canary deployments - Security in the pipeline: SAST, DAST, SCA, IaC scanning - GitOps with Flux/Argo on EKS ### Key AWS services - AWS CodePipeline - AWS CodeBuild - AWS CodeDeploy - AWS CDK - AWS CloudFormation - Amazon Inspector ### Curated external resources - [AWS DevOps overview](https://aws.amazon.com/devops/) - [AWS Cloud Development Kit (CDK)](https://aws.amazon.com/cdk/) - [DevOps Research and Assessment (DORA)](https://dora.dev/) - [AWS DevOps Blog](https://aws.amazon.com/blogs/devops/) - [AWS Well-Architected DevOps Guidance](https://docs.aws.amazon.com/wellarchitected/latest/devops-guidance/devops-guidance.html) ### Live monitored sources (Parallel AI) - [GitHub - Siddhant-K-code/agent-trace: strace for AI agents. Capture and replay every tool call, prompt, and response from Claude Code, Cursor, Gemini CLI or any MCP client · GitHub](https://github.com/Siddhant-K-code/agent-trace) — github.com (2026-05-04): The 'agent-trace' developer tool (GitHub: Siddhant-K-code/agent-trace) has launched significant new monitoring and control features: 1) A 'watch' mode that automatically terminates agents (using SIGSTOP or SIGTERM) when specific rules in a .watch-rules.json file are triggered, su - [Releases · microsoft/agent-framework · GitHub](https://github.com/microsoft/agent-framework/releases) — github.com (2026-05-08): Microsoft Agent Framework released version dotnet-1.5.0, introducing a significant shift to standardize on the Model Context Protocol (MCP) for toolbox consumption. The update also adds hosted agent observability samples and implements message filtering for non-portable content t - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [Announcing Agent Control: The Open Source Control Plane for ...](https://galileo.ai/blog/announcing-agent-control) — galileo.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca ### Sessions on this topic (8) - PRT215-S — The Visibility Gap: Turning Observability into DevSecOps Signals (https://aws-summit-2026-kb.pages.dev/sessions/PRT215-S) - AIM201 — From demo to deployment: solving agentic AI's toughest challenges (https://aws-summit-2026-kb.pages.dev/sessions/AIM201) - DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/DEV204) - DEV209 — CI/CD Guardrails for Agentic Coding Workflows (https://aws-summit-2026-kb.pages.dev/sessions/DEV209) - AIM301 — Commbank pioneering AI-driven DevSecOps with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/AIM301) - SEC302 — Leap ahead in Cloud Operations with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/SEC302) - PRT106-S — The AI Challenge You Don't Yet Know About - Software Supply Chain (https://aws-summit-2026-kb.pages.dev/sessions/PRT106-S) - DEV208 — Production-Grade Platforms: Real-World IaC Practices on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV208) --- ## Observability & Monitoring URL: https://aws-summit-2026-kb.pages.dev/topics/observability Tagline: Logs, metrics, traces — and now AI-powered insights. ### Overview Observability is more than monitoring: it's the ability to ask new questions of your system without shipping new code. AWS's native stack includes Amazon CloudWatch (metrics, logs, dashboards, alarms), AWS X-Ray (distributed tracing), Amazon Managed Service for Prometheus, and Amazon Managed Grafana. CloudWatch Application Signals automatically discovers services and shows golden signals. AI-driven tools like Amazon Q in CloudWatch and partner solutions (Datadog, New Relic, Dynatrace, Splunk) help triage incidents. ### Key concepts - Three pillars: logs, metrics, traces — plus events and profiles - OpenTelemetry as the vendor-neutral standard - SLOs, SLIs, error budgets — the SRE language - Anomaly detection and automated root-cause analysis - Cost-aware observability — sampling, retention, log-class tiers ### Key AWS services - Amazon CloudWatch - AWS X-Ray - Amazon Managed Service for Prometheus - Amazon Managed Grafana - AWS Distro for OpenTelemetry ### Curated external resources - [Amazon CloudWatch](https://aws.amazon.com/cloudwatch/) - [AWS X-Ray](https://aws.amazon.com/xray/) - [OpenTelemetry — open source standard](https://opentelemetry.io/) - [Google SRE Books (free)](https://sre.google/books/) - [AWS Observability Best Practices](https://aws-observability.github.io/observability-best-practices/) ### Live monitored sources (Parallel AI) - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [TraceRoot](http://traceroot.ai/) — traceroot.ai (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [Introducing the Agent Command Center and Devin in ...](http://windsurf.com/blog/windsurf-2-0) — indsurf.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [Edge Delta Makes All Telemetry Pipelines Data ...](http://prnewswire.com/news-releases/edge-delta-makes-all-telemetry-pipelines-data-throughput-limitless-and-free-for-all-customers-302736808.html) — prnewswire.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [grafana.com](https://grafana.com/press/2026/04/21/grafana-labs-targets-the-ai-blind-spot-with-new-observability-tools-announced-at-grafanacon-2026/) — grafana.com (2026-04-25): Grafana Labs announced several AI-focused observability and agent tools on April 21, 2026: 1) AI Observability in Grafana Cloud for real-time monitoring of agent inputs, outputs, and execution flows; 2) Expanded Grafana Assistant with a new API, Automations, and Remote MCP server ### Sessions on this topic (18) - PRT215-S — The Visibility Gap: Turning Observability into DevSecOps Signals (https://aws-summit-2026-kb.pages.dev/sessions/PRT215-S) - AIM201 — From demo to deployment: solving agentic AI's toughest challenges (https://aws-summit-2026-kb.pages.dev/sessions/AIM201) - DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/DEV204) - ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/ISV302) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - COP301 — Elevating your Agentic AI Observability (https://aws-summit-2026-kb.pages.dev/sessions/COP301) - ISV210 — Boost performance and reduce costs with Aurora: Canva's story (https://aws-summit-2026-kb.pages.dev/sessions/ISV210) - PRT209-S — How Auto & General leverage observability foundations for AI (https://aws-summit-2026-kb.pages.dev/sessions/PRT209-S) - PRT101-S — Accelerating Innovation with GitLab DAP Powered by Amazon Bedrock (https://aws-summit-2026-kb.pages.dev/sessions/PRT101-S) - PRT203-S — Design, Deploy, and Govern AI Agents with Boomis Agentstudio 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT203-S) - DEV312 — Strands Agents on Lambda: Observability With Powertools & X-Ray (https://aws-summit-2026-kb.pages.dev/sessions/DEV312) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - DEV305 — Agents in the enterprise: Best practices with Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/DEV305) - SEC302 — Leap ahead in Cloud Operations with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/SEC302) - DEV210 — AI-Driven Incident Triage: From Slack Alert to Root Cause (https://aws-summit-2026-kb.pages.dev/sessions/DEV210) - WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health (https://aws-summit-2026-kb.pages.dev/sessions/WPS203) - DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops (https://aws-summit-2026-kb.pages.dev/sessions/DEV206) - SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR (https://aws-summit-2026-kb.pages.dev/sessions/SMB203) --- ## Cost Optimization & FinOps URL: https://aws-summit-2026-kb.pages.dev/topics/cost Tagline: Build a culture of accountability for cloud spend. ### Overview FinOps is the operational practice of bringing financial accountability to the variable spend model of cloud. AWS provides AWS Cost Explorer, AWS Budgets, AWS Cost & Usage Reports (CUR), and AWS Compute Optimizer to find waste. Savings Plans and Reserved Instances commit to usage for steep discounts. Tagging strategy and account structure (AWS Organizations, AWS Control Tower) are the foundation. The FinOps Foundation defines a phased maturity model: Inform → Optimize → Operate. ### Key concepts - Right-sizing, scheduling, and Spot for variable workloads - Savings Plans (Compute, EC2 Instance, SageMaker) vs. RIs - Tagging strategy and showback/chargeback - Architecting for cost: serverless, Graviton, S3 storage classes - Cost anomaly detection and forecasting ### Key AWS services - AWS Cost Explorer - AWS Budgets - AWS Compute Optimizer - Savings Plans - AWS Trusted Advisor ### Curated external resources - [AWS Cloud Financial Management](https://aws.amazon.com/aws-cost-management/) - [FinOps Foundation](https://www.finops.org/) - [AWS Well-Architected Cost Pillar](https://docs.aws.amazon.com/wellarchitected/latest/cost-optimization-pillar/welcome.html) - [AWS Cost Optimization Workshop](https://catalog.workshops.aws/well-architected-cost-optimization/) ### Live monitored sources (Parallel AI) - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [Talent Harbor | Sales Recruitment as a Service (RaaS)](http://talentharbor.com/) — talentharbor.com (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [Outcome-based pricing for AI Agents - Sierra](http://sierra.ai/blog/outcome-based-pricing-for-ai-agents) — sierra.ai (2026-05-11): Sierra announced an outcome-based pricing model for its AI agents, ensuring that the company is only paid when its AI agents drive real, tangible results for the business, aligning cost directly with success. - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-moves-beyond-the-sidecar-AI-era-giving-customers-a-complete-AI-native-experience-across-all-products-and-packages/default.aspx) — newsroom.servicenow.com (2026-04-09): ServiceNow announced on 2026-04-09 that its platform is now AI-native with a new tiered offer model spanning AI assistance, agentic automation, and fully autonomous operations. Pricing approach: introduces tiered packaging (including ESM Foundation for midsize customers); no indi - [globenewswire.com](https://www.globenewswire.com/news-release/2026/04/06/3268529/0/en/Yuno-Launches-Payments-Concierge-An-Always-On-AI-Agent-for-Payment-Operations.html) — globenewswire.com (2026-04-06): Yuno announced Payments Concierge (Apr 6, 2026), an always-on autonomous payments agent that monitors, detects anomalies, and takes configured actions (adjust routing, enable/disable providers) to optimize acceptance, costs, and conversion. Pricing approach: the announcement emph ### Sessions on this topic (3) - AIM201 — From demo to deployment: solving agentic AI's toughest challenges (https://aws-summit-2026-kb.pages.dev/sessions/AIM201) - COP302 — Applying AI for FinOps and FinOps for AI (https://aws-summit-2026-kb.pages.dev/sessions/COP302) - DAT401 — Real-Time DataLakes with Apache Iceberg, Amazon MSK, and Amazon S3 (https://aws-summit-2026-kb.pages.dev/sessions/DAT401) --- ## Migration & Modernization URL: https://aws-summit-2026-kb.pages.dev/topics/migration Tagline: Move workloads to AWS — and modernize them along the way. ### Overview AWS Migration Hub orchestrates large migrations. AWS Application Migration Service (MGN) lift-and-shifts servers; AWS Database Migration Service (DMS) handles databases. AWS Transform is the new agentic AI service for VMware migrations and .NET / mainframe modernization. The 7 R's framework — Retire, Retain, Relocate, Rehost, Replatform, Repurchase, Refactor — guides each application's strategy. ### Key concepts - The 7 R's migration strategies - Migration Acceleration Program (MAP) funding and methodology - Application discovery and dependency mapping - Database migration patterns: homogeneous vs. heterogeneous - Mainframe and VMware modernization with agentic AI ### Key AWS services - AWS Migration Hub - AWS Application Migration Service - AWS Database Migration Service - AWS Transform - AWS Schema Conversion Tool ### Curated external resources - [AWS Cloud Migration](https://aws.amazon.com/cloud-migration/) - [AWS Transform](https://aws.amazon.com/transform/) - [The 7 R's of Cloud Migration](https://docs.aws.amazon.com/prescriptive-guidance/latest/strategy-migration/welcome.html) - [AWS Migration Blog](https://aws.amazon.com/blogs/migration-and-modernization/) ### Sessions on this topic (12) - MAM302 — Agentic AI for VMware migrations with AWS Transform for VMware (https://aws-summit-2026-kb.pages.dev/sessions/MAM302) - MAM301 — From tech debt to competitive advantage: Migrate & modernize with AWS (https://aws-summit-2026-kb.pages.dev/sessions/MAM301) - MAM303 — Digital transformation excellence using agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/MAM303) - ISV210 — Boost performance and reduce costs with Aurora: Canva's story (https://aws-summit-2026-kb.pages.dev/sessions/ISV210) - MAM307 — Modernise legacy code using fine-tuned Gen AI models (https://aws-summit-2026-kb.pages.dev/sessions/MAM307) - TNC202 — Accelerate Your Cloud Journey with AWS Transform (https://aws-summit-2026-kb.pages.dev/sessions/TNC202) - ISV205 — AWS Graviton: The best price performance for your AWS workloads (https://aws-summit-2026-kb.pages.dev/sessions/ISV205) - MAM305 — Legacy App modernization and reverse engineering using Kiro (https://aws-summit-2026-kb.pages.dev/sessions/MAM305) - FSI201 — BELIEVE: The Impossible Migration That Transformed Australian Banking (https://aws-summit-2026-kb.pages.dev/sessions/FSI201) - WPS204 — Safe Transport Victoria's Migration to AWS Cloud (https://aws-summit-2026-kb.pages.dev/sessions/WPS204) - DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) (https://aws-summit-2026-kb.pages.dev/sessions/DEV310) - INO102 — Partnering for Scale & Innovation (https://aws-summit-2026-kb.pages.dev/sessions/INO102) --- ## IoT & Edge Computing URL: https://aws-summit-2026-kb.pages.dev/topics/iot Tagline: Connect, secure, and analyze billions of devices. ### Overview AWS IoT spans device software (FreeRTOS, AWS IoT Greengrass), connectivity (AWS IoT Core supports MQTT, MQTT-over-WebSockets, HTTPS), data services (IoT SiteWise for industrial, IoT TwinMaker for digital twins, IoT FleetWise for vehicles), and analytics. For edge compute, AWS Outposts brings AWS to your data center, AWS Wavelength brings it to 5G networks, and Local Zones bring it to metro areas. ### Key concepts - MQTT and pub/sub messaging - Device shadows and desired vs. reported state - Greengrass — Lambda and ML at the edge - Digital twins and industrial data models - OTA firmware updates and fleet management ### Key AWS services - AWS IoT Core - AWS IoT Greengrass - AWS IoT SiteWise - AWS IoT TwinMaker - AWS Outposts ### Curated external resources - [AWS IoT](https://aws.amazon.com/iot/) - [AWS IoT Greengrass](https://aws.amazon.com/greengrass/) - [AWS Outposts](https://aws.amazon.com/outposts/) - [AWS IoT Blog](https://aws.amazon.com/blogs/iot/) ### Sessions on this topic (1) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) --- ## Sustainability & Green IT URL: https://aws-summit-2026-kb.pages.dev/topics/sustainability Tagline: Build with the planet in mind. ### Overview AWS aims to power operations with 100% renewable energy and is on a path to be water positive by 2030. The AWS Customer Carbon Footprint Tool reports your emissions in tCO2e. The AWS Well-Architected Sustainability Pillar gives you design principles to reduce the environmental impact of your workloads — choose efficient regions, right-size, use Graviton, choose serverless, and optimize data storage and transfer. ### Key concepts - Cloud vs. on-prem carbon — the efficiency dividend - Region selection by carbon intensity - Architectural patterns for efficiency (Graviton, serverless, async) - Measuring and reporting Scope 1/2/3 emissions - Sustainable data: lifecycle, tiering, deduplication ### Key AWS services - AWS Customer Carbon Footprint Tool - AWS Graviton - AWS Well-Architected Sustainability Pillar ### Curated external resources - [AWS Sustainability](https://sustainability.aboutamazon.com/products-services/aws-cloud) - [AWS Well-Architected Sustainability Pillar](https://docs.aws.amazon.com/wellarchitected/latest/sustainability-pillar/sustainability-pillar.html) - [Green Software Foundation](https://greensoftware.foundation/) --- ## Resilience & Disaster Recovery URL: https://aws-summit-2026-kb.pages.dev/topics/resilience Tagline: Design for failure — because everything fails, all the time. ### Overview Resilience is the ability of a workload to recover from failures. AWS offers AWS Resilience Hub (continuous resilience assessment), AWS Backup (centralized backup), AWS Elastic Disaster Recovery (DRS — sub-minute RPO), and AWS Fault Injection Service (chaos engineering). The four common DR strategies are Backup & Restore, Pilot Light, Warm Standby, and Multi-Site Active/Active — trading cost for RTO/RPO. Multi-region active-active with Aurora DSQL or DynamoDB global tables is now feasible for many applications. ### Key concepts - RTO (Recovery Time Objective) and RPO (Recovery Point Objective) - Four DR strategies and the cost/RTO trade-off - Chaos engineering and game days - Multi-AZ vs. multi-region — what each protects against - Cell-based architectures and bulkheads ### Key AWS services - AWS Resilience Hub - AWS Backup - AWS Elastic Disaster Recovery - AWS Fault Injection Service ### Curated external resources - [AWS Resilience Hub](https://aws.amazon.com/resilience-hub/) - [AWS Well-Architected Reliability Pillar](https://docs.aws.amazon.com/wellarchitected/latest/reliability-pillar/welcome.html) - [Disaster Recovery of Workloads on AWS — whitepaper](https://docs.aws.amazon.com/whitepapers/latest/disaster-recovery-workloads-on-aws/disaster-recovery-workloads-on-aws.html) - [Principles of Chaos Engineering](https://principlesofchaos.org/) ### Live monitored sources (Parallel AI) - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AI Agent Rate Limiting Strategies & Best Practices](https://fast.io/resources/ai-agent-rate-limiting/) — fast.io (2026-05-10): Arcjet introduced 'Guards,' a runtime security service for AI agent workflows that enables enforcement of per-user token budgets and spend limits inside agent loops and can detect prompt injection in tool results. - [Circuit Breakers for AI Agents: How We Stop Cascading ...](https://cencori.com/blog/circuit-breakers-for-ai-agents) — cencori.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [Multi-Agent Orchestration Patterns Drive Enterprise ROI in 2026](https://insights.reinventing.ai/articles/ai-agents-orchestration-patterns-2026-03-18) — insights.reinventing.ai (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. ### Sessions on this topic (11) - PRT104-S — Building Resilience for AI Data Foundations and Cloud-Native Apps 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT104-S) - PRT207-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform (https://aws-summit-2026-kb.pages.dev/sessions/PRT207-S) - AIM301 — Commbank pioneering AI-driven DevSecOps with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/AIM301) - ARC201 — Building on AWS resilience: Innovations for critical success (https://aws-summit-2026-kb.pages.dev/sessions/ARC201) - ARC307 — AI Powered Resilience Lifecycle (https://aws-summit-2026-kb.pages.dev/sessions/ARC307) - ARC402 — DynamoDB: Resilience & lessons from the Oct 2025 service disruption (https://aws-summit-2026-kb.pages.dev/sessions/ARC402) - PRT111-S — From Risk to Resilience - How Mimecast Works with AWS (https://aws-summit-2026-kb.pages.dev/sessions/PRT111-S) - FSI201 — BELIEVE: The Impossible Migration That Transformed Australian Banking (https://aws-summit-2026-kb.pages.dev/sessions/FSI201) - ISV202 — Architecting for growth and resilience: Cell based design deep dive (https://aws-summit-2026-kb.pages.dev/sessions/ISV202) - WPS204 — Safe Transport Victoria's Migration to AWS Cloud (https://aws-summit-2026-kb.pages.dev/sessions/WPS204) - WPS302 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS302) --- ## Voice & Conversational AI URL: https://aws-summit-2026-kb.pages.dev/topics/voice-ai Tagline: Build natural voice and chat experiences for customers. ### Overview Amazon Connect is AWS's cloud contact center — set up an omnichannel center in minutes with built-in voice, chat, tasks, email, and now AI. Amazon Q in Connect gives agents real-time, generative-AI assistance from your knowledge base. Amazon Lex builds chatbots and voicebots, while Amazon Polly does text-to-speech (with neural and generative voices) and Amazon Transcribe does speech-to-text. New voice-to-voice models from Amazon and partners enable natural, low-latency conversations. ### Key concepts - Speech-to-text (ASR), TTS, and end-to-end voice models - Intent classification, slot filling, and dialog management - Real-time agent assist and post-call analytics - Sentiment and quality management at scale - Voice biometrics and fraud detection ### Key AWS services - Amazon Connect - Amazon Q in Connect - Amazon Lex - Amazon Polly - Amazon Transcribe ### Curated external resources - [Amazon Connect](https://aws.amazon.com/connect/) - [Amazon Q in Connect](https://aws.amazon.com/connect/q/) - [Amazon Lex](https://aws.amazon.com/lex/) - [AWS Contact Center Blog](https://aws.amazon.com/blogs/contact-center/) ### Sessions on this topic (27) - COP301 — Elevating your Agentic AI Observability (https://aws-summit-2026-kb.pages.dev/sessions/COP301) - ISV303 — From hours to minutes: SafetyCulture's journey to 90% faster analytics (https://aws-summit-2026-kb.pages.dev/sessions/ISV303) - PRT207-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform (https://aws-summit-2026-kb.pages.dev/sessions/PRT207-S) - PRT103-S — Cloud Anywhere: Architectural Freedom for Unified Data and AI 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects (https://aws-summit-2026-kb.pages.dev/sessions/PRT103-S) - DAT402 — Deep dive into database integrations with AWS Zero-ETL (https://aws-summit-2026-kb.pages.dev/sessions/DAT402) - ISV304 — Managing AI Agents with Confidence and Control using Kasada & AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV304) - ARC307 — AI Powered Resilience Lifecycle (https://aws-summit-2026-kb.pages.dev/sessions/ARC307) - AIM101 — AI League Championship | 14-May | 08:00 - 16:00 (https://aws-summit-2026-kb.pages.dev/sessions/AIM101) - PRT210-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform (https://aws-summit-2026-kb.pages.dev/sessions/PRT210-S) - FSI207 — From enterprise data mesh to AI with Amazon SageMaker Unified Studio (https://aws-summit-2026-kb.pages.dev/sessions/FSI207) - DEV306 — Taming Legacy Code: Multi-Agent AI in Brownfield Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV306) - STP204 — How Heidi Health Fine-Tunes Speech-to-Text Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/STP204) - FSI204 — Agentic AI in Financial Services: Architectural Patterns That Work (https://aws-summit-2026-kb.pages.dev/sessions/FSI204) - IND204 — How Transurban Transformed Customer Experience with AI Agents on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND204) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) - STP212 — How Apate AI uses Amazon Bedrock and voice AI to catch scammers (https://aws-summit-2026-kb.pages.dev/sessions/STP212) - IDE102 — Power of Possibility: Leading Through Innovation and Connection (https://aws-summit-2026-kb.pages.dev/sessions/IDE102) - ISV101 — How AI is Transforming Pharmacy Care with Amazon Nova:MedAdvisor Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV101) - BIZ201 — AI-Everywhere: Transform Customer Interactions into Memorable Moments (https://aws-summit-2026-kb.pages.dev/sessions/BIZ201) - IND201 — Transforming software license efficiency - Human-centered AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND201) - ISV203 — AI Monetization and Pricing Strategies (https://aws-summit-2026-kb.pages.dev/sessions/ISV203) - SMB204 — Accelerated Insights from Amazon Connect using AI (https://aws-summit-2026-kb.pages.dev/sessions/SMB204) - IND101 — Test, Learn, Iterate: Amazon Connect Success (https://aws-summit-2026-kb.pages.dev/sessions/IND101) - IND206 — How scalable data foundations helped TGE unlock the power of AI (https://aws-summit-2026-kb.pages.dev/sessions/IND206) - STP214 — Create hyper-personalized voice interactions with Amazon Nova Sonic (https://aws-summit-2026-kb.pages.dev/sessions/STP214) - ISV213 — From GRC Platform to AI-Native Risk Intelligence on AWS:Protecht Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV213) - MAE204 — How Amazon Ads Creative Agent uses AWS to democratize ad creation (https://aws-summit-2026-kb.pages.dev/sessions/MAE204) --- ## Data Governance & Privacy URL: https://aws-summit-2026-kb.pages.dev/topics/governance Tagline: Make data discoverable, trustworthy, and compliant. ### Overview Modern data governance balances access and control. Amazon DataZone provides a business-friendly data catalog, AWS Lake Formation enforces fine-grained access on the data lake, AWS Glue Data Catalog is the technical metadata store, and Amazon Macie discovers PII in S3. Active metadata, lineage (via OpenLineage), and data contracts are emerging best practices. For AI specifically, model cards, data sheets, and AI guardrails extend governance to ML/LLM systems. ### Key concepts - Data catalog vs. data marketplace vs. data product - Fine-grained access: row, column, cell-level - Lineage and impact analysis - PII discovery and classification - AI governance: model cards, evaluation, watermarking ### Key AWS services - Amazon DataZone - AWS Lake Formation - AWS Glue Data Catalog - Amazon Macie - AWS Audit Manager ### Curated external resources - [Amazon DataZone](https://aws.amazon.com/datazone/) - [AWS Lake Formation](https://aws.amazon.com/lake-formation/) - [OpenLineage — open standard](https://openlineage.io/) - [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) ### Live monitored sources (Parallel AI) - [CISA, US and International Partners Release Guide to Secure ...](https://www.cisa.gov/news-events/news/cisa-us-and-international-partners-release-guide-secure-adoption-agentic-ai) — cisa.gov (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [The Context Graph Revolution: Why Enterprise AI ... - Medium](https://medium.com/@thanapong_18619/the-context-graph-revolution-why-enterprise-ai-needs-decision-lineage-c01d90fd1db4) — medium.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [Comment and Control: GitHub AI Agents as Credential ...](https://labs.cloudsecurityalliance.org/research/csa-research-note-comment-control-github-prompt-injection-20/) — labs.cloudsecurityalliance.org (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against ### Sessions on this topic (7) - ANT301 — A practitioners guide to data for agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/ANT301) - ARC301 — Build an AI-ready data foundation (https://aws-summit-2026-kb.pages.dev/sessions/ARC301) - STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI (https://aws-summit-2026-kb.pages.dev/sessions/STP208) - WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health (https://aws-summit-2026-kb.pages.dev/sessions/WPS203) - FSI207 — From enterprise data mesh to AI with Amazon SageMaker Unified Studio (https://aws-summit-2026-kb.pages.dev/sessions/FSI207) - STP209 — How Cartesian Turns AI Agents from SaaS Killer to SaaS Moat (https://aws-summit-2026-kb.pages.dev/sessions/STP209) - IDE101 — From principles to practice: Scaling AI responsibly (https://aws-summit-2026-kb.pages.dev/sessions/IDE101) --- ## Industry Spotlight: Financial Services URL: https://aws-summit-2026-kb.pages.dev/topics/industries-fsi Tagline: AI, real-time payments, fraud, and regulation in FSI. ### Overview Financial services on AWS uses generative AI for personalised banking, AI agents for KYC/AML automation, real-time fraud detection on streaming data, and core-banking modernization off mainframes. Compliance frameworks like APRA CPS 234 (Australia), SOC 2, PCI-DSS, and ISO 27001 are supported via AWS Artifact and a deep partner network. Quality validation agents accelerate regulated change. ### Key concepts - Confidential computing for sensitive workloads - Streaming fraud detection patterns - Open banking APIs and tokenization - Core-banking and mainframe modernization - Model risk management and explainability ### Key AWS services - Amazon Bedrock - Amazon Fraud Detector - AWS Clean Rooms - AWS Mainframe Modernization ### Curated external resources - [AWS Financial Services](https://aws.amazon.com/financial-services/) - [AWS Compliance Programs](https://aws.amazon.com/compliance/programs/) - [AWS Mainframe Modernization](https://aws.amazon.com/mainframe-modernization/) ### Live monitored sources (Parallel AI) - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [Stripe introduces Link, a digital wallet that autonomous AI ...](https://postofday.com/2026/05/01/stripe-introduces-link-a-digital-wallet-that-autonomous-ai-agents-can-use-too/) — postofday.com (2026-05-05): Reports indicate that MoonPay and the Agentic Experience Protocol (AXP) have launched functional agent payment infrastructure (April-May 2026), with AXP extending the Universal Commerce Protocol (UCP) to support unified agentic commerce experiences and rich product data. - [Beyond humans: Lily Liu says Solana is building the payment rails for the 'AI machine economy'](https://www.coindesk.com/business/2026/05/06/beyond-humans-lily-liu-says-solana-is-building-the-payment-rails-for-the-ai-machine-economy) — coindesk.com (2026-05-07): Binance founder Changpeng Zhao (CZ) stated that the BNB Chain is the optimal payments rail for automated transactions between AI agents, highlighting the chain's existing infrastructure and the adoption of agent-specific standards like BAP-578. - [Meow Technologies launches the first agentic banking ...](http://thenextweb.com/news/meow-technologies-agentic-banking-ai-agents) — thenextweb.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. ### Sessions on this topic (9) - STP215 — How Sonder Improve 24/7 Employee Wellbeing with AWS AI (https://aws-summit-2026-kb.pages.dev/sessions/STP215) - STP201 — Scaling Security at Startup Speed: Hnry's AI-Powered Approach (https://aws-summit-2026-kb.pages.dev/sessions/STP201) - FSI206 — Agentic AI Transforming Quality at Cloud Speed (https://aws-summit-2026-kb.pages.dev/sessions/FSI206) - WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health (https://aws-summit-2026-kb.pages.dev/sessions/WPS203) - FSI201 — BELIEVE: The Impossible Migration That Transformed Australian Banking (https://aws-summit-2026-kb.pages.dev/sessions/FSI201) - ISV209 — From dev tools to customer value: BGL's agentic AI journey (https://aws-summit-2026-kb.pages.dev/sessions/ISV209) - FSI204 — Agentic AI in Financial Services: Architectural Patterns That Work (https://aws-summit-2026-kb.pages.dev/sessions/FSI204) - DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV203) - ISV102 — From documents to voice - building AI products on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV102) --- ## Industry Spotlight: Public Sector & Government URL: https://aws-summit-2026-kb.pages.dev/topics/industries-public Tagline: Citizen services, defense and sovereign cloud capability. ### Overview AWS supports government and public sector with sovereign-cloud options, IRAP (Australia), FedRAMP/IL5 (US), and dedicated regions. Generative AI is being applied to citizen service portals, document processing, defense logistics, and digital workforce upskilling. AWS GovCloud and partner-led sovereign clouds offer data-residency and operator controls. ### Key concepts - Sovereign cloud and data residency - IRAP / Essential Eight / FedRAMP / IL5 - Citizen-facing AI assistants - Defense and intelligence workloads - Workforce upskilling at scale ### Key AWS services - AWS GovCloud - AWS European Sovereign Cloud - Amazon Bedrock - AWS Skill Builder ### Curated external resources - [AWS Public Sector](https://aws.amazon.com/government-education/) - [AWS for Government](https://aws.amazon.com/government-education/government/) - [IRAP — Australian Government](https://www.cyber.gov.au/irap) ### Live monitored sources (Parallel AI) - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [hitehouse.gov](https://www.whitehouse.gov/omb/) — hitehouse.gov (2026-05-02): {"event_type": "company_overview", "extracted_fact": "The Office of Management and Budget (OMB) is a federal government agency headquartered in Washington, D.C., operating in the public administration sector. It is federally funded and serves as the primary budgetary and manageme - [Agentic AI Enterprise Use Cases — 30+ Real Deployments (2026)](https://www.ampcome.com/post/post-agentic-ai-enterprise-use-cases) — ampcome.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [80% of Fortune 500 use active AI Agents: Observability ...](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/) — microsoft.com (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 ### Sessions on this topic (7) - SEC305 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC305) - SEC301 — Inside the Attack Chain: Emerging Threat Actor Tactics and Techniques (https://aws-summit-2026-kb.pages.dev/sessions/SEC301) - SEC401 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads (https://aws-summit-2026-kb.pages.dev/sessions/SEC401) - ARC201 — Building on AWS resilience: Innovations for critical success (https://aws-summit-2026-kb.pages.dev/sessions/ARC201) - PRT108-S — From Experiment to Production: Unlock AI Deployment Bottlenecks (https://aws-summit-2026-kb.pages.dev/sessions/PRT108-S) - WPS302 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS302) - WPS202 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS202) --- ## Industry Spotlight: Healthcare & Life Sciences URL: https://aws-summit-2026-kb.pages.dev/topics/industries-hls Tagline: AI for clinicians, researchers, and patients. ### Overview AWS supports healthcare with HIPAA-eligible services, AWS HealthLake for FHIR data, Amazon HealthScribe for clinical note generation, and Amazon Bedrock for life-sciences research (drug discovery, clinical-trial analysis). AWS HealthOmics handles genomic, transcriptomic, and other omics data at scale. Privacy-preserving techniques (federated learning, AWS Clean Rooms) enable cross-organization collaboration on sensitive data. ### Key concepts - FHIR and interoperability standards - Clinical NLP and ambient AI scribes - Genomic and multi-omics workflows - Privacy-preserving collaboration - Real-world evidence and clinical trials ### Key AWS services - AWS HealthLake - Amazon HealthScribe - AWS HealthOmics - Amazon Comprehend Medical ### Curated external resources - [AWS for Healthcare](https://aws.amazon.com/health/) - [AWS HealthLake](https://aws.amazon.com/healthlake/) - [AWS HealthOmics](https://aws.amazon.com/healthomics/) ### Live monitored sources (Parallel AI) - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [80% of Fortune 500 use active AI Agents: Observability ...](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/) — microsoft.com (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [techlifesci.com](https://www.techlifesci.com/p/everyone-is-building-ai-agents) — techlifesci.com (2026-04-04): Industry analysis (TechLifeSci, 2026-04-04) surveying agent activity across biopharma and enterprise. Summary: infrastructure and investments (Roche, Eli Lilly GPU builds) are large; a few agent systems are reported running in production workflows (operational automation, data-re - [onixnet.com](https://www.onixnet.com/news/onix-deepens-strategic-collaboration-with-google-cloud-to-help-accelerate-enterprise-scale-cloud-data-and-agentic-ai-transformation/) — onixnet.com (2026-04-06): Onix (Apr 6, 2026) announced an expanded strategic collaboration with Google Cloud centered on its Wingspan agentic AI platform. The release states Wingspan has already driven "thousands of AI agents deployed in production environments across Fortune 500 companies," positions Win ### Sessions on this topic (7) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health (https://aws-summit-2026-kb.pages.dev/sessions/WPS203) - WPS301 — AWS for healthcare analytics: accelerating time to insights (https://aws-summit-2026-kb.pages.dev/sessions/WPS301) - STP204 — How Heidi Health Fine-Tunes Speech-to-Text Models on AWS (https://aws-summit-2026-kb.pages.dev/sessions/STP204) - ISV101 — How AI is Transforming Pharmacy Care with Amazon Nova:MedAdvisor Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV101) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) - SMB205 — How Blackmores accelerated SAP RISE connectivity with an EBA and Kiro (https://aws-summit-2026-kb.pages.dev/sessions/SMB205) --- ## Media & Entertainment URL: https://aws-summit-2026-kb.pages.dev/topics/media Tagline: Stream, create, and monetise content with AWS. ### Overview AWS Media Services provide a complete pipeline: AWS Elemental MediaLive for live encoding, MediaPackage for origination, MediaConvert for VOD, MediaTailor for personalized ads, and Amazon IVS for low-latency streaming. Generative AI is transforming content creation — Amazon Nova Reel for video generation, Amazon Nova Canvas for images, and partner solutions for VFX and post-production. ### Key concepts - Live vs. VOD streaming architectures - DRM, watermarking, and content protection - Personalized ad insertion (SSAI) - Cloud-native production and remote editing - Generative AI in creative pipelines ### Key AWS services - AWS Elemental MediaLive - AWS Elemental MediaConvert - Amazon IVS - Amazon Nova Reel ### Curated external resources - [AWS for M&E](https://aws.amazon.com/media/) - [AWS Elemental services](https://aws.amazon.com/media-services/) - [AWS M&E Blog](https://aws.amazon.com/blogs/media/) ### Sessions on this topic (16) - DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/DEV204) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - DEV304 — Building Agentic AI: Amazon Nova Act and Strands Agents in Practice (https://aws-summit-2026-kb.pages.dev/sessions/DEV304) - STP302 — Unleash Live: Cloud-Powered Vision for Infrastructure (https://aws-summit-2026-kb.pages.dev/sessions/STP302) - AIM301 — Commbank pioneering AI-driven DevSecOps with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/AIM301) - ISV206 — Scaling RAG to Millions of Vectors: The Squiz Story (https://aws-summit-2026-kb.pages.dev/sessions/ISV206) - MAE202 — Seven's AWS Journey: Streaming Premium Content at the Speed of Innovation (https://aws-summit-2026-kb.pages.dev/sessions/MAE202) - IND202 — How Zuru Uses AI to Analyze TikTok Trends for Rapid Content Creation (https://aws-summit-2026-kb.pages.dev/sessions/IND202) - STP301 — AI-Native Remediation with Pleri: Your Security Engineer That Ships (https://aws-summit-2026-kb.pages.dev/sessions/STP301) - INO203 — Behind the curtain: How Amazons AI innovations are powered by AWS (https://aws-summit-2026-kb.pages.dev/sessions/INO203) - MAE205 — AI at Speed of News: Unlocking Value from Media with Generative AI (https://aws-summit-2026-kb.pages.dev/sessions/MAE205) - DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops (https://aws-summit-2026-kb.pages.dev/sessions/DEV206) - DEV208 — Production-Grade Platforms: Real-World IaC Practices on AWS (https://aws-summit-2026-kb.pages.dev/sessions/DEV208) - MAE204 — How Amazon Ads Creative Agent uses AWS to democratize ad creation (https://aws-summit-2026-kb.pages.dev/sessions/MAE204) - SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR (https://aws-summit-2026-kb.pages.dev/sessions/SMB203) - SEC501 — Where Big Ideas Live: How to Actually Read Research Papers (https://aws-summit-2026-kb.pages.dev/sessions/SEC501) --- ## Manufacturing & Industry 4.0 URL: https://aws-summit-2026-kb.pages.dev/topics/manufacturing Tagline: Smart factories, digital twins, and predictive operations. ### Overview AWS for Industrial connects OT (operational technology) and IT to drive Industry 4.0. AWS IoT SiteWise collects, structures, and analyzes industrial equipment data at scale. AWS IoT TwinMaker builds digital twins. Amazon Lookout for Equipment / Vision applies ML for predictive maintenance and quality. Generative AI is now used for anomaly explanations, root-cause analysis, and operator assistance. ### Key concepts - OT/IT convergence and the Purdue model - Asset hierarchies and industrial data models - Predictive maintenance and condition monitoring - Digital twins and simulation - Edge processing for latency and resilience ### Key AWS services - AWS IoT SiteWise - AWS IoT TwinMaker - Amazon Lookout for Equipment - Amazon Lookout for Vision ### Curated external resources - [AWS for Industrial](https://aws.amazon.com/industrial/) - [AWS IoT SiteWise](https://aws.amazon.com/iot-sitewise/) - [AWS IoT TwinMaker](https://aws.amazon.com/iot-twinmaker/) ### Sessions on this topic (19) - ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/ISV302) - DEV207 — Data Observability Without the Pain - Lessons from a Production System (https://aws-summit-2026-kb.pages.dev/sessions/DEV207) - STP215 — How Sonder Improve 24/7 Employee Wellbeing with AWS AI (https://aws-summit-2026-kb.pages.dev/sessions/STP215) - STP210 — TeamForm's Generative Dashboards with Strands & Bedrock AgentCore (https://aws-summit-2026-kb.pages.dev/sessions/STP210) - ARC304 — Demystifying Agent Identity (https://aws-summit-2026-kb.pages.dev/sessions/ARC304) - STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI (https://aws-summit-2026-kb.pages.dev/sessions/STP208) - ISV304 — Managing AI Agents with Confidence and Control using Kasada & AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV304) - SEC302 — Leap ahead in Cloud Operations with AWS DevOps Agent (https://aws-summit-2026-kb.pages.dev/sessions/SEC302) - DEV210 — AI-Driven Incident Triage: From Slack Alert to Root Cause (https://aws-summit-2026-kb.pages.dev/sessions/DEV210) - ARC307 — AI Powered Resilience Lifecycle (https://aws-summit-2026-kb.pages.dev/sessions/ARC307) - WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health (https://aws-summit-2026-kb.pages.dev/sessions/WPS203) - DEV306 — Taming Legacy Code: Multi-Agent AI in Brownfield Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV306) - ISV211 — Scaling Conversation Intelligence with Agentic AI on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV211) - IND204 — How Transurban Transformed Customer Experience with AI Agents on AWS (https://aws-summit-2026-kb.pages.dev/sessions/IND204) - STP301 — AI-Native Remediation with Pleri: Your Security Engineer That Ships (https://aws-summit-2026-kb.pages.dev/sessions/STP301) - IDE101 — From principles to practice: Scaling AI responsibly (https://aws-summit-2026-kb.pages.dev/sessions/IDE101) - STP101 — Driving Profitable Growth with Generative AI: From Prompt to Product (https://aws-summit-2026-kb.pages.dev/sessions/STP101) - DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops (https://aws-summit-2026-kb.pages.dev/sessions/DEV206) - ISV214 — Grounding AI Agents: How to give your AI real-world data with MCP (https://aws-summit-2026-kb.pages.dev/sessions/ISV214) --- ## Startups & Innovation URL: https://aws-summit-2026-kb.pages.dev/topics/startups Tagline: Build fast, scale globally, fail cheap. ### Overview AWS Activate provides up to $100K in credits, technical mentorship, and co-marketing for eligible startups. The AWS Generative AI Accelerator and AWS Marketplace help startups distribute products. Patterns favored by startups: serverless-first, single-table DynamoDB designs, Amplify for full-stack mobile/web, Bedrock for AI features, and event-driven architectures that scale linearly with users. ### Key concepts - Build-measure-learn loops and lean architecture - Serverless-first patterns to keep COGS low - Multi-tenant SaaS isolation models - Going from MVP to product-market-fit on AWS - AWS Activate, Marketplace, and ISV co-sell ### Key AWS services - AWS Amplify - AWS Lambda - Amazon DynamoDB - Amazon Bedrock - AWS Activate ### Curated external resources - [AWS Activate for Startups](https://aws.amazon.com/startups/) - [AWS SaaS Factory](https://aws.amazon.com/partners/programs/saas-factory/) - [AWS Startup Blog](https://aws.amazon.com/blogs/startups/) ### Live monitored sources (Parallel AI) - [Top Tech News Today, May 1, 2026 - Tech Startups](https://techstartups.com/2026/05/01/top-tech-news-today-may-1-2026/) — techstartups.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-05): The CSAI Foundation (Cloud Security Alliance) announced milestones to secure the agentic control plane, including the strategic acquisition/stewardship of two foundational specifications: the Autonomous Action Runtime Management (AARM) specification (for securing AI-driven action - [techstartups.com](https://techstartups.com/2026/04/29/top-tech-news-today-april-29-2026/) — techstartups.com (2026-04-29): OpenAI expanded distribution by making its models (including Codex and managed agents) available on AWS Bedrock, broadening cloud provider options for enterprises. ### Sessions on this topic (6) - STP201 — Scaling Security at Startup Speed: Hnry's AI-Powered Approach (https://aws-summit-2026-kb.pages.dev/sessions/STP201) - ISV301 — Rolling to Scale: Roller's Multi-Tenant SaaS platform on AWS (https://aws-summit-2026-kb.pages.dev/sessions/ISV301) - STP213 — AI-Powered Farming: How Halter's ML Models Transform Dairy Operations (https://aws-summit-2026-kb.pages.dev/sessions/STP213) - STP101 — Driving Profitable Growth with Generative AI: From Prompt to Product (https://aws-summit-2026-kb.pages.dev/sessions/STP101) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) - STP202 — Stop Vibing, Start Shipping: How Startups Build with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/STP202) --- ## Diversity, Equity & Inclusion in Tech URL: https://aws-summit-2026-kb.pages.dev/topics/dei Tagline: Build inclusive teams and equitable AI. ### Overview AWS's diversity-in-tech track focuses on closing skill gaps with programs like AWS re/Start, AWS GetIT, and the AWS Skill Builder free tier. Equitable AI requires diverse teams and intentional bias evaluation across data, models, and outcomes. Programs like Women in Tech, AWS She Builds, and partnerships with organizations like Code Like a Girl support representation across the cloud workforce. ### Key concepts - Inclusive design and accessibility (WCAG, ARIA) - Bias detection and fairness metrics in ML - Allyship, mentorship, and sponsorship - Inclusive hiring and skills-based assessments - Re-skilling and AI literacy programs ### Key AWS services - AWS re/Start - AWS Skill Builder - SageMaker Clarify (bias detection) ### Curated external resources - [AWS Diversity, Equity & Inclusion](https://aws.amazon.com/diversity-inclusion/) - [AWS re/Start](https://aws.amazon.com/training/restart/) - [AWS She Builds Community](https://community.aws/she-builds) ### Sessions on this topic (1) - IDE301 — Diversity In Tech - AI Literacy Skills - Rapid prototyping with Kiro (https://aws-summit-2026-kb.pages.dev/sessions/IDE301) --- ## AWS Well-Architected Framework URL: https://aws-summit-2026-kb.pages.dev/topics/well-architected Tagline: Six pillars for building great cloud workloads. ### Overview The AWS Well-Architected Framework codifies architecture best practices across six pillars: Operational Excellence, Security, Reliability, Performance Efficiency, Cost Optimization, and Sustainability. The Well-Architected Tool runs free reviews against your accounts; lenses extend the framework with domain-specific guidance (Generative AI Lens, SaaS, Serverless, IoT, Machine Learning, Financial Services, Games). Use it to identify high-risk issues and track remediation over time. ### Key concepts - Six pillars and their design principles - Lenses for domain-specific guidance - Risk identification: high (HRI) and medium (MRI) - Trade-off analysis between pillars - Continuous improvement with periodic reviews ### Key AWS services - AWS Well-Architected Tool - AWS Trusted Advisor ### Curated external resources - [AWS Well-Architected Framework](https://aws.amazon.com/architecture/well-architected/) - [Well-Architected Generative AI Lens](https://docs.aws.amazon.com/wellarchitected/latest/generative-ai-lens/generative-ai-lens.html) - [AWS Architecture Center](https://aws.amazon.com/architecture/) ### Sessions on this topic (2) - FSI206 — Agentic AI Transforming Quality at Cloud Speed (https://aws-summit-2026-kb.pages.dev/sessions/FSI206) - FSI204 — Agentic AI in Financial Services: Architectural Patterns That Work (https://aws-summit-2026-kb.pages.dev/sessions/FSI204) --- ## Mobile & Cross-Platform Development URL: https://aws-summit-2026-kb.pages.dev/topics/mobile Tagline: Ship full-stack mobile and web apps in days, not months. ### Overview AWS Amplify gives developers a full toolkit — auth (Cognito), data (AppSync GraphQL or REST), storage (S3), functions (Lambda), and hosting — to build and deploy React, Next.js, React Native, Flutter, iOS, and Android apps. Amplify Gen 2 uses TypeScript-first infrastructure as code via AWS CDK. AWS Device Farm tests on real devices in the cloud. ### Key concepts - Backend as code with Amplify Gen 2 / CDK - GraphQL with AWS AppSync vs. REST - Mobile auth flows with Amazon Cognito - Push notifications and offline-first sync - Real-device testing with Device Farm ### Key AWS services - AWS Amplify - AWS AppSync - Amazon Cognito - AWS Device Farm - Amazon Pinpoint ### Curated external resources - [AWS Amplify](https://aws.amazon.com/amplify/) - [AWS AppSync](https://aws.amazon.com/appsync/) - [Amazon Cognito](https://aws.amazon.com/cognito/) - [AWS Mobile Blog](https://aws.amazon.com/blogs/mobile/) ### Sessions on this topic (4) - PRT301-S — Unite Teams, Tools, and AI to Drive Transformation at Scale (https://aws-summit-2026-kb.pages.dev/sessions/PRT301-S) - AIM204 — Get to know Amazon Quick, your new agentic teammate (https://aws-summit-2026-kb.pages.dev/sessions/AIM204) - IDE102 — Power of Possibility: Leading Through Innovation and Connection (https://aws-summit-2026-kb.pages.dev/sessions/IDE102) - INO203 — Behind the curtain: How Amazons AI innovations are powered by AWS (https://aws-summit-2026-kb.pages.dev/sessions/INO203) --- ## Gaming & Interactive Media URL: https://aws-summit-2026-kb.pages.dev/topics/gaming Tagline: Power online games — from indie to AAA. ### Overview AWS for Games offers Amazon GameLift (managed dedicated game servers and matchmaking), Amazon GameLift Streams (cloud streaming), and a broad ecosystem of partner middleware (Unreal, Unity, Photon). Generative AI is increasingly used for NPC dialog, content generation, and player support. The new generation of agentic NPCs powered by Bedrock unlocks emergent gameplay. ### Key concepts - Dedicated server fleets and FlexMatch matchmaking - Game backend patterns: leaderboards, inventory, social - Cloud streaming and pixel streaming - Anti-cheat and player safety - Generative AI in game development workflows ### Key AWS services - Amazon GameLift - Amazon GameLift Streams - Amazon Lumberyard / Open 3D Engine - Amazon Bedrock ### Curated external resources - [AWS for Games](https://aws.amazon.com/gametech/) - [Amazon GameLift](https://aws.amazon.com/gamelift/) - [AWS Game Tech Blog](https://aws.amazon.com/blogs/gametech/) ### Sessions on this topic (4) - GHJ301 — R1 — AWS Game Day : Secret Agent Unicorns (https://aws-summit-2026-kb.pages.dev/sessions/GHJ301) - GHJ301 — R2 — AWS Game Day : Secret Agent Unicorns (https://aws-summit-2026-kb.pages.dev/sessions/GHJ301) - AIM403 — AI League (https://aws-summit-2026-kb.pages.dev/sessions/AIM403) - WPS202 — Secure and Resilient Agentic AI for High-Assurance Environments (https://aws-summit-2026-kb.pages.dev/sessions/WPS202) --- ## Code Generation & AI-Assisted Development URL: https://aws-summit-2026-kb.pages.dev/topics/ai-coding Tagline: AI pair programmers that actually ship. ### Overview AI-assisted coding has moved from autocomplete to full agentic workflows. Amazon Q Developer and Kiro provide inline suggestions, transformations (Java upgrades, .NET porting), security scans, and full agentic edits across files. The shift from "vibe-coding" to spec-driven development addresses quality, traceability, and trust. Effective use combines AI with strong testing, code review, and observability. ### Key concepts - Inline suggestions vs. agentic edits vs. autonomous swarms - Spec-driven development with Kiro - Code transformation: Java upgrades, .NET porting, mainframe - Security scanning and SBOM with Q Developer - Productivity measurement (DORA metrics for AI) ### Key AWS services - Amazon Q Developer - Kiro - AWS CodeWhisperer (legacy) ### Curated external resources - [Amazon Q Developer](https://aws.amazon.com/q/developer/) - [Kiro IDE](https://kiro.dev/) - [GitHub research: AI and developer productivity](https://github.blog/news-insights/research/) ### Live monitored sources (Parallel AI) - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [GitHub Copilot in Visual Studio Code, April releases](http://github.blog/changelog/2026-05-06-github-copilot-in-visual-studio-code-april-releases) — github.blog (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [Best AI Code Generation Tools 2026 - TrustRadius](http://trustradius.com/categories/ai-code-generation) — trustradius.com (2026-05-11): Devin updated its pricing structure to include several tiers: Free (limited usage, Devin Review, DeepWiki), Pro ($20/month with usage quotas and integrations), Max ($200/month with increased quotas), Teams ($80/month with unlimited members and centralized billing), and Enterprise - [2026](https://docs.devin.ai/release-notes/2026) — docs.devin.ai (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [Cursor AI 2026: Complete Guide to New Features, Tips ...](https://anycap.ai/page/en-US/blog/cursor-ai-2026-new-features-guide) — anycap.ai (2026-05-09): Reports indicate that GitHub Copilot paused new sign-ups for its Pro, Pro+, and Student tiers and removed Claude Opus models from the Pro tier in May 2026. ### Sessions on this topic (6) - DVT201 — Building Software like never before with agentic AI (https://aws-summit-2026-kb.pages.dev/sessions/DVT201) - DEV209 — CI/CD Guardrails for Agentic Coding Workflows (https://aws-summit-2026-kb.pages.dev/sessions/DEV209) - ISV201 — MCP on EKS: Xero's AI-Driven Developer Experience (https://aws-summit-2026-kb.pages.dev/sessions/ISV201) - TNC203 — Structured Approach to AI coding with Spec-Driven Development on Kiro (https://aws-summit-2026-kb.pages.dev/sessions/TNC203) - DEV306 — Taming Legacy Code: Multi-Agent AI in Brownfield Systems (https://aws-summit-2026-kb.pages.dev/sessions/DEV306) - STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade (https://aws-summit-2026-kb.pages.dev/sessions/STP216) --- # PART 2 — SESSIONS ## PRT109-S — Hello Future, Meet Reality: Enterprise AI Lessons URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT109-S Level: foundational Type: Breakout session Category: Partner Showcase Topics: Security, Identity & Compliance Hello Future, Meet Reality is a fast, frank panel discussion with leaders on what actually held up when AI hit real-world conditions: legacy systems, security teams, governance, data mess, change fatigue... the lot. No magic wand, just real decisions, real trade-offs, and get their real view of what ""future-ready"" looks like in an enterprise. ### Playbook (editorial commentary) **The concept.** A retrospective on what actually broke when generative AI met production: legacy integrations, security pushback, governance vacuum, change fatigue. **Why it matters.** Most public AI ROI stories are pilot-phase. This is where you hear what survived contact with month 6. **The hard parts.** The blockers are rarely technical. They are: who owns the data, who signs off, "we never documented that integration," and people who simply don't trust the model's output enough to act on it. **Playbook moves.** (1) Bring an honest pilot post-mortem template — winners, losers, why. (2) Track ratio of integration time vs. modeling time. If integration is less than 70%, your pilot was unrealistically clean. (3) Identify your "Schelling point" stakeholder — the one whose nod breaks the org logjam. Brief them first. **The surprise.** The most expensive failure pattern isn't model performance — it's data ownership politics. When the data steward and the AI team report to different VPs, projects die in handoffs, not in inference. --- ### Live monitored sources - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [CISA, US and International Partners Release Guide to Secure ...](https://www.cisa.gov/news-events/news/cisa-us-and-international-partners-release-guide-secure-adoption-agentic-ai) — cisa.gov (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Releases · microsoft/agent-governance-toolkit · GitHub](https://github.com/microsoft/agent-governance-toolkit/releases) — github.com (2026-05-08): Microsoft released v3.5.0 of the Agent Governance Toolkit, adding enterprise-grade agent identity via Citadel Integration (Entra identity bridge), Multi-Agent Collective Policies for workflow-wide constraints, Intent-Based Authorization for structured lifecycle management (declar --- ## PRT215-S — The Visibility Gap: Turning Observability into DevSecOps Signals URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT215-S Level: intermediate Type: Breakout session Category: Partner Showcase Topics: Observability & Monitoring; DevOps, CI/CD & DevSecOps; Security, Identity & Compliance The Visibility Gap: Turning Observability into DevSecOps Signals (sponsored by Datadog)Security teams and dev teams share the same production environment but operate from different signals. Without runtime context, security monitoring has blind spots with misconfigured infrastructure and threats in flight. This session draws on Fone Dynamics' ISO 27001 journey to show how runtime telemetry, cloud audit logs, and code scanning give DevSecOps and SecOps teams shared context. ### Playbook (editorial commentary) **The concept.** SecOps and DevOps see the same systems through different lenses. Runtime telemetry becomes the shared substrate that makes them speak the same language. **Why it matters.** Vulnerability scans tell you what *could* be wrong. Runtime telemetry tells you what *is* wrong, right now. ISO 27001 (and most modern audits) want evidence of controls *operating*, not just existing. **The hard parts.** Mapping abstract control language ("access is restricted to authorised users") to concrete log events ("user X read object Y from bucket Z") is laborious. Most orgs duck it. **Playbook moves.** (1) Build a control-to-signal map: one row per control, one column per evidence source. (2) Anything without a signal source is a paper control — burn those down. (3) Tag existing logs with control IDs at ingestion, not at audit time. **The surprise.** The fastest path to ISO 27001 evidence isn't more controls — it's tagging existing logs to existing control IDs. Most enterprises already have 70%+ of evidence; they just can't find it on demand. --- ### Live monitored sources - [GitHub - Siddhant-K-code/agent-trace: strace for AI agents. Capture and replay every tool call, prompt, and response from Claude Code, Cursor, Gemini CLI or any MCP client · GitHub](https://github.com/Siddhant-K-code/agent-trace) — github.com (2026-05-04): The 'agent-trace' developer tool (GitHub: Siddhant-K-code/agent-trace) has launched significant new monitoring and control features: 1) A 'watch' mode that automatically terminates agents (using SIGSTOP or SIGTERM) when specific rules in a .watch-rules.json file are triggered, su - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [AI Agent Authentication & Authorization in 2026: What Works ...](https://api.aport.io/blog/best-ai-agent-authentication-authorization-2026) — api.aport.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [AI Agent Delegation Patterns: Four Best Architectures for 2026 | Fastio](https://fast.io/resources/ai-agent-delegation-patterns) — fast.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca --- ## PRT104-S — Building Resilience for AI Data Foundations and Cloud-Native Apps 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT104-S Level: foundational Type: Lightning talk Category: Partner Showcase Topics: Resilience & Disaster Recovery; Security, Identity & Compliance; Generative AI & Foundation Models AI innovation depends on consistent, trusted data. When disrupted, AI systems and the business decisions they support are at risk. In this session, learn how cloudnative protection models support AI pipelines, reduce recovery time after disruptions, and minimise operational overhead. Discover best practices to protect AI and cloudnative applications in AWS while innovating with confidence. ### Playbook (editorial commentary) **The concept.** AI training data and feature stores are the new crown jewels. Recovery cost from corruption now exceeds the cost of regenerating model weights. **Why it matters.** A poisoned RAG corpus or corrupted training set can ship bad behaviour to thousands of users before you notice. Silent data corruption is worse than data loss. **The hard parts.** Backups for ephemeral pipeline state are weird. Idempotent reprocessing usually beats restoring intermediate state. Integrity checking on training data is not the same as backup. **Playbook moves.** (1) Define RPO/RTO per data class — raw, processed features, embeddings, model weights. They differ. (2) Set integrity hashes on training corpora; recheck before fine-tune jobs. (3) Test restore quarterly, not annually. **The surprise.** Most teams over-protect model weights (cheap to retrain) and under-protect feature stores (expensive to rebuild from raw). Your backup budget is probably allocated wrongly. --- ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Top Tech News Today, May 1, 2026 - Tech Startups](https://techstartups.com/2026/05/01/top-tech-news-today-may-1-2026/) — techstartups.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. --- ## PRT202-S — 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT202-S Level: intermediate Type: Lightning talk Category: Partner Showcase Topics: Agentic AI; Security, Identity & Compliance; Generative AI & Foundation Models This session demonstrates how quickly and easily enterprise-grade AI security can be applied to your Amazon Bedrock projects. See the new Prisma AIRS AI-security platform in action on AWS: We explore Model Scanning and AI Posture Management; then dive into Runtime Security; and finally examine automated AI Red Teaming and AI Agent Security. Unlock your AI-led innovation - securely. ### Playbook (editorial commentary) **The concept.** AI security splits into four distinct surfaces: model supply chain (scanning), configuration (posture), inference traffic (runtime), and adversarial probing (red teaming). Each needs different tooling. **Why it matters.** Securing AI is not "put it behind a WAF." Prompt injection bypasses traditional perimeters. Models can leak training data. Tool-calling agents have privileges your APIs never granted. **The hard parts.** Threat models for agents look different from threat models for APIs. Your existing playbook misses the new vectors. **Playbook moves.** (1) Threat-model the agent's tool surface specifically. (2) Separate "data exfil via response" from "command injection via tool call" — they need different controls. (3) Schedule automated red-team runs into the release pipeline. **The surprise.** The riskiest agent vulnerability isn't prompt injection — it's *authorisation confusion*. The agent inherits user identity for context but operates with system privileges for execution. Whoever can manipulate the prompt can effectively act as the system. This is the agent-era equivalent of SSRF, and almost no one is testing for it. --- ### Live monitored sources - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Amazon Builds AI Agent Payments With Coinbase and Stripe](https://thedefiant.io/news/infrastructure/amazon-builds-ai-agent-payments-with-coinbase-and-stripe) — thedefiant.io (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [AI Agent Governance After RSAC 2026: $2.5B Invested, Zero ...](https://ienable.ai/blog/ai-agent-governance-after-rsac-2026-state-of-the-market.html) — ienable.ai (2026-05-05): The CSAI Foundation (Cloud Security Alliance) announced milestones to secure the agentic control plane, including the strategic acquisition/stewardship of two foundational specifications: the Autonomous Action Runtime Management (AARM) specification (for securing AI-driven action --- ## PRT204-S — Optimising GenAI at Runtime with Experimentation and Guardrails 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT204-S Level: intermediate Type: Lightning talk Category: Partner Showcase Topics: Security, Identity & Compliance; Generative AI & Foundation Models Generative AI systems evolve constantly, and the impact of prompt or model changes often isnt clear until real users interact with them in production. In this session, learn how teams using Amazon Bedrock safely experiment with AI at runtime, testing models and prompts with targeted rollouts, evaluating system outputs online, and optimising against real business result ### Playbook (editorial commentary) **The concept.** Treat prompts and models as feature flags. Ramp them, A/B them, kill-switch them. Don't ship a model change to 100% of users at once. **Why it matters.** Model upgrades are silent breaking changes. Without runtime control, every prompt tweak is a high-blast-radius deployment. **The hard parts.** Evaluating LLM outputs in production is harder than for traditional features. Click-through doesn't capture "the answer was confidently wrong." You need offline evals + online signals together. **Playbook moves.** (1) Tag every prompt change with a flag. Make rollback a one-click operation. (2) Define explicit output evaluators: faithfulness, toxicity, latency, cost-per-call. (3) Roll to 1% before 100%. Always. **The surprise.** The best leading indicator of a bad prompt change isn't user satisfaction — it's the *variance* of agent token consumption. Misaligned prompts cause agents to retry, second-guess, and burn tokens. Track token-spend variance per session; spikes there precede user complaints by hours. --- ### Live monitored sources - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [MCP Governance (2026): Policy Gates for MCP Servers](https://cordum.io/blog/mcp-governance-servers) — cordum.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [Experian Announces Agent Trust to Power Trusted AI ...](http://businesswire.com/news/home/20260430719198/en/Experian-Announces-Agent-Trust-to-Power-Trusted-AI-Driven-Commerce) — businesswire.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [MCP Security Gateway - Agent Governance Toolkit](https://microsoft.github.io/agent-governance-toolkit/tutorials/07-mcp-security-gateway/) — microsoft.github.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [Releases · microsoft/autogen · GitHub](https://github.com/microsoft/autogen/releases) — github.com (2026-05-07): CrewAI released pre-release version 1.14.5a3 on 2026-05-06. Key changes include: - Refactored the CLI into a standalone `crewai-cli` package. - Fixed the status endpoint path from `/{kickoff_id}/status` to `/status/{kickoff_id}`. - Updated the `gitpython` dependency to version >= --- ## AIM201 — From demo to deployment: solving agentic AI's toughest challenges URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM201 Level: intermediate Type: Breakout session Category: AI & Machine Learning Topics: Cost Optimization & FinOps; Observability & Monitoring; Security, Identity & Compliance; Agentic AI; DevOps, CI/CD & DevSecOps; Retrieval Augmented Generation (RAG) Most AI agent projects stall when moving from prototype to production. This session tackles the top challenges builders face when deploying agentic AI at scale. You'll learn how to answer the fundamental question of whether to build custom agents or leverage pre-built agents for DevOps, security, development, and business productivity use cases. Then you'll discover how to address the critical production challenges of reliability, observability, cost management, security, and evaluation. Drawing from real customer deployments and AWS's portfolio of agentic AI capabilities, you'll gain actionable approaches for building agents that don't just demo well but ship and scale. ### Playbook (editorial commentary) **The concept.** The pilot-to-production gap for agents is real and the failure modes are predictable: reliability, observability, cost, security, evals. Build vs. buy decisions are harder than for SaaS because build-cost is hidden in ops. **Why it matters.** Industry estimates suggest 70%+ of agent pilots never reach production. You'll be in that bucket if you don't pre-empt the predictable failures. **The hard parts.** "It works in the notebook" is the new "it works on my machine." Demo environments hide everything that kills production: rate limits, cold starts, multi-user state, partial failures, cost runaway. **Playbook moves.** (1) Write the production runbook *before* building. If you can't describe how to triage a failure, you're not ready to ship. (2) Pre-commit a budget per task. Cap it at the agent layer. (3) Define an offline eval suite on day one, even if small. **The surprise.** The single highest-leverage practice in agent ops is the offline eval suite. It's tedious to build but it unlocks everything downstream — model upgrades, prompt iteration, regression testing, vendor swaps. Teams that skip evals end up trapped on a single model and prompt forever. --- ### Live monitored sources - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [80% of Fortune 500 use active AI Agents: Observability ...](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/) — microsoft.com (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [MCP Governance (2026): Policy Gates for MCP Servers](https://cordum.io/blog/mcp-governance-servers) — cordum.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes --- ## AIM401 — Beyond API Dependency: Fine-tuning Cost-Effective Models on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM401 Level: expert Type: Breakout session Category: AI & Machine Learning Topics: Machine Learning & SageMaker; Generative AI & Foundation Models As API costs for general-purpose LLMs rise, relying solely on off-the-shelf models can quickly undermine both cost control and system reliability. In this session, we share how Nearmap moved beyond API dependency by fine-tuning and distilling domain-specific models on AWS to analyze 300 million building permits for roof modifications. Well discuss our approach to generating and structuring training data, distilling large models into smaller, production-ready alternatives, evaluating trade-offs across model architectures, and making data-driven accuracy-versus-cost decisions before deployment. Attendees will leave with concrete patterns for shipping efficient, specialized models into production. ### Playbook (editorial commentary) **The concept.** Distillation + fine-tuning a smaller model on your domain can beat frontier-model APIs on narrow tasks at a fraction of the cost (often 50–100×). **Why it matters.** Frontier model APIs are an OpEx tax that scales with usage. For specialised tasks (classification, extraction, narrow generation), you're paying for generality you don't use. **The hard parts.** Generating high-quality training data is the actual challenge. Fine-tuning is the easy bit. Eval is hard at this scale — you need a held-out set the trained model never sees. **Playbook moves.** (1) Use frontier models as labelers for the small model — they generate the training data. (2) Set explicit accuracy budgets, not just cost targets. (3) Plan for periodic refresh; data drifts. **The surprise.** For genuinely domain-specific tasks, a fine-tuned 7B-class model often *beats* a frontier model on the metric that matters — because it overfits to *your* distribution. That's not a bug; it's the feature you're paying for. --- ### Live monitored sources - [Top Tech News Today, May 1, 2026 - Tech Startups](https://techstartups.com/2026/05/01/top-tech-news-today-may-1-2026/) — techstartups.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [Releases · crewAIInc/crewAI · GitHub](https://github.com/crewAIInc/crewAI/releases) — github.com (2026-05-12): CrewAI (crewAIInc/crewAI) published release 1.14.5a4 (pre-release) on 2026-05-08. Highlights: updated LLM listings and dependency adjustments (moved textual dependency into crewai-cli and added certifi), several bug fixes and changelog/documentation updates. Migration implication - [2026 - TechCrunch](https://techcrunch.com/2026/) — techcrunch.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [aws.amazon.com](https://aws.amazon.com/about-aws/whats-new/2026/04/bedrock-openai-models-codex-managed-agents/) — aws.amazon.com (2026-04-29): Amazon Bedrock (AWS) now offers OpenAI models, Codex, and Managed Agents (Limited Preview) — announced 2026-04-28. What changed: OpenAI models and Managed Agents are available inside AWS Bedrock limited preview, letting AWS customers run OpenAI models and managed agent capabiliti - [Basata Raises $21M Series A to Expand AI Healthcare ...](https://www.citybiz.co/article/844227/basata-raises-21m-series-a-to-expand-ai-healthcare-operations-platform) — citybiz.co (2026-05-09): Phoenix-based healthcare AI startup Basata raised $21 million in Series A funding to expand its AI-driven healthcare operations automation tools. --- ## ANT301 — A practitioners guide to data for agentic AI URL: https://aws-summit-2026-kb.pages.dev/sessions/ANT301 Level: advanced Type: Breakout session Category: Analytics & Big Data Topics: Model Context Protocol (MCP); Data Governance & Privacy; Streaming & Real-Time Data; Generative AI & Foundation Models; Databases & Aurora; OpenSearch & Vector Search; Data Lakes, Lakehouse & AI-Ready Data; Agentic AI; Machine Learning & SageMaker; Retrieval Augmented Generation (RAG) In this session, gain the skills needed to deploy end-to-end agentic AI applications using your most valuable data. This session focuses on data management using processes like Model Context Protocol (MCP) and Retrieval Augmented Generation (RAG), and provides concepts that apply to other methods of customizing agentic AI applications. Discover best practice architectures using AWS database services like Amazon Aurora and OpenSearch Service, along with analytical, data processing and streaming experiences found in SageMaker Unified Studio. Learn data lake, governance, and data quality concepts and how Amazon Bedrock AgentCore and Bedrock Knowledge Bases, and other features tie solution components together. ### Playbook (editorial commentary) **The concept.** Agents need both fresh data (real-time signals) and rich context (historical, semantic). MCP and RAG are the two delivery mechanisms; they coexist rather than compete. **Why it matters.** A great agent on stale or wrong data is just confidently wrong, faster. The data stack is the constraint, not the model. **The hard parts.** Data quality, lineage, freshness, and access control are not solved by "buy a vector DB." They're solved by data engineering discipline that most orgs lack. **Playbook moves.** (1) Inventory data sources by latency tier (sub-second / minutes / hours / batch). Map agents to the tier they need. (2) Audit what data the agent *actually* uses, not what you intended. (3) Apply governance at the source, not at the agent — agents will route around perimeter controls. **The surprise.** RAG retrieval quality is dominated by chunking strategy, not embedding model. Boring but true. Spend a week on chunk size, overlap, and semantic boundaries before you spend a dollar on a fancier embedder. --- ### Live monitored sources - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## ARC301 — Build an AI-ready data foundation URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC301 Level: advanced Type: Breakout session Category: Architecture Topics: Agentic AI; Data Governance & Privacy; Generative AI & Foundation Models An unparalleled level of interest in generative AI and agentic AI is driving organizations to rethink their data strategy. While there is a need for data foundation constructs such as data pipelines, data architectures, data stores and data governance to evolve, there are business elements that need to stay constant like cost-efficiency and effectively collaborating across data estates. In this session we will cover how building your data foundation on AWS provides the tools and the building blocks to balance both needs, and empower organizations to grow their data strategy for building AI-ready applications. ### Playbook (editorial commentary) **The concept.** Foundation = pipelines + storage + governance + lineage. Without all four, AI initiatives are theatre. **Why it matters.** Every "AI strategy" deck has a token slide labeled "data foundation." That slide is actually 80% of the work. **The hard parts.** Data mesh vs. data lake vs. lakehouse debates miss the point. The point is *contracts between domains* — who owns what, what they guarantee, how breaking changes get communicated. **Playbook moves.** (1) Codify data contracts in writing, with SLAs. (2) Treat downstream AI as just another consumer subject to those contracts. (3) Invest in lineage tooling early; retrofitting later is brutal. **The surprise.** Cost-efficiency in data foundations comes from eliminating duplicate ingestion (the same data landing in three lakes), not from cheaper storage. Storage is rounding error in 2026; egress and re-processing are not. --- ### Live monitored sources - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios --- ## DAT304 — AI-Native by Design: How Deputy Rewired Its Operating Model on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/DAT304 Level: advanced Type: Breakout session Category: Databases Topics: Retrieval Augmented Generation (RAG) Ciaran Hale shares how Deputy moved from fragmented AI experiments to a centralised AI and data foundation on AWS. By standardising infrastructure, introducing governance guardrails, and building reusable components, Deputy enabled teams to scale AI with confidence. This shift also transformed internal adoptionembedding AI into everyday workflowsand unlocked new customer value through Deputy AI, delivering smarter automation and insights in-product. A practical blueprint for building an AI-native organisation. ### Playbook (editorial commentary) **The concept.** Going from "scattered AI experiments" to "AI-native operating model" requires a centralised platform team, not just decentralised exploration. Standards, governance guardrails, reusable components. **Why it matters.** Without standards, every team builds its own broken stack. You end up with five vector DBs, three orchestration frameworks, and zero shared evals. **The hard parts.** Platform teams without enforcement get ignored. Platform teams with enforcement become bottlenecks. Both fail. **Playbook moves.** (1) Make the golden path so easy that paving your own is the harder option. Reusable components beat policy memos. (2) Embed platform engineers in product teams for the first quarter. (3) Measure platform adoption as the team's KPI. **The surprise.** "AI-native" is mistakenly framed as a tech change. It's actually a *procurement* change. Your buying decisions need to weight "agent-friendly APIs" and "structured outputs" as first-class criteria. Half the AI roadblocks come from vendors whose APIs aren't built for this. --- ### Live monitored sources - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Notion launches its first AI agents for data analysis and task ... - MLQ.ai](http://mlq.ai/news/notion-launches-its-first-ai-agents-for-data-analysis-and-task-automation) — mlq.ai (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. --- ## MAM302 — Agentic AI for VMware migrations with AWS Transform for VMware URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM302 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Agentic AI; Migration & Modernization; Compute: EC2, Graviton & Nitro Accelerate your VMware migration journey with AWS Transform, the first agentic AI service for large-scale VMware workload migrations to Amazon EC2. Discover how to migrate from on-premises VMware infrastructure to a modernized, cloud-native architecture while overcoming challenges like evolving licensing models and vendor lock-in. Meet the team behind AWS Transform and see a live demonstration showcasing automated application discovery, dependency mapping, network translation, wave planning, and server migration with optimized EC2 instance selection. Learn practical approaches to streamline large-scale migrations and modernize VMware workloads to AWS with greater speed and confidence. ### Playbook (editorial commentary) **The concept.** Apply agentic AI to migration steps: discovery, dependency mapping, network translation, wave planning, cutover. **Why it matters.** Broadcom's licensing changes have made VMware-on-prem economically unstable for many shops. Migration timelines that used to be optional are now urgent. **The hard parts.** Discovery is the unsexy bottleneck. Most VMware estates have inaccurate inventories — the CMDB and reality diverged years ago. **Playbook moves.** (1) Run discovery first; defer migration decisions until the dependency graph is verified. Don't trust the CMDB. (2) Quarantine "tier 0" workloads for human review; let agents handle the long tail. (3) Build the rollback plan before the cutover. **The surprise.** Agent-driven migration shines on the *long tail* of small workloads, not the strategic flagship apps. Target tier-3/4 apps first to bank fast wins and build trust. The flagship workloads will need bespoke human attention regardless of tooling. --- ### Live monitored sources - [The agent control plane becomes the new enterprise buying surface](https://www.linkedin.com/pulse/agent-control-plane-becomes-new-enterprise-buying-andrew-mcpherson-gcd4f) — linkedin.com (2026-05-07): At Cloud Next 2026, Google committed $750 million to a partner fund designed to accelerate the development of agentic AI builds, supporting partners like Accenture and KPMG in scaling AI agent deployment. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Google launches $750M partner fund at Cloud Next 2026 to ...](https://thenextweb.com/news/google-cloud-750m-partner-fund-agentic-ai) — thenextweb.com (2026-05-07): At Cloud Next 2026, Google committed $750 million to a partner fund designed to accelerate the development of agentic AI builds, supporting partners like Accenture and KPMG in scaling AI agent deployment. - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [galvnews.com](https://www.galvnews.com/general-compute-launches-asic-first-inference-cloud-for-autonomous-ai-agents/article_1b97eb72-57ac-5b09-a820-dc6daadb2a64.html) — galvnews.com (2026-04-19): General Compute announced an ASIC-first inference cloud targeted at autonomous AI agents and agent workloads. Capabilities: infrastructure optimized for inference at scale for autonomous/agentic systems, likely to improve latency/cost for running many agents. Limitations: infrast --- ## MAM306 — Adding Agentic AI to legacy apps with Amazon Bedrock AgentCore URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM306 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Security, Identity & Compliance; Generative AI & Foundation Models; Serverless: Lambda & Step Functions; Agentic AI; Retrieval Augmented Generation (RAG) In this code-first session, we demonstrate how to add agentic AI capabilities and augment a legacy application using Amazon Bedrock AgentCore and the Amazon Strands Agents SDK. We will explore how to build AI-powered features for a legacy application without modifying the existing backend code. We will showcase how to leverage existing APIs and Lambda functions as the backbone for your agentic AI experience. You'll learn how to execute code in isolated sandbox environments, ensuring security while accessing internal data sources with Amazon Bedrock AgentCore Code Interpreter. ### Playbook (editorial commentary) **The concept.** You don't need to rewrite a legacy app to make it agentic. Wrap it. Use existing APIs and Lambda functions as the agent's tools; run untrusted code in sandboxed Code Interpreter environments. **Why it matters.** Most enterprise IP lives in legacy code. Rewriting kills you. Wrapping unlocks AI value without the rewrite tax. **The hard parts.** Legacy APIs lack the granularity agents need (too coarse, idempotency unclear, error handling inconsistent). Auth boundaries don't match agent needs. **Playbook moves.** (1) Inventory existing APIs; score them for agent suitability — idempotency, granularity, auth context, error semantics. (2) Build a thin agent-API adapter layer rather than expose internals. (3) Use Code Interpreter sandboxes for any agent that needs to compute, not call. **The surprise.** The Code Interpreter sandbox is the safest pattern most teams ignore. It lets you give agents *capability* without giving them *prod access*. Sandbox + result-passing handles 80% of the "agent needs to run code" problem with a fraction of the blast radius. --- ### Live monitored sources - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [How to Build an Agentic AI Strategy With Process Intelligence](http://skan.ai/blogs/process-intelligence-for-agentic-ai-enterprise-automation) — skan.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## DEV204 — AI-Powered EKS Troubleshooting with AWS DevOps Agent URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV204 Level: intermediate Category: Developer Tools Topics: Containers: EKS, ECS & Fargate; Observability & Monitoring; Media & Entertainment; Security, Identity & Compliance; Networking & Edge; DevOps, CI/CD & DevSecOps Managing EKS clusters means correlating logs, metrics, IAM policies, and network configurations under pressure. The AWS DevOps Agent, announced at re:Invent 2025, changes this workflow fundamentally. In this session, you'll watch a live demonstration where the DevOps Agent autonomously investigates an EKS service failuretracing issues from Pod logs through VPC Security Groups without manual intervention. You'll learn how the agent correlates cross-service dependencies, generates verified remediation plans, and integrates into existing SRE workflows. ### Playbook (editorial commentary) **The concept.** An LLM agent that does SRE-style triage: correlates Pod logs, IAM policies, VPC security groups, network configs, and proposes verified remediation. **Why it matters.** Mean time to resolution in Kubernetes is dominated by *correlation* across surfaces. Engineers spend the first 20 minutes of an incident gathering context. Agents are good at that. **The hard parts.** Generic SRE agents miss your idioms — your naming conventions, your custom controllers, your "we always restart that pod first" tribal knowledge. **Playbook moves.** (1) Curate your runbooks as MCP tools the agent can call. (2) Treat your existing playbooks as agent training data — version them. (3) Wire the agent into ChatOps so its proposals are visible to the team. **The surprise.** Agents are best at the boring 80% of incidents. The hard 20% they'll fumble — that's where humans still win. So measure success on *time-to-page-the-human*, not on full autoresolution. The agent's job is to short-circuit the easy stuff and hand off cleanly when it's stuck. --- ### Live monitored sources - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [draft-klrc-aiagent-auth-01 - AI Agent Authentication and ...](https://datatracker.ietf.org/doc/draft-klrc-aiagent-auth/) — datatracker.ietf.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Arcjet is en guarde in agentic workflows with Guard](http://computerweekly.com/blog/CW-Developer-Network/Arcjet-is-en-guarde-in-agentic-workflows-with-Guard) — computerweekly.com (2026-05-10): Arcjet introduced 'Guards,' a runtime security service for AI agent workflows that enables enforcement of per-user token budgets and spend limits inside agent loops and can detect prompt injection in tool results. - [gdplabs/gl-iam-cookbook · GitHub - GitHub](http://github.com/gdplabs/gl-iam-cookbook) — github.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management --- ## ISV302 — Architecting Scalable AI Agents using Amazon Bedrock AgentCore URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV302 Level: advanced Type: Lightning talk Category: ISV & Partners Topics: Observability & Monitoring; Generative AI & Foundation Models; Manufacturing & Industry 4.0; Analytics, Redshift & Generative BI; Agentic AI; Retrieval Augmented Generation (RAG) Discover how to build powerful AI agents using Amazon Bedrock's suite of tools, with a focus on Amazon Bedrock AgentCore. This session explores how Parrot Analytics leveraged the modular components of Amazon Bedrock AgentCore and Amazon Nova foundational models to achieve 10x the processing speed of manual classification across 2M+ entities. We will dive into prompt and context engineering, knowledge bases, and observability for production agentic workloads. ### Playbook (editorial commentary) **The concept.** Modular AgentCore components + Nova foundation models for high-throughput entity classification (10× speedup over manual on 2M+ entities). **Why it matters.** Classification at scale is a sweet spot for agents — well-defined output, clear evals, repetitive task. If your business has classification as a bottleneck, this pattern works. **The hard parts.** Prompt drift over millions of entities. Evals become statistical, not anecdotal. Inter-rater agreement (when humans review agent output) becomes a meaningful metric. **Playbook moves.** (1) Stratified sampling for evals — don't just look at random 100. (2) Re-eval on every prompt change. (3) Track inter-rater agreement between agent and humans over time. **The surprise.** Modular AgentCore decomposition lets you swap models per stage. Use a cheap model for triage ("is this even worth processing?"), a mid-tier for the bulk, and an expensive model only for ambiguous cases that fail confidence checks. Don't run uniform inference. The cost difference is 10×. --- ### Live monitored sources - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [How to Build an Agentic AI Strategy With Process Intelligence](http://skan.ai/blogs/process-intelligence-for-agentic-ai-enterprise-automation) — skan.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Devin AI Guide 2026: Features, Pricing, How to Use & Complete ...](https://aitoolsdevpro.com/ai-tools/devin-guide/) — aitoolsdevpro.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s --- ## STP205 — How Dovetail powers Multi-Tenant Agents with Vector Indexing at Scale URL: https://aws-summit-2026-kb.pages.dev/sessions/STP205 Level: intermediate Category: Startups Topics: OpenSearch & Vector Search; Retrieval Augmented Generation (RAG) When you're building multi-tenant vector search but can't control what customers throw at you, every indexing decision — isolation, embedding, chunking, partitioning — becomes a bet you're making blind, and this talk gives you the framework to make the right ones. ### Playbook (editorial commentary) **The concept.** Multi-tenant vector search has a four-axis decision tree: isolation, embedding choice, chunking strategy, partitioning. Each is a bet you're making blind unless you frame it explicitly. **Why it matters.** Customer A's bizarre 10,000-page PDF cannot be allowed to performance-tax customer B's queries. In B2B SaaS, this is non-negotiable. **The hard parts.** Tenant isolation in shared indexes is hard. Tenant-per-index is operationally simple but doesn't scale past a few hundred tenants. Tenant-aware filtering inside a shared index is harder to get right but is the only path that scales. **Playbook moves.** (1) Adopt tenant-aware filtering at the index level, not the application level. (2) Pre-decide who controls embedding model upgrades — you, or the customer? Both have valid arguments. (3) Cap per-tenant index size to bound worst-case latency. **The surprise.** Most teams optimise for retrieval *quality* and forget *tail latency*. A single cold tenant with a giant document set will kill P99 for everyone unless you isolate aggressively. Per-tenant query budgets are a feature, not a limitation. --- ### Live monitored sources - [prnewswire.com](https://www.prnewswire.com/news-releases/constructive-open-sources-agentic-db-the-postgres-memory-layer-for-ai-agents-302755269.html) — prnewswire.com (2026-04-29): Constructive announced and open-sourced "agentic-db": a purpose-built Postgres "memory layer" for AI agents providing long-term episodic memory, conversation and tool-call event logs, token accounting, a versioned skills/tools registry, rules/behavioral policies for governance, t - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Agentic RAG Explained: AI Agents + RAG in 2026](https://freeacademy.ai/blog/agentic-rag-ai-agents-supercharge-retrieval-2026) — freeacademy.ai (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [kr.teradata.com](https://kr.teradata.com/press-releases/2026/teradata-enables-ai-agents) — kr.teradata.com (2026-04-19): Teradata announced agentic, multi-modal capabilities for Teradata Enterprise Vector Store: unified structured+unstructured data, automated ingestion (docs/images/audio), multi-modal embeddings (up to 8K dims), hybrid/ fusion search (semantic + lexical + metadata), LangChain integ - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## DEV207 — Data Observability Without the Pain - Lessons from a Production System URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV207 Level: intermediate Category: Developer Tools Topics: Storage: S3, EBS & EFS; Observability & Monitoring; Media & Entertainment; Security, Identity & Compliance; IoT & Edge Computing; Serverless: Lambda & Step Functions; Manufacturing & Industry 4.0; Industry Spotlight: Healthcare & Life Sciences; Retrieval Augmented Generation (RAG) Modern IoT platforms are inherently data platforms. Events flow through APIs, queues, AWS Lambda Serverless functions, storage systems, and device networks before becoming meaningful data. When something goes wrong, tracing a single event across these distributed components quickly becomes painfuland the question shifts from _what happened_ to _where do I even start looking Ill walk through three practical observability patterns drawn from building and operating a production, event-driven IoT healthcare platform on AWS that processes tens of thousands of device events daily. Using OpenTelemetry, AWS X-Ray and Honeycomb, well explore techniques for gaining visibility into asynchronous event pipelines, correlating activity across services, and tracing events as they move through distributed systems. Youll leave with three concrete patterns you can apply immediately to your own event-driven data systems. ### Playbook (editorial commentary) **The concept.** Event-driven systems make tracing painful. OpenTelemetry + X-Ray + Honeycomb give end-to-end visibility across asynchronous IoT pipelines. **Why it matters.** When events traverse 5+ services (API → SQS → Lambda → S3 → device), "where do I look?" is the longest part of incident response. **The hard parts.** Async correlation. A single event ID needs to span SQS messages, Lambda invocations, S3 writes, device callbacks. Without enforcement, IDs get dropped at boundaries. **Playbook moves.** (1) Inject correlation IDs at the edge (the API or device boundary). (2) Log them at every hop. (3) Make trace assembly a one-query operation in your observability tool. **The surprise.** The biggest observability win is not tools — it's a *correlation ID standard* the team enforces. Pick one (the X-Ray trace ID is fine), enforce it everywhere, and stop debating. Tooling matters far less than you think once the IDs are consistent. --- ### Live monitored sources - [Edge Delta Makes All Telemetry Pipelines Data ...](http://prnewswire.com/news-releases/edge-delta-makes-all-telemetry-pipelines-data-throughput-limitless-and-free-for-all-customers-302736808.html) — prnewswire.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. --- ## ISV208 — From One Month to Two Days: How Xero Transformed Their DLC with AI URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV208 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Security, Identity & Compliance What if your engineering team could deliver a production-ready feature in two days that would traditionally take one month When Xero partnered with AWS, that's exactly what happened. In this session, we'll take you inside a real-world AI-Driven Development Life Cycle (AIDLC) workshop run against a live production use case. Working across an existing brownfield codebase, the Xero engineering team reached MVP in just two days against a feature that would have taken a month using traditional development methodology. We'll unpack what made that possible, what we learned along the way, and how those lessons are now shaping a plan to scale the AIDLC methodology across the entire engineering organisation. ### Playbook (editorial commentary) **The concept.** AI-Driven Development Lifecycle (AIDLC) — pair-programming with AI through the entire SDLC, not just code completion. Brownfield codebase, MVP in 2 days for what used to take a month. **Why it matters.** If a 30× delivery speedup is achievable on real production work — even partially — your release cadence assumptions need re-baselining. **The hard parts.** Brownfield is much harder than greenfield. AI hallucinates against bad context. The gain depends entirely on the quality of context you can feed the model. **Playbook moves.** (1) Invest in repo-level context — `CLAUDE.md` / `AGENTS.md` files, ADRs, architecture diagrams the AI can read. (2) Spec-first, not code-first. The spec is the leverage point. (3) Verify with tests, not vibes — and let the AI help write the tests. **The surprise.** The 90%+ time saving is mostly in *design + scaffolding*, not raw code typing. The lever is "decisions made fast," not "characters typed fast." Senior engineers who already produced fast scaffolds will see smaller gains than juniors who got stuck on architecture. --- ### Live monitored sources - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [draft-klrc-aiagent-auth-01 - AI Agent Authentication and ...](https://datatracker.ietf.org/doc/draft-klrc-aiagent-auth/) — datatracker.ietf.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## STP215 — How Sonder Improve 24/7 Employee Wellbeing with AWS AI URL: https://aws-summit-2026-kb.pages.dev/sessions/STP215 Level: intermediate Category: Startups Topics: Manufacturing & Industry 4.0; Industry Spotlight: Financial Services Sonder's mission is to empower people to be at their best, delivering the right care at the right time to over 1 million members globally across employers, universities, and insurance partners. To scale this mission without compromising quality, Sonder built an AI Copilot entirely on AWS, a 100% human-in-the-loop system designed with safety-first approach where Care Specialists remain at the centre of every interaction, with reinforcement learning from human feedback (RLHF) continuously improving the Copilot's performance. This session explores the architecture behind it, including real-time triage, knowledge retrieval, and escalation workflows and shares the tangible business impact: faster response times, greater operational efficiency, and new service and revenue opportunities unlocked by AI. ### Playbook (editorial commentary) **The concept.** Human-in-the-loop AI with reinforcement learning from human feedback (RLHF), in a safety-critical domain (mental health triage). Care Specialists remain at the centre of every interaction. **Why it matters.** You cannot ship autonomous AI in safety-critical contexts without massive liability exposure. HITL is the floor, not a stopgap. **The hard parts.** HITL only works if humans actually review. Reviewer fatigue, banner blindness, and rubber-stamping kill it within months unless you design against them. **Playbook moves.** (1) Surface uncertainty explicitly in the UI — visually distinguish low-confidence cases. (2) Track human override rate as a quality signal. Override rate trending to zero means humans have stopped reviewing. (3) Audit a sample of low-override cases monthly. **The surprise.** RLHF without operational rigor is just opinion-collection dressed up. The signal quality of human feedback drops fast under load — tired reviewers click through. Audit your reviewers, not just your model. The reviewer pool's calibration *is* your safety budget. --- ### Live monitored sources - [How Fortune 500 Companies Are Moving Agentic AI Into Production](https://ai2.work/blog/how-fortune-500-companies-are-moving-agentic-ai-into-production) — ai2.work (2026-05-05): Anthropic has deployed production AI agents using Claude Opus 4.7 at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa for high-stakes financial workflows including KYC, underwriting, and insurance claims. AIG reported that these agents scored 88% as accurate as human experts on - [Fortune 500 Deploy Agentic AI in 2026 | Gheware DevOps AI Blog](https://devops.gheware.com/blog/posts/fortune-500-agentic-ai-deployment-playbook-2026.html) — devops.gheware.com (2026-05-05): Anthropic has deployed production AI agents using Claude Opus 4.7 at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa for high-stakes financial workflows including KYC, underwriting, and insurance claims. AIG reported that these agents scored 88% as accurate as human experts on - [Anthropic deepens Wall Street push with new AI agents, and ...](https://fortune.com/2026/05/05/anthropic-wall-street-financial-services-agents-jamie-dimon/) — fortune.com (2026-05-05): Anthropic has deployed production AI agents using Claude Opus 4.7 at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa for high-stakes financial workflows including KYC, underwriting, and insurance claims. AIG reported that these agents scored 88% as accurate as human experts on - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [80% of Fortune 500 use active AI Agents: Observability ...](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/) — microsoft.com (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 --- ## AIM302 — Agentic AI Meets Responsible AI - Science, Strategy and Practice URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM302 Level: advanced Type: Breakout session Category: AI & Machine Learning Topics: Agentic AI AI agents offer powerful capabilities — and introduce fundamentally new risks that require more than traditional controls. This session explores responsible agentic AI through three lenses: the science, the framework, and a real-world customer story. Understand the scientific frontiers that make agents different — from emergent behaviour and agent-to-agent trust to the challenges of governing systems that plan, negotiate, and act autonomously. Learn the four areas of the AWS Responsible AI framework where agents change the rules, and hear how one of Australia's leading health insurer is putting responsible AI into practice — from strategy to governance to real-world trade-offs. ### Playbook (editorial commentary) **The concept.** Agents introduce fundamentally new risks: emergent behaviour, agent-to-agent trust, autonomous action. Old controls (designed for predictive ML: input → score → decision) don't fit. **Why it matters.** Your existing AI risk framework was designed for the predictive era. Agents *plan*, *negotiate*, *act*. The audit surface is different. **The hard parts.** How do you audit a system that plans? Logs of what happened don't explain *why*. The reasoning is the artefact you need to capture, and the reasoning is in natural language and tool calls, not structured logs. **Playbook moves.** (1) Capture agent reasoning (tool calls, plans, decisions, intermediate scratchpads), not just outputs. (2) Build an agent-specific risk taxonomy — most existing AI risk frameworks won't translate. (3) Create a clear escalation path for agent decisions that exceed predefined authority levels. **The surprise.** The hardest agent risk isn't "agent does bad thing" — it's "agent confidently fabricates evidence to *justify* the bad thing." Your auditing must catch the justification narrative, not just the action. Agents can produce coherent-looking reasoning for incorrect actions, and that reasoning is what humans will trust. --- ### Live monitored sources - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [ServiceNow expands AI agent governance through deeper ...](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-expands-AI-agent-governance-through-deeper-integration-with-Microsoft/default.aspx) — newsroom.servicenow.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes --- ## AIM303 — AWS Security Agent: Proactive AppSec from Design to Deployment URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM303 Level: advanced Type: Breakout session Category: AI & Machine Learning Topics: Security, Identity & Compliance; Retrieval Augmented Generation (RAG) Application security teams face an impossible challenge: scale security expertise across growing application portfolios while maintaining development velocity. Traditional approaches force organizations to choose between speed and security. In this session, discover how AWS Security Agent transforms application security from reactive to proactive through AI-powered automation. Learn how this frontier agent conducts automated security reviews customized to your organizational requirements and delivers on-demand penetration testing tailored to your applications. Join us to see how you can scale security coverage and prevent vulnerabilities early in the development lifecycle while maintaining the speed of modern development. ### Playbook (editorial commentary) **The concept.** Automated security review + on-demand penetration testing, customised to your organisation's policies and threat model. **Why it matters.** AppSec teams are perpetually under-resourced. Automating the "tier 1" review work gives you real leverage without compromising rigor. **The hard parts.** Customisation matters — generic checks miss your specific threat model and create alert fatigue. Out-of-the-box config is rarely good enough. **Playbook moves.** (1) Encode your threat model as agent prompts/configurations. Treat them as code. (2) Re-run on every PR, not just every release. (3) Tune the false-positive rate aggressively — alert fatigue kills adoption. **The surprise.** AI pentest finds *different* bugs than human pentest. Don't replace; complement. AI is great at exhaustive coverage (every endpoint, every parameter combination); humans are great at creative chaining (using bug A to enable bug B to escalate to bug C). Run both, expect different findings, don't be surprised when they don't overlap much. --- ### Live monitored sources - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Agentic Identity and Access Management](https://www.coalitionforsecureai.org/wp-content/uploads/2026/04/agentic-identity-and-access-control.pdf) — coalitionforsecureai.org (2026-05-08): New implementation patterns for AI agent identity (updated May 6, 2026) highlight the convergence of the Model Context Protocol (MCP) for agent-server handshakes and OAuth 2.1 with Dynamic Client Registration (DCR) for runtime credential issuance. A key pattern is the use of 'dis - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [proofpoint.com](https://www.proofpoint.com/us/products/ai-mcp-security) — proofpoint.com (2026-05-01): Proofpoint provides an MCP Security Platform to secure AI connectivity at scale by routing MCP traffic through a central gateway. The platform enables centralized discovery and risk classification of 'shadow' MCP servers, enforces authentication via OAuth 2.0, controls user and a - [Prompt Injection Attack to Tool Selection in LLM Agents](https://www.ndss-symposium.org/wp-content/uploads/2026-s675-paper.pdf) — ndss-symposium.org (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf --- ## ARC305 — Transforming from SaaS to multi-tenant agentic SaaS URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC305 Level: advanced Type: Breakout session Category: Architecture Topics: Agentic AI; Model Context Protocol (MCP); Retrieval Augmented Generation (RAG) Existing SaaS providers must determine how and where agents best fit into their offerings. Getting there requires organizations to transform existing IP and functionality into agent-powered experiences. This breakout will dig into the details of this transformation, examining the patterns, strategies, and techniques that can be used to introduce agents into an existing multi-tenant system. Well focus heavily on identifying the target agents, digging into how/where theyre built and introduced, how theyre integrated, and so on. Well also dig into how multi-tenancy lands in new agents, integrating with MCP servers, using RAG, applying tenant isolation, supporting onboarding, and on on. ### Playbook (editorial commentary) **The concept.** Existing multi-tenant SaaS platforms bolt on agents, requiring new thinking on tenant isolation, MCP server scoping, RAG, onboarding, and pricing. **Why it matters.** If a competitor ships agents into your category before you do, you've ceded the AI-native narrative. Renewals get harder. **The hard parts.** Tenant isolation in agent state — memory, vectors, available tools — is genuinely novel. Your existing isolation model was designed for HTTP requests, not stateful agent conversations. **Playbook moves.** (1) Build tenant-scoped MCP servers — never share an MCP across tenants. (2) Audit every tool call for tenant boundary correctness. (3) Re-think onboarding: agents need access to tenant data, which means new permission flows. **The surprise.** The right pricing model for agentic SaaS isn't seats or usage — it's *outcome-based* (per resolved ticket, per successful workflow, per accurate document processed). Almost no SaaS has the billing infra to do this. Start the billing rebuild now, not after launch. Your finance team will become a bottleneck if you don't. --- ### Live monitored sources - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [GitHub - agentgateway/agentgateway: Next Generation Agentic Proxy for AI Agents and MCP servers · GitHub](https://github.com/agentgateway/agentgateway) — github.com (2026-05-07): Agentgateway released version v1.2.0-alpha.1, continuing the development of its open-source AI-native proxy for agent-to-agent and agent-to-tool communication. The project maintains approximately 2.6k GitHub stars and is part of the Linux Foundation. - [Rate Limiting Controls | GitHub Agentic Workflows](https://github.github.com/gh-aw/reference/rate-limiting-controls/) — github.github.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits - [What’s New in Agent 365: May 2026 | Microsoft Community Hub](https://techcommunity.microsoft.com/blog/agent-365-blog/what%E2%80%99s-new-in-agent-365-may-2026/4516340) — techcommunity.microsoft.com (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. --- ## COP301 — Elevating your Agentic AI Observability URL: https://aws-summit-2026-kb.pages.dev/sessions/COP301 Level: advanced Type: Breakout session Category: Other Topics: Voice & Conversational AI; Observability & Monitoring; Databases & Aurora; Analytics, Redshift & Generative BI; Agentic AI Gain deep visibility into the performance and reliability of autonomous agents with Amazon CloudWatch. This session showcases how CloudWatch delivers endtoend observability for agentic AI workloadstracking decision quality, token efficiency, and workflow execution at scale. Explore prebuilt dashboards and advanced metrics that help you optimize agent performance, control operational costs, and maintain consistent behavior across complex intelligent systems. Walk away ready to implement productiongrade observability that ensures your AI agents operate reliably, make optimal decisions, and deliver measurable outcomes at scale. ### Playbook (editorial commentary) **The concept.** Beyond logs and metrics — observability for agent decisions, token efficiency, multi-step workflow execution, and decision drift over time. **Why it matters.** Agents are stateful, multi-step, and expensive. Traditional APM misses the cost dimension entirely and treats decision quality as opaque. **The hard parts.** Defining the right metrics is non-obvious. "Request rate" doesn't mean what it used to. "Decision quality" isn't directly measurable. **Playbook moves.** (1) Track four axes minimum: cost per task, success rate, latency per task (end-to-end, not per LLM call), decision drift over time. (2) Build dashboards per agent, not per service. (3) Alert on cost spikes, not just error spikes. **The surprise.** The metric most agentic systems should track and don't is *loop count* — how many tool calls per completed task. It's the canary for prompt regression, model drift, and broken tools. When loop count starts trending up week-over-week, something is wrong even if all your other metrics look fine. --- ### Live monitored sources - [Gartner 2026 Confirms It: The Context Graph Is the Missing ...](https://thecontextgraph.co/memos/gartner-2026-ai-agents-decision-intelligence-sales) — thecontextgraph.co (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [AI Agent Context Window Cost: Why Bills Multiply [2026]](https://www.waxell.ai/blog/ai-agent-context-window-cost) — axell.ai (2026-05-05): Waxell published an analysis on the compounding cost of AI agent context windows, detailing how naive history management leads to 3x-5x budget underestimation. They proposed a runtime enforcement architecture (Waxell Runtime) that operates in the execution path to enforce hard to - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes --- ## DVT201 — Building Software like never before with agentic AI URL: https://aws-summit-2026-kb.pages.dev/sessions/DVT201 Level: intermediate Type: Breakout session Category: Other Topics: Agentic AI; Code Generation & AI-Assisted Development; Kiro & Spec-Driven Development Discover the future of software development as we explore the transformative power of agents with Kiros spec-driven development in modernizing your entire software development lifecycle (SDLC). Agentic AI is impacting software development by automating tasks, improving efficiency, and enabling more autonomous workflows throughout the development lifecycle. It allows teams to go beyond simple code generation to handle project planning, designing, testing, documentation, building agents into workflows, and retiring technical debt. Join us to explore how these powerful capabilities work together to help organizations accelerate from prototype to production ready applications. ### Playbook (editorial commentary) **The concept.** Spec-driven development — you write specs; the agent builds code, tests, docs, infrastructure. Code generation got commoditised; spec quality is the differentiator. **Why it matters.** If everyone has the same code-gen models, the engineering moat shifts to who can describe what they want most precisely. **The hard parts.** Specs are hard to write well. Most engineers haven't been trained on it. The skill is closer to product management than to coding. **Playbook moves.** (1) Treat spec authoring as a senior-engineer activity initially; pair juniors with seniors. (2) Version specs in the same repo as the code. Review them like code. (3) Build a library of "spec patterns" for common features. **The surprise.** The team skill that matters most in spec-driven development is *product clarity*, not coding skill. Your PMs and tech leads become the bottleneck. Plan for that — invest in PM training on spec writing now, before the AI tooling forces the issue. --- ### Live monitored sources - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [Red Hat adds support for agentic AI development | CIO](https://www.cio.com/article/4169833/red-hats-message-to-enterprises-you-dont-need-to-re-platform-for-ai-agents-2.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs --- ## MAM301 — From tech debt to competitive advantage: Migrate & modernize with AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM301 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Agentic AI; Migration & Modernization Your legacy applications and infrastructure don't have to be a liability. In this session, discover how leading enterprises are converting tech debt into modern cloud advantages that unlock AI capabilities, cut costs, and accelerate innovation. The path to AI starts with migration and modernization. We'll walk through the complete transformation journey: infrastructure, applications, data, and AI-ready business outcomes. You'll see how new agentic AI automation is making migration and modernization faster, simpler, and more cost-effective. We'll explore proven pathways for your most critical workloads (Microsoft, VMware, SAP, and mainframe), backed by 20 years of AWS expertise. ### Playbook (editorial commentary) **The concept.** Migration + modernization unlocks AI capabilities that legacy infrastructure can't run. The path to AI starts with migration. **Why it matters.** AI workloads need scale and architectural primitives that mainframes and on-prem VMware don't provide. Migrating *is* the AI strategy in many cases. **The hard parts.** Mainframe migration is risk-laden. "Move-and-improve" is hype; lift-and-shift just becomes technical debt with new bills. **Playbook moves.** (1) Score workloads on AI-readiness (data accessibility, API surface, ability to scale horizontally). (2) Migrate high-AI-leverage workloads first, even if they're not the cheapest. (3) Build the AI use case for each workload before migrating it — that frames the modernisation choices. **The surprise.** The best migration ROI argument to the board isn't cost savings — it's *capability unlock*. Frame the budget ask accordingly: "we cannot do X, Y, Z on the current stack." Cost-only arguments lose to "wait six months and revisit." Capability arguments don't. --- ### Live monitored sources - [IBM Consulting Expands AI Capabilities to Accelerate Enterprise Transformation](https://newsroom.ibm.com/2026-05-06-ibm-consulting-expands-ai-capabilities-to-accelerate-enterprise-transformation) — newsroom.ibm.com (2026-05-08): IBM announced an expansion of its AI capabilities through 'IBM Enterprise Advantage' and 'IBM Consulting Advantage,' including the 'Agent2Agent (A2A)' interoperability standard to allow multi-agent orchestration across enterprise ecosystems (e.g., watsonx Orchestrate and SAP's Jo - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Agentic AI at Scale: The Rackspace Story](http://rackspace.com/resources/agentic-ai-scale-rackspace-story) — rackspace.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## MAM303 — Digital transformation excellence using agentic AI URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM303 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Agentic AI; Migration & Modernization; Retrieval Augmented Generation (RAG) Discover how customers are leveraging AWS AI-driven solutions to accelerate their cloud transformation journey, moving beyond traditional migration and modernization to achieve digital transformation excellence. This session showcases real-world experiences where organizations have integrated AI-powered accelerators with AWS cloud services to achieve 3x faster migration timelines and drive unprecedented business value. Learn actionable strategies and see demonstrations of how a large transformation project can become a digital enterprise enabler for your organization, ### Playbook (editorial commentary) **The concept.** AI accelerators applied to migration produce 3× faster timelines via automated discovery, dependency mapping, test generation, and cutover dry-runs. **Why it matters.** Migration projects historically run long because of human-bottlenecked discovery and cutover planning. Agents compress those phases dramatically. **The hard parts.** 3× is achievable on greenfield-ish migrations; brownfield with regulatory baggage gets less. The 3× is real but conditional. **Playbook moves.** (1) Apply agents to discovery first, then test generation, then cutover dry-runs. (2) Don't try to AI-ify the political and stakeholder negotiations — that's still human work. (3) Pre-approve migration patterns at the architecture review board so individual migrations don't restart approval cycles. **The surprise.** The biggest barrier to fast migration isn't tech — it's stakeholder approval cycles. AI doesn't help there directly. The transformation programs that move fastest *pre-approve* common patterns in advance, so individual migrations bypass the slow path. That's an org-design move, not a tech move. --- ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and --- ## DEV205 — Securing Amazon Bedrock AgentCore: A Practical Framework URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV205 Level: intermediate Category: Developer Tools Topics: Agentic AI; Security, Identity & Compliance; Generative AI & Foundation Models Explore a practical framework to think about and build secure AI agents on Amazon Bedrock AgentCore. This session covers threat modeling specific to agentic workloads, including how agents interact with tools, memory, and external systems, and what you need to watch out for. You'll learn how to apply AWS security best practices across AgentCore services, and walk away with actionable patterns. Suitable for developers and architects building AI agents who want to move from prototype to production with confidence. ### Playbook (editorial commentary) **The concept.** Threat-modeling agentic workloads specifically — tool surface, memory, external system interactions. Apply AWS security best practices across AgentCore components. **Why it matters.** Agent threat models differ meaningfully from API threat models. Your existing security playbook misses tool injection, memory poisoning, and prompt-via-data attacks. **The hard parts.** Tool injection (an agent tricked into calling a tool with malicious args). Memory poisoning (an agent's long-term memory corrupted by an attacker). Prompt-via-data (instructions embedded in retrieved content). **Playbook moves.** (1) Run STRIDE per agent component (memory, tool registry, planner, executor). (2) Specifically check: who can mutate memory, who can register new tools, who validates tool outputs before they're trusted. (3) Sandbox every tool execution that touches external systems. **The surprise.** The most common agent vulnerability in real deployments is *over-broad tool scopes*. Engineers grant tools more permission than needed because it's faster. Audit tool permissions like you audit IAM. Most don't, and it shows up as blast radius when things go wrong. --- ### Live monitored sources - [CISA, US and International Partners Release Guide to Secure ...](https://www.cisa.gov/news-events/news/cisa-us-and-international-partners-release-guide-secure-adoption-agentic-ai) — cisa.gov (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Comment and Control: GitHub AI Agents as Credential ...](https://labs.cloudsecurityalliance.org/research/csa-research-note-comment-control-github-prompt-injection-20/) — labs.cloudsecurityalliance.org (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Decision Traces for Agentic Operations: Why Agents Need ...](https://xmpro.com/decision-traces-for-agentic-operations-why-agents-need-operational-memory/) — xmpro.com (2026-05-08): XMPro introduced the concept of 'operational memory' powered by decision traces, which capture the reasoning behind specific actions (including exceptions and human judgment) rather than just general rules. This is implemented via a decision trace layer in the orchestration path --- ## ISV210 — Boost performance and reduce costs with Aurora: Canva's story URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV210 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Observability & Monitoring; Migration & Modernization; Databases & Aurora From initial idea to executionlearn about Canva's journey migrating MySQL workloads from Amazon RDS to Aurora at scale. Discover how Canva achieved meaningful performance improvements, cost savings, and operational efficiencies through this strategic migration. This lightning talk shares real-world insights on planning and executing large-scale database migrations, key Aurora best practices for optimizing cost and performance, and how the latest monitoring features help maintain efficiency as you scale. Learn how AWS Countdown Premium (CDP) accelerated and de-risked Canva's migration, delivering tangible business value while minimizing operational disruption. ### Playbook (editorial commentary) **The concept.** RDS MySQL → Aurora migration at scale. Performance gains, cost savings, operational efficiencies. Real-world insights on planning and de-risking. **Why it matters.** At Canva-scale data volumes, even single-digit % gains translate to millions per year. The migration math compounds. **The hard parts.** Schema changes, connection pool management, replication lag during cutover, application-level retries. Aurora isn't a drop-in for MySQL in every case. **Playbook moves.** (1) Use Aurora Backtrack as your safety net during cutover — it's a proper "undo" you don't get with MySQL. (2) Run a failover drill before production cutover, not after. (3) Right-size connection pools — Aurora's connection limits behave differently. **The surprise.** The most underrated Aurora feature isn't performance — it's *storage decoupling for clones*. You can spin up a full prod-fidelity clone in minutes for testing, with no storage cost duplication. Use clones for staging, load testing, and pre-prod validation. Most teams underuse this. --- ### Live monitored sources - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Decision Traces: Essential AI Infrastructure for Enterprise Scale](https://atlan.com/know/what-are-decision-traces-for-ai-agents/) — atlan.com (2026-05-06): Core idea: The Context Graph resource guide defines context graphs as a 'living record of decision traces' used for execution validation and temporal reasoning. It distinguishes them from knowledge graphs (static entities) and vector databases (semantic similarity) by their nativ - [Arango News & Press - ArangoDB](http://arango.ai/news-press) — arango.ai (2026-05-08): XMPro introduced the concept of 'operational memory' powered by decision traces, which capture the reasoning behind specific actions (including exceptions and human judgment) rather than just general rules. This is implemented via a decision trace layer in the orchestration path - [7 best tools for debugging AI agents in production (2026)](https://www.braintrust.dev/articles/best-ai-agent-debugging-tools-2026) — braintrust.dev (2026-05-10): Microsoft moved Agent 365 from preview to general availability as a unified control plane to discover, monitor, and govern AI agents across clouds and third-party SaaS; emphasis on discovering and controlling "shadow" local agents installed by employees. Pricing details not discl - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca --- ## STP201 — Scaling Security at Startup Speed: Hnry's AI-Powered Approach URL: https://aws-summit-2026-kb.pages.dev/sessions/STP201 Level: intermediate Category: Startups Topics: Security, Identity & Compliance; Startups & Innovation; Industry Spotlight: Financial Services Hnry, a fast-growing fintech startup, faced a critical challenge: their lean engineering team lacked capacity for thorough security reviews and penetration testing. Traditional solutions like hiring security specialists or engaging consulting firms would compromise their velocity and budget. By adopting AWS Security Agent, Hnry gained enterprise-grade security capabilities without enterprise overhead. The AI-powered tool integrated into their development workflow, providing real-time security feedback, automated design reviews, and on-demand penetration testing. This enabled Hnry to maintain startup speed while achieving robust security posture, proving small teams can deliver enterprise-level security through intelligent automation. ### Playbook (editorial commentary) **The concept.** Lean fintech replaces hiring AppSec specialists with an AI security agent — design reviews, real-time PR feedback, on-demand penetration testing. **Why it matters.** SOC2 and regulatory requirements demand security review even for small companies, but security talent is unaffordable for early-stage startups. **The hard parts.** Auditors want evidence in specific formats. The AI agent's output must map to those formats — generic reports get rejected. **Playbook moves.** (1) Configure agent output to match your audit evidence formats (the auditor's checklist, not yours). (2) Treat the AI as a junior, not a senior — review its critical findings before acting. (3) Track agent suggestions you accepted vs. dismissed; use that as a feedback loop. **The surprise.** The strongest argument for AI security in startups isn't cost — it's *consistency*. A junior human's first 6 months are inconsistent (skipped checks, varied rigor). The AI is consistent from day one. For a startup chasing SOC2, that consistency is what auditors actually reward. --- ### Live monitored sources - [ServiceNow expands AI agent governance through deeper ...](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-expands-AI-agent-governance-through-deeper-integration-with-Microsoft/default.aspx) — newsroom.servicenow.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Guild Raises $44M to Build the Agent Control Plane](https://www.guild.ai/knowledge/guild-raises-44m-agent-control-plane) — guild.ai (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Prompt Injection Attack to Tool Selection in LLM Agents](https://www.ndss-symposium.org/wp-content/uploads/2026-s675-paper.pdf) — ndss-symposium.org (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [Tekst Raises $13.5 Million Series A | The SaaS News](http://thesaasnews.com/news/tekst-raises-13-5-million-series-a) — thesaasnews.com (2026-05-11): China's Ministry of Industry and Information Technology (MIIT) released a national standard for AI terminal intelligence grading, which includes the implementation of the smart cockpit level 3 standard. - [When prompts become shells: RCE vulnerabilities in AI agent ...](https://www.microsoft.com/en-us/security/blog/2026/05/07/prompts-become-shells-rce-vulnerabilities-ai-agent-frameworks) — microsoft.com (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf --- ## GHJ301 — R1 — AWS Game Day : Secret Agent Unicorns URL: https://aws-summit-2026-kb.pages.dev/sessions/GHJ301 Level: advanced Type: Gamified learning Category: Other Topics: Agentic AI; Gaming & Interactive Media AWS Game Day : Secret Agent UnicornsAWS GameDay is a gamified learning event that challenges participants to use AWS solutions to solve real-world technical problems in a risk-free setting. As a new hire at Unicorn.Rentals, the worlds largest mythical creature rental company, youll test your AWS knowledge in an interactive, team-based, risk-free environment! The Secret Agentic Unicorns GameDay covers the different components of AgentCore, guiding participants through agent creating using Strands, A2A, and other AgentCore technologies. Youll get real world experience creating and learning about AgentCore agents, and have fun along the way. ### Playbook (editorial commentary) **The concept.** Hands-on AgentCore lab — Strands SDK, agent-to-agent (A2A) protocols, agent building under simulated incident pressure. **Why it matters.** You don't learn agents from slides. You learn by debugging them at 2 AM when something is broken and the demo gods have fled. **The hard parts.** GameDays are time-boxed. Trade-offs and shortcuts mirror real life — that's the point. **Playbook moves.** (1) Send 2–3 engineers, not one. The team dynamics are part of the lesson. (2) Debrief afterwards as a written internal post-mortem. (3) Pick people who'll bring lessons back to their teams, not just enthusiasts. **The surprise.** GameDay's hidden value is meeting *other practitioners*. The networking IS the curriculum. Hallway conversations during breaks generate more "huh, we should try that" than the labs themselves. Optimise for that. --- ### Live monitored sources - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## DEV313 — From Timeout to Throughput: Scaling Resilient Agentic Systems URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV313 Level: advanced Category: Developer Tools Topics: Agentic AI; Generative AI & Foundation Models Moving an AI agent from prototype to production requires more than optimism. This session tackles the "Day 2" engineering challenges of scaling resilient agentic architectures on AWS. Learn practical patterns for handling traffic spikes, optimizing throughput, and controlling costs using Amazon Bedrock models and AgentCore Runtime. We'll cover tool filtering strategies, when multi-agent architectures make sense, how to apply evaluations effectively, and how to harden your APIs against real-world load. Leave with concrete techniques to transform brittle GenAI prototypes into production-grade systems that survive viral launches and demanding enterprise workloads. ### Playbook (editorial commentary) **The concept.** Day-2 ops for agentic systems — traffic spikes, throughput optimisation, resilience under load. Tool filtering, multi-agent architectures, evals, API hardening. **Why it matters.** Viral launches kill prototypes. The Hacker News front page has destroyed more agent demos than bugs have. **The hard parts.** Tool fan-out (agent calls 20 tools in parallel and overwhelms downstream). Retry storms when tools fail. Multi-agent coordination overhead. Backpressure that doesn't propagate cleanly. **Playbook moves.** (1) Tool filtering — don't expose all tools to all queries. Reduces token cost and decision space. (2) Circuit breakers per tool, not just per service. (3) Async fan-in patterns where possible. **The surprise.** When to use multi-agent vs. single-agent: only when single-agent context is genuinely too big OR when subdomains have different model needs. Otherwise, multi-agent is just complexity tax dressed up as architecture. The default should be single-agent until you can name the specific reason it's not enough. --- ### Live monitored sources - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Fetched web page](https://mem0.ai/blog/6-techniques-to-cut-ai-agent-memory-cost-beyond-basic-retrieval) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## ISV303 — From hours to minutes: SafetyCulture's journey to 90% faster analytics URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV303 Level: advanced Type: Lightning talk Category: ISV & Partners Topics: Voice & Conversational AI; Data Lakes, Lakehouse & AI-Ready Data; Analytics, Redshift & Generative BI Discover how SafetyCulture, the global workplace operations platform used by 70,000+ organizations, achieved a 90% reduction in daily data pipeline execution time while processing the same data volumes on Amazon Redshift. Join this session with SafetyCulture data engineering team to learn how their team transformed a complex, slow-running data warehouse into a high-performance, AI-ready analytics platform using modern lakehouse architecture principles. ### Playbook (editorial commentary) **The concept.** Lakehouse modernisation on Redshift achieving 90% reduction in pipeline execution time on the same data volumes. Modern lakehouse architecture principles. **Why it matters.** Pipelines that take all night kill iteration speed for analysts and data scientists. AI workloads need fresh data, not yesterday's. **The hard parts.** Old SQL doesn't translate cleanly to lakehouse patterns. Materialisation strategies, incremental loading, partition design — all need rethinking. **Playbook moves.** (1) Identify "tall pole" pipelines (the slowest 5%) first — they're usually 50%+ of total runtime. (2) Rewrite those for materialised views + incremental loads, not full table scans. (3) Measure end-to-end, not per-job. **The surprise.** Most pipeline 90% wins come from *removing redundant computation* (the same aggregation computed three times across three pipelines), not from faster compute. Profile before re-platforming. You'll find that half the pipelines are doing the same work. --- ### Live monitored sources - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Event Sourcing for Agents: Log-Based Architecture for ...](https://understandingdata.com/posts/event-sourcing-agents/) — understandingdata.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ --- ## STP210 — TeamForm's Generative Dashboards with Strands & Bedrock AgentCore URL: https://aws-summit-2026-kb.pages.dev/sessions/STP210 Level: intermediate Category: Startups Topics: Containers: EKS, ECS & Fargate; Databases & Aurora; Manufacturing & Industry 4.0; Analytics, Redshift & Generative BI; Agentic AI Most teams are still piloting AI - TeamForm is shipping it. In this session, we show how we built enterprise and production-ready generative dashboards in weeks on AWS Bedrock and AgentCore, and how an AI-native operating model made that velocity possible. Learn what it actually takes to operationalise AI across product and engineering, not just prototype it. ### Playbook (editorial commentary) **The concept.** Generative dashboards — users describe what they want; the agent constructs the dashboard, query, visualisation. Built in weeks on Bedrock + AgentCore. **Why it matters.** Static dashboards don't keep up with question variety. The BI team's backlog is the bottleneck for analytical insight in most organisations. **The hard parts.** Trust. Users won't believe a dashboard the AI generated unless they can audit how it was built. Hallucinated metrics are catastrophic. **Playbook moves.** (1) Show the SQL/spec the agent generated — make it auditable. (2) Let users edit. Treat the agent as a *draft generator*, not the final authority. (3) Govern the underlying data model rigorously — the agent only as good as the schema it queries. **The surprise.** Generative dashboards work because you can drop the BI team's backlog. The hidden cost: someone needs to govern the *data model* the agent queries. That used to be the BI team's job. If you displace BI without replacing the governance, you get fast-but-wrong dashboards. Reassign the data modelling responsibility before you ship. --- ### Live monitored sources - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Devin AI Guide 2026: Features, Pricing, How to Use & Complete ...](https://aitoolsdevpro.com/ai-tools/devin-guide/) — aitoolsdevpro.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s --- ## PRT102-S — Efficiency to Innovation: How Agentic AI Unlocks New Business Models URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT102-S Level: foundational Type: Breakout session Category: Partner Showcase Topics: Agentic AI This joint discussion explores how organizations are moving from simplistic operational AI to agentic AI that drives innovation and new revenue. Through industry use cases and platform strategies, it shows how AI-enabled systems are disrupting markets and reshaping competitive advantage, with practical guidance for growth. ### Playbook (editorial commentary) **The concept.** Move from operational AI (cost cutting, automation) to innovation AI (new revenue streams, new products). Industry use cases and platform strategies. **Why it matters.** Cost cutting hits a floor — you can only cut so far. Revenue ceilings are higher. The companies winning with AI are the ones building new products, not just optimising old ones. **The hard parts.** New product lines need new go-to-market motions, not just new tech. The hardest part of an AI-native product is selling it, not building it. **Playbook moves.** (1) Pick one ambitious AI-native product per year. Resource it as a startup-within-a-company. (2) Don't dilute it with feature work for existing products. (3) Set a 12-month kill criterion — if it hasn't found product-market fit, kill it cleanly. **The surprise.** The companies that win with agentic AI will look more like *platforms* than products. Plan to expose APIs that *your customers' agents* can call. The customers aren't buying your UI in two years; they're letting their agents shop your APIs. --- ### Live monitored sources - [CISA, US and International Partners Release Guide to Secure ...](https://www.cisa.gov/news-events/news/cisa-us-and-international-partners-release-guide-secure-adoption-agentic-ai) — cisa.gov (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes --- ## PRT201-S — Postman and the Future of AI-Driven API Development in 2026 URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT201-S Level: intermediate Type: Breakout session Category: Partner Showcase Topics: Agentic AI; Model Context Protocol (MCP); Kiro & Spec-Driven Development; Retrieval Augmented Generation (RAG) Software development has fundamentally changed in 2026, driven by vibe coding, AI agents, and RAG/MCP. APIs are the interface layer for AI systems to perform meaningful work. For this to succeed, your APIs must be discoverable, consistent, and usable by both developers and agents. Postman is now central to designing, managing, and iterating on your APIs to be sustainable in this new era. ### Live monitored sources - [Agentic Benchmarks 2026: Tool Use, Browsing, Computer Use | BenchLM.ai](https://benchlm.ai/agentic) — benchlm.ai (2026-05-12): BenchLM.ai updated its Agentic Benchmarks leaderboard on 2026-05-11. The update introduced two new benchmarks: 1) OpenHands Index, a holistic coding-agent benchmark covering issue resolution, frontend work, greenfield development, testing, and information gathering; and 2) SWE-At - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [AI Agent Identity and MCP: Authenticating Non-Human Identities](https://guptadeepak.com/ciam-compass/guides/ai-agent-identity-mcp) — guptadeepak.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [GitHub - agentgateway/agentgateway: Next Generation Agentic Proxy for AI Agents and MCP servers · GitHub](https://github.com/agentgateway/agentgateway) — github.com (2026-05-07): Agentgateway released version v1.2.0-alpha.1, continuing the development of its open-source AI-native proxy for agent-to-agent and agent-to-tool communication. The project maintains approximately 2.6k GitHub stars and is part of the Linux Foundation. - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent --- ## PRT207-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT207-S Level: intermediate Type: Breakout session Category: Partner Showcase Topics: Voice & Conversational AI; Resilience & Disaster Recovery; Retrieval Augmented Generation (RAG) Geopolitical instability, rising CX demands, rapid tech shifts, and escalating cyber threats converge faster than manual processes can handle. Join our expert panel as they leverage AWS and AI to build customer solutions, elevate engagement, and neutralise cyber threats. We'll share real deployments, proven governance, and measurable gains in efficiency, resilience, and customer impact. ### Playbook (editorial commentary) **The concept.** A cohesive AI-enabled customer engagement platform spanning engagement, governance, and security. Built for a world of geopolitical instability and rising CX expectations. **Why it matters.** CX disruption from AI is fast. Standalone AI tools fragment the experience — chat with bot, then transfer to human, then back to bot. Users hate it. **The hard parts.** Governance + speed are usually at odds. AI makes the gap worse — model changes that make UX better may compromise governance. **Playbook moves.** (1) Pre-approve AI patterns for common interactions; deploy fast within those patterns. (2) Build a "slow lane" for novel interactions that need governance review. (3) Instrument handoffs (bot → human, human → bot) as a primary metric. **The surprise.** The CX value of AI is rarely full automation — it's making humans *look smarter* to customers. Warm context, faster recall of customer history, faster handoffs. The "AI-augmented agent" outperforms both pure-AI and pure-human in most studies. Optimise for human-AI teaming, not displacement. --- ### Live monitored sources - [Scaling Autonomous Agent Swarms with Distributed Task ...](https://martinuke0.github.io/posts/2026-03-31-scaling-autonomous-agent-swarms-with-distributed-task-orchestration-and-low-latency-communication-protocols/) — martinuke0.github.io (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [The AI Agent Infrastructure Landscape in 2026: A Practitioner ...](https://chenagent.dev/articles/ai-agent-infrastructure-landscape-2026) — chenagent.dev (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. - [Circuit Breakers for AI Agents: How We Stop Cascading ...](https://cencori.com/blog/circuit-breakers-for-ai-agents) — cencori.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [definity definitely defines agentic data engineering platform](https://www.computerweekly.com/blog/CW-Developer-Network/definity-definitely-defines-agentic-data-engineering-platform) — computerweekly.com (2026-05-12): Waxell detailed the necessity of 'AgentOps' as a distinct discipline from MLOps, emphasizing the need for a runtime governance layer that intercepts and enforces policy on tool calls, data access, and cost boundaries (budget hard stops) before they are executed by the agent to pr - [Multi-Agent Orchestration Patterns Drive Enterprise ROI in 2026](https://insights.reinventing.ai/articles/ai-agents-orchestration-patterns-2026-03-18) — insights.reinventing.ai (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. --- ## PRT209-S — How Auto & General leverage observability foundations for AI URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT209-S Level: intermediate Type: Breakout session Category: Partner Showcase Topics: Observability & Monitoring; Retrieval Augmented Generation (RAG) As one of Australia's leading general insurers, Auto & General knew AI would play an important part in their future IT operations. To ensure success, the team embarked on a comprehensive observability maturity journey to build solid foundations, governance, and structure. Learn how A&G worked with New Relic to successfully lay observability foundations for the AI age. ### Playbook (editorial commentary) **The concept.** Observability maturity is a *precondition* for AI ops, not a follow-up. Foundations, governance, structure. **Why it matters.** You cannot run AI on systems you cannot see. AI ops compound the visibility problem because AI behaviours are stochastic and hard to reproduce. **The hard parts.** Maturity isn't tools — it's discipline. Consistent telemetry, consistent naming, clear ownership. Tools without discipline produce dashboards no one trusts. **Playbook moves.** (1) Score every service on observability maturity (logs / metrics / traces / ownership clarity). (2) Block AI rollouts on services that haven't been scored. (3) Make ownership explicit before tooling investment. **The surprise.** The fastest observability uplift comes from forcing *service ownership clarity*. Tools don't help if no one owns the service well enough to know what "normal" looks like. The org-design move (clear owners) precedes the tooling move. --- ### Live monitored sources - [Daxn — AI Agent Governance](http://daxn.ai/) — daxn.ai (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [AI Agent Audit Trail + RBAC: 2026 Enterprise Requirements](https://www.teamazing.com/blog/ai-agent-audit-trail-rbac-requirements/) — teamazing.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [Artificial Intelligence May 2026 - arXiv.org](https://arxiv.org/list/cs.AI/current) — arxiv.org (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [The Context Graph Revolution: Why Enterprise AI ... - Medium](https://medium.com/@thanapong_18619/the-context-graph-revolution-why-enterprise-ai-needs-decision-lineage-c01d90fd1db4) — medium.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. --- ## PRT213-S — How NAB is Conquering Multi-Cloud to Secure the Enterprise URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT213-S Level: intermediate Type: Breakout session Category: Partner Showcase Topics: Security, Identity & Compliance; Retrieval Augmented Generation (RAG) For the National Australia Bank (NAB), operating across multiple cloud environments and delivery platforms is a necessity but results in a critical challenge - fragmented visibility and the costly reality of "design once, build thrice" security controls. Learn how NAB moved beyond compliance to empower every engineering team with a unified view of risk, achieving security at speed and scale. ### Playbook (editorial commentary) **The concept.** Multi-cloud security via unified posture management. Stop "design once, build thrice" by abstracting policy from cloud-specific primitives. **Why it matters.** Banks live multi-cloud whether they want to or not — M&A, regional regulations, vendor diversification, redundancy. Single-cloud policies don't scale. **The hard parts.** Each cloud's primitives differ. Abstractions leak. A policy that works for IAM on AWS doesn't translate cleanly to Azure RBAC. **Playbook moves.** (1) Codify policies in a cloud-agnostic layer — a CSPM tool or policy-as-code framework. (2) Push enforcement into dev pipelines (block at PR), not at runtime. (3) Accept that some policies will need cloud-specific implementations; document why. **The surprise.** Multi-cloud security is mostly a *metadata* problem. What assets exist, where, owned by whom, classified how? Solve metadata first; controls fall out of metadata. Most orgs put the controls before the metadata and end up with controls they can't enforce. --- ### Live monitored sources - [proofpoint.com](https://www.proofpoint.com/us/products/ai-mcp-security) — proofpoint.com (2026-05-01): Proofpoint provides an MCP Security Platform to secure AI connectivity at scale by routing MCP traffic through a central gateway. The platform enables centralized discovery and risk classification of 'shadow' MCP servers, enforces authentication via OAuth 2.0, controls user and a - [Prompt Injection in Production Agents: 2026 Taxonomy](https://www.digitalapplied.com/blog/prompt-injection-production-agents-2026-taxonomy) — digitalapplied.com (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [Proofpoint Unveils Industry's Newest Intent-Based AI ...](http://proofpoint.com/us/newsroom/press-releases/proofpoint-unveils-industrys-newest-intent-based-ai-security-solution) — proofpoint.com (2026-05-10): Research Publication: The Cloud Security Alliance (CSA) released a research note on May 8, 2026, formalizing 'Promptware'—a class of prompt injection attacks that function as malware to create Agentic Command and Control (C2) infrastructure. Risk: The research highlights that L --- ## PRT216-S — Postman and the Future of AI-Driven API Development in 2026 URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT216-S Level: intermediate Type: Breakout session Category: Partner Showcase Topics: Agentic AI; Model Context Protocol (MCP); Kiro & Spec-Driven Development; Retrieval Augmented Generation (RAG) Postman and the Future of AI-Driven API Development in 2026 (sponsored by Postman, Inc)Software development has fundamentally changed in 2026, driven by vibe coding, AI agents, and RAG/MCP. APIs are the interface layer for AI systems to perform meaningful work. For this to succeed, your APIs must be discoverable, consistent, and usable by both developers and agents. Postman is now central to designing, managing, and iterating on your APIs to be sustainable in this new era. ### Live monitored sources - [Agentic Benchmarks 2026: Tool Use, Browsing, Computer Use | BenchLM.ai](https://benchlm.ai/agentic) — benchlm.ai (2026-05-12): BenchLM.ai updated its Agentic Benchmarks leaderboard on 2026-05-11. The update introduced two new benchmarks: 1) OpenHands Index, a holistic coding-agent benchmark covering issue resolution, frontend work, greenfield development, testing, and information gathering; and 2) SWE-At - [AI Agent Identity and MCP: Authenticating Non-Human Identities](https://guptadeepak.com/ciam-compass/guides/ai-agent-identity-mcp) — guptadeepak.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [GitHub - agentgateway/agentgateway: Next Generation Agentic Proxy for AI Agents and MCP servers · GitHub](https://github.com/agentgateway/agentgateway) — github.com (2026-05-07): Agentgateway released version v1.2.0-alpha.1, continuing the development of its open-source AI-native proxy for agent-to-agent and agent-to-tool communication. The project maintains approximately 2.6k GitHub stars and is part of the Linux Foundation. - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent --- ## PRT301-S — Unite Teams, Tools, and AI to Drive Transformation at Scale URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT301-S Level: advanced Type: Breakout session Category: Partner Showcase Topics: Mobile & Cross-Platform Development; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) Today's leaders face pressure to prove AI ROI, yet many struggle to realise value on fragmented foundations. Miro on AWS solves this by uniting teams, tools, and collaborative AI workflows in one workspace — amplifying the way teams actually work with enterprise scale and security. In this fireside chat with Culture Amp, hear how turning AI ambition into real outcomes works in practice. ### Playbook (editorial commentary) **The concept.** Collaborative AI workflows in a single workspace, uniting teams, tools, and AI across the work surface. Fireside chat with Culture Amp on real outcomes. **Why it matters.** AI tools fragmenting workflows is the new sprawl. Every team has its own AI tool; nothing connects. **The hard parts.** Whiteboard tools are sticky but rarely the system of record. Their value depends on whether teams actually keep them current. **Playbook moves.** (1) Pick what's source-of-truth (whiteboard? doc? Jira? Notion?) and stick to it. AI flows from there. (2) Audit tool sprawl quarterly — kill duplicates. (3) Resist adopting more AI tools until you've integrated the ones you have. **The surprise.** The actual productivity gain in collaborative AI tools is in what they *replace*, not what they add. If a new AI tool doesn't have a clear "this kills X meeting" or "this kills Y handoff" answer, it's just adding cognitive load. Audit before adopting. --- ### Live monitored sources - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. --- ## PRT103-S — Cloud Anywhere: Architectural Freedom for Unified Data and AI 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT103-S Level: foundational Type: Lightning talk Category: Partner Showcase Topics: Voice & Conversational AI; Security, Identity & Compliance; Generative AI & Foundation Models Need flexibility across cloud and on-premises This session reveals how delivering cloud anywhere gives you the architectural freedom to run data and AI across AWS, clouds and hybrid environments. Discover how a unified data fabric simplifies governance and security, accelerating AI-driven innovation. Achieve unified control and compliance for highly regulated environments. ### Playbook (editorial commentary) **The concept.** Hybrid + multi-cloud data fabric. Run AI workloads where the data lives. Unified governance across environments. **Why it matters.** Some data can't move — sovereignty laws, data gravity (TBs that cost too much to egress), regulatory constraints. The "consolidate to one cloud" dream isn't reality for most enterprises. **The hard parts.** Unified governance across environments is genuinely hard. Identity federation, policy translation, audit log aggregation — none of it is solved by one tool. **Playbook moves.** (1) Classify data by mobility — what can move, what can't, what shouldn't. (2) Architect for stationary data first; design pipelines that bring compute to data, not the other way. (3) Plan egress costs into the architecture explicitly. **The surprise.** "Cloud anywhere" sounds great in theory but adds substantial complexity. If 90% of your data already lives in one cloud, multi-cloud data fabric is overkill — you're paying complexity tax for the 10%. Decide whether you're truly multi-cloud before architecting like you are. --- ### Live monitored sources - [Releases · microsoft/agent-governance-toolkit · GitHub](https://github.com/microsoft/agent-governance-toolkit/releases) — github.com (2026-05-08): Microsoft released v3.5.0 of the Agent Governance Toolkit, adding enterprise-grade agent identity via Citadel Integration (Entra identity bridge), Multi-Agent Collective Policies for workflow-wide constraints, Intent-Based Authorization for structured lifecycle management (declar - [recordonline.com](https://www.recordonline.com/press-release/story/45494/langguard-ai-unveils-an-open-ai-control-plane-to-accelerate-enterprise-agentic-roi) — recordonline.com (2026-04-22): LangGuard.AI launched an 'Open AI Control Plane' to govern agentic workflows. Key features include Governance AI Run-time Links (GRAIL) for visibility into agent behavior and tool actions, and an Open Policy-as-Code framework for intent-driven enforcement across security and audi - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [aws.amazon.com](https://aws.amazon.com/about-aws/whats-new/2026/04/bedrock-openai-models-codex-managed-agents/) — aws.amazon.com (2026-04-29): Amazon Bedrock (AWS) now offers OpenAI models, Codex, and Managed Agents (Limited Preview) — announced 2026-04-28. What changed: OpenAI models and Managed Agents are available inside AWS Bedrock limited preview, letting AWS customers run OpenAI models and managed agent capabiliti - [cloudsecurityalliance.org](https://cloudsecurityalliance.org/blog/2026/04/29/securing-the-agentic-control-plane-key-progress-at-the-csai-foundation) — cloudsecurityalliance.org (2026-04-30): The CSAI Foundation (Cloud Security Alliance) announced its 2026 mission to secure the 'Agentic Control Plane' on 2026-04-29. Key initiatives include the Autonomous Action Runtime Management (AARM) framework (aarm.dev), an open specification for securing AI-driven actions at runt --- ## PRT112-S — Empower Data with Oracle AI Database and Native AI Services on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT112-S Level: foundational Category: Partner Showcase Topics: OpenSearch & Vector Search; Databases & Aurora Organisations today depend on fast, secure access to data to support mission-critical operations and evolving AI workloads. With Oracle AI Database services available on AWS, your team can streamline data integration and deliver high-impact solutions for semantic search, fraud detection, quality control, and much more. ### Playbook (editorial commentary) **The concept.** Oracle AI Database services running on AWS. Vector + relational together. Use cases: semantic search, fraud detection, quality control. **Why it matters.** Mature vector + RDBMS reduces stack complexity. One database instead of two reduces ops burden, sync issues, consistency problems. **The hard parts.** Performance trade-offs of vector operations in row-oriented RDBMS. Generic benchmarks don't tell you what your specific workload will do. **Playbook moves.** (1) Benchmark vector + relational hybrid queries against your real workload patterns. (2) Compare against the alternative (separate vector DB + RDBMS) honestly. (3) Factor in operational overhead, not just query performance. **The surprise.** Putting vectors next to transactional data unlocks real-time RAG over fresh data. Your warehouse can't do that — there's always replication lag. If your AI use case requires acting on fresh transactional data (fraud detection, real-time personalisation), the consolidated DB option becomes more compelling than its raw benchmarks suggest. --- ### Live monitored sources - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [kr.teradata.com](https://kr.teradata.com/press-releases/2026/teradata-enables-ai-agents) — kr.teradata.com (2026-04-19): Teradata announced agentic, multi-modal capabilities for Teradata Enterprise Vector Store: unified structured+unstructured data, automated ingestion (docs/images/audio), multi-modal embeddings (up to 8K dims), hybrid/ fusion search (semantic + lexical + metadata), LangChain integ - [businesswire.com](https://www.businesswire.com/news/home/20260422902027/en/Bedrock-Datas-ArgusAI-Now-Governs-AI-Agents-Built-on-Google-Vertex-AI-and-Dialogflow) — businesswire.com (2026-04-22): Bedrock Data announced that ArgusAI now provides governance and agent-aware access control for AI agents built on Google Vertex AI Search and Dialogflow. The platform implements a 'Data Bill of Materials (DBOM)' to automatically discover and map the data stores accessed by agents --- ## PRT217-S — Your Agents Should Be Durable URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT217-S Level: intermediate Type: Lightning talk Category: Partner Showcase Topics: Agentic AI; Containers: EKS, ECS & Fargate Your Agents Should Be Durable (sponsored by Temporal)Building AI agents is easy — making them production-ready is hard. Crashes, API failures, and state management are just a few challenges when moving from PoC to production. Learn how durable execution with Temporal makes it simple to build reliable agents that run for days, weeks, or months, using a code-first approach developers love. ### Playbook (editorial commentary) **The concept.** Durable execution for agents — state persistence, retries, long-running operations measured in days or weeks. Code-first. **Why it matters.** Naive agent code is fragile. Crashes lose state. Retries become infinite loops. Long-running workflows fail in nasty, partial ways. **The hard parts.** Distributed state is genuinely hard. Agents amplify the hardness because they have non-deterministic decision points throughout the workflow. **Playbook moves.** (1) Use durable execution for any agent that runs >5 min OR has external side effects. Don't roll your own. (2) Design workflows assuming any step might fail and retry; test that. (3) Make "human review" an explicit, durable workflow step, not an ad-hoc pause. **The surprise.** The most useful Temporal pattern for agents isn't retry — it's the ability to *pause* an agent for human review and *resume* hours or days later without state loss. This is the missing primitive in most agent frameworks. Once you have it, "human in the loop" becomes practical at scale; without it, HITL is a constant battle against state expiration. --- ### Live monitored sources - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Decision Traces: Essential AI Infrastructure for Enterprise Scale](https://atlan.com/know/what-are-decision-traces-for-ai-agents/) — atlan.com (2026-05-06): Core idea: The Context Graph resource guide defines context graphs as a 'living record of decision traces' used for execution validation and temporal reasoning. It distinguishes them from knowledge graphs (static entities) and vector databases (semantic similarity) by their nativ - [How to Build an Agentic AI Strategy With Process Intelligence](http://skan.ai/blogs/process-intelligence-for-agentic-ai-enterprise-automation) — skan.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## AIM204 — Get to know Amazon Quick, your new agentic teammate URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM204 Level: intermediate Type: Breakout session Category: AI & Machine Learning Topics: Security, Identity & Compliance; Databases & Aurora; Mobile & Cross-Platform Development; Analytics, Redshift & Generative BI; Agentic AI; Amazon Q & AI Assistants Most of us spend more time hunting for information than using it. Amazon Quick changes that. It reaches across all your company's data — documents, databases, emails, Slack threads, dashboards, Jira tickets — and lets you search it, ask questions, and take action, all from one place. Available across web, mobile, Slack, and Microsoft tools with multi-model intelligence, Quick delivers consumer-grade AI with enterprise-grade security and governance. No vendor lock-in, no siloed copilots. Just one AI teammate that works wherever you do. ### Playbook (editorial commentary) **The concept.** Cross-corpus enterprise search agent — documents, databases, emails, Slack threads, dashboards, Jira. Web, mobile, Slack, Microsoft tools. Multi-model intelligence. **Why it matters.** Enterprise search has been broken for 20 years. Most knowledge work begins with hunting for information. Agentic search reframes the problem. **The hard parts.** Permissions inheritance across systems. The agent must respect every source system's ACLs without papering over inconsistencies. A document leaked via search is the same as a document leaked any other way — but harder to detect. **Playbook moves.** (1) Audit access boundaries before connecting any source. (2) Beware accidental info leakage — the agent will surface things people forgot they had access to. (3) Plan for cross-system permission cleanup as part of rollout, not after. **The surprise.** Cross-corpus search exposes years of *permission-leak debt* you never knew existed. People had access to channels they should have been removed from; service accounts have read access to everything; old projects retain over-broad permissions. The agent will surface all of it. Plan for cleanup as a *deliberate* phase of rollout — pretending it's not there means it'll show up as a security incident. --- ### Live monitored sources - [Enterprise AI Agents Move From Pilot to Production: What 2026 ...](https://insights.reinventing.ai/articles/ai-agents-enterprise-production-2026-02-25) — insights.reinventing.ai (2026-05-05): Microsoft's 2026 Data Security Index reports that more than 80% of Fortune 500 companies are now running active AI agents in production, integrated across sales, finance, customer service, and security workflows. - [Releases · microsoft/agent-governance-toolkit · GitHub](https://github.com/microsoft/agent-governance-toolkit/releases) — github.com (2026-05-08): Microsoft released v3.5.0 of the Agent Governance Toolkit, adding enterprise-grade agent identity via Citadel Integration (Entra identity bridge), Multi-Agent Collective Policies for workflow-wide constraints, Intent-Based Authorization for structured lifecycle management (declar - [Artificial Intelligence May 2026 - arXiv.org](https://arxiv.org/list/cs.AI/current) — arxiv.org (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [The Context Graph Revolution: Why Enterprise AI ... - Medium](https://medium.com/@thanapong_18619/the-context-graph-revolution-why-enterprise-ai-needs-decision-lineage-c01d90fd1db4) — medium.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi --- ## DEV202 — AI Native Development: Strategies and Impact across Amazon and AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV202 Level: intermediate Type: Breakout session Category: Developer Tools Topics: Model Context Protocol (MCP); Kiro & Spec-Driven Development; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) AI Native Development: Strategies and Impact across Amazon and AWSAmazon and AWS have evolved beyond AI-assisted development to embrace AI Native practices, integrating AI as a partner throughout the software development lifecycle. Learn how their teams leverage AWS foundational tools including Kiro, and Amazon Bedrock. Discover effective Prompt Driven Development methodologies and grassroots adoption strategies from early champions. See how Amazon enables teams to provide AI with right context through strategic use of MCP, RAG and custom models trained on Amazon technical knowledge. Understand the culture transformation required across multi-thousand person organizations, where every role must evolve. Gain actionable insights to accelerate your AI Native journey. ### Live monitored sources - [GitHub Copilot in Visual Studio Code, April releases](http://github.blog/changelog/2026-05-06-github-copilot-in-visual-studio-code-april-releases) — github.blog (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [2026](https://docs.devin.ai/release-notes/2026) — docs.devin.ai (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Belitsoft Releases AI Agent Development Forecast 2026: 40% of ...](https://www.abnewswire.com/pressreleases/belitsoft-releases-ai-agent-development-forecast-2026-40-of-enterprise-applications-to-include-taskspecific-agents-by-year-end_800878.html) — abnewswire.com (2026-05-05): DeepClaude, a new open-source tool, has been released enabling the use of the Claude Code agent loop with DeepSeek V4 Pro, allowing Claude to orchestrate DeepSeek models for multi-step tasks. - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent --- ## DEV211 — Evolution of Automation: Orchestration to Intent-Based Supervision URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV211 Level: intermediate Type: Breakout session Category: Developer Tools Topics: Agentic AI Automation has long relied on orchestration: predefined workflows, rigid control flows, and human-managed exceptions. That model breaks down as systems become adaptive, distributed, and autonomous. In this session, we explore the shift from orchestration to intent-based supervision, where humans define purpose, constraints, and authority, while agents decide how to act within those bounds. Drawing on real-world agentic architectures, the talk shows how dynamic discovery, semantic negotiation, and closed-loop feedback replace static workflows. Attendees will learn how to design automation that scales safely, remains governable, and adapts as context changes, without removing human judgment from the system. ### Live monitored sources - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Enterprise AI Agents 2026: Mid-Year Report on What's Working](https://www.ampcome.com/post/enterprise-ai-agents-2026-mid-year-report) — ampcome.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM --- ## DEV301 — Evolution of Automation: Orchestration to Intent-Based Supervision URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV301 Level: advanced Type: Breakout session Category: Developer Tools Topics: Agentic AI Evolution of Automation: Orchestration to Intent-Based SupervisionAutomation has long relied on orchestration: predefined workflows, rigid control flows, and human-managed exceptions. That model breaks down as systems become adaptive, distributed, and autonomous. In this session, we explore the shift from orchestration to intent-based supervision, where humans define purpose, constraints, and authority, while agents decide how to act within those bounds. Drawing on real-world agentic architectures, the talk shows how dynamic discovery, semantic negotiation, and closed-loop feedback replace static workflows. Attendees will learn how to design automation that scales safely, remains governable, and adapts as context changes, without removing human judgment from the system. ### Live monitored sources - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Enterprise AI Agents 2026: Mid-Year Report on What's Working](https://www.ampcome.com/post/enterprise-ai-agents-2026-mid-year-report) — ampcome.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM --- ## DEV304 — Building Agentic AI: Amazon Nova Act and Strands Agents in Practice URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV304 Level: advanced Type: Breakout session Category: Developer Tools Topics: Agentic AI; Media & Entertainment Explore practical applications of Agentic AI through two real-world case studies. First, dive into a Hong Kong weather agent built with Amazon Nova Act, featuring conversation flow design and meteorological data integration. Watch a live demo showcasing natural language interaction for weather information retrieval and trend analysis. Next, discover a mathematics teaching AI agent developed using the Strands Agents framework, demonstrating personalized math video creation through automatic content generation. Both cases include comprehensive code demonstrations, providing developers with concrete references for AI agent application development on AWS. ### Playbook (editorial commentary) **The concept.** Practical agent patterns with Amazon Nova Act + Strands SDK. Two case studies referenced: a Hong Kong weather agent (conversation flow design + meteorological data integration) and a mathematics teaching agent (description cuts off). **Why it matters.** SDK choice meaningfully shapes agent design. Strands is the AWS-blessed path; using it gets you the integrations and the support relationships, at the cost of less framework flexibility than alternatives. **The hard parts.** Conversation flow design is where most demos fail in production. Happy paths look great; recovery from misunderstandings, ambiguous queries, and context loss is where the work is. **Playbook moves.** (1) Look at conversation flow design first — that's where most demos fail. (2) Pay attention to how the demo handles edge cases (ambiguous queries, missing data, follow-up questions). (3) Map the SDK's strengths to your use case before adopting. **The surprise.** The Strands SDK pushes you toward stateful conversation patterns by default. That's the right default for most assistant-style agents but the wrong default for stateless task-running agents (e.g., classify this document, extract these fields). Know which kind you're building before you start; the architectural commitment is real. > *Note: this session's source description was truncated mid-sentence. The above is reconstructed from the partial text plus context.* --- ### Live monitored sources - [GitHub - agentgateway/agentgateway: Next Generation Agentic Proxy for AI Agents and MCP servers · GitHub](https://github.com/agentgateway/agentgateway) — github.com (2026-05-07): Agentgateway released version v1.2.0-alpha.1, continuing the development of its open-source AI-native proxy for agent-to-agent and agent-to-tool communication. The project maintains approximately 2.6k GitHub stars and is part of the Linux Foundation. - [Edge Delta Makes All Telemetry Pipelines Data ...](http://prnewswire.com/news-releases/edge-delta-makes-all-telemetry-pipelines-data-throughput-limitless-and-free-for-all-customers-302736808.html) — prnewswire.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca --- ## DEV314 — AI Native Development: Strategies and Impact across Amazon and AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV314 Level: advanced Type: Breakout session Category: Developer Tools Topics: Model Context Protocol (MCP); Kiro & Spec-Driven Development; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) Amazon and AWS have evolved beyond AI-assisted development to embrace AI Native practices, integrating AI as a partner throughout the software development lifecycle. Learn how their teams leverage AWS foundational tools including Kiro, and Amazon Bedrock. Discover effective Prompt Driven Development methodologies and grassroots adoption strategies from early champions. See how Amazon enables teams to provide AI with right context through strategic use of MCP, RAG and custom models trained on Amazon technical knowledge. Understand the culture transformation required across multi-thousand person organizations, where every role must evolve. Gain actionable insights to accelerate your AI Native journey. ### Live monitored sources - [GitHub Copilot in Visual Studio Code, April releases](http://github.blog/changelog/2026-05-06-github-copilot-in-visual-studio-code-april-releases) — github.blog (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [2026](https://docs.devin.ai/release-notes/2026) — docs.devin.ai (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Belitsoft Releases AI Agent Development Forecast 2026: 40% of ...](https://www.abnewswire.com/pressreleases/belitsoft-releases-ai-agent-development-forecast-2026-40-of-enterprise-applications-to-include-taskspecific-agents-by-year-end_800878.html) — abnewswire.com (2026-05-05): DeepClaude, a new open-source tool, has been released enabling the use of the Claude Code agent loop with DeepSeek V4 Pro, allowing Claude to orchestrate DeepSeek models for multi-step tasks. - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent --- ## MAM307 — Modernise legacy code using fine-tuned Gen AI models URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM307 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Agentic AI; Machine Learning & SageMaker; Migration & Modernization; Generative AI & Foundation Models Rio Tintos data science team saw an opportunity to preserve institutional knowledge and improve developer productivity by modernizing a legacy codebase. Rather than attempting a full system overhaul, the team focused first on adding generative AI capabilities to their critical legacy application. By using the proven, open, and trusted data foundation of AWS, the company laid the groundwork for incremental modernization without disrupting core operations. Learn about model fine tuning against legacy codebases, Amazon Nova, SageMaker Jumpstart and AgentCore in this deep dive with AWS & Rio Tinto ### Live monitored sources - [Decision Traces: Essential AI Infrastructure for Enterprise Scale](https://atlan.com/know/what-are-decision-traces-for-ai-agents/) — atlan.com (2026-05-06): Core idea: The Context Graph resource guide defines context graphs as a 'living record of decision traces' used for execution validation and temporal reasoning. It distinguishes them from knowledge graphs (static entities) and vector databases (semantic similarity) by their nativ - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [AI Agent Audit Trail: Complete Guide for 2026 | Fastio](https://fast.io/resources/ai-agent-audit-trail/) — fast.io (2026-05-04): Core idea: Atlan defines 'Decision Traces' as structured records of how and why organizational decisions were made (capturing reasoning paths, policies, and precedents) rather than just data movement. It positions decision traces as a key layer of a 'Context Graph' (alongside kno - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [agent-audit-trail · PyPI](https://pypi.org/project/agent-audit-trail/) — pypi.org (2026-05-04): Core idea: Atlan defines 'Decision Traces' as structured records of how and why organizational decisions were made (capturing reasoning paths, policies, and precedents) rather than just data movement. It positions decision traces as a key layer of a 'Context Graph' (alongside kno --- ## PRT101-S — Accelerating Innovation with GitLab DAP Powered by Amazon Bedrock URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT101-S Level: foundational Type: Lightning talk Category: Partner Showcase Topics: Agentic AI; Observability & Monitoring; Security, Identity & Compliance; Generative AI & Foundation Models Learn how GitLab Duo Agent Platform (DAP) powered by Amazon Bedrock brings agentic AI into every stage of the software development lifecycle while keeping data, logs, and inference traffic inside your AWS environment. We'll show how teams can orchestrate AI-assisted workflows for planning, coding, security, and compliance using Amazon Bedrock foundation models behind GitLab's AI Gateway. ### Live monitored sources - [Context Graph: The Definitive Resource for AI Decision Traces](https://www.contextgraph.tech/) — contextgraph.tech (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Belitsoft Releases AI Agent Development Forecast 2026: 40% of ...](https://www.abnewswire.com/pressreleases/belitsoft-releases-ai-agent-development-forecast-2026-40-of-enterprise-applications-to-include-taskspecific-agents-by-year-end_800878.html) — abnewswire.com (2026-05-05): DeepClaude, a new open-source tool, has been released enabling the use of the Claude Code agent loop with DeepSeek V4 Pro, allowing Claude to orchestrate DeepSeek models for multi-step tasks. --- ## PRT203-S — Design, Deploy, and Govern AI Agents with Boomis Agentstudio 5 Steps to Enterprise-Grade AI Security for Amazon Bedrock Projects URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT203-S Level: intermediate Type: Lightning talk Category: Partner Showcase Topics: Agentic AI; Observability & Monitoring; Security, Identity & Compliance; Generative AI & Foundation Models Transform enterprise automation with Boomi's AI agent ecosystem. Learn to use Agent Designer to visually build agents that connect across systems, and Agent Control Tower for centralised governance, compliance, and performance monitoring. Securely orchestrate your AI lifecycle at scale with Amazon Bedrock. ### Live monitored sources - [Guild Raises $44M to Build the Agent Control Plane](https://www.guild.ai/knowledge/guild-raises-44m-agent-control-plane) — guild.ai (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [ServiceNow expands AI agent governance through deeper ...](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-expands-AI-agent-governance-through-deeper-integration-with-Microsoft/default.aspx) — newsroom.servicenow.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## PRT205-S — The AI Challenge You Don't Yet Know About - Software Supply Chain URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT205-S Level: intermediate Type: Lightning talk Category: Partner Showcase In an era where AI accelerates software development, organisations must address supply chain risks. Using insights from Chainguard's research and observed industry patterns, we walk through the mechanics of these compromises without the hype. Attendees will leave with straightforward approaches for enhancing software integrity and building long-term trust in their development pipeline on AWS. --- ## SEC305 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads URL: https://aws-summit-2026-kb.pages.dev/sessions/SEC305 Level: advanced Type: Breakout session Category: Security, Identity & Compliance Topics: Agentic AI; Industry Spotlight: Public Sector & Government; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) Dive deep into advanced security architectures for AI workloads, exploring how to protect your workload against sophisticated attack vectors. Through technical examples, we'll implement secure architectures for AI workloads, covering identity, fine-grained access policies, and secure foundation model deployment patterns. Learn how to harden generative and agentic AI applications using AWS security capabilities, implementing least-privilege controls, and building secure architectures at scale. ### Live monitored sources - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-05): The CSAI Foundation (Cloud Security Alliance) announced milestones to secure the agentic control plane, including the strategic acquisition/stewardship of two foundational specifications: the Autonomous Action Runtime Management (AARM) specification (for securing AI-driven action - [AI Agent Delegation Patterns: Four Best Architectures for 2026 | Fastio](https://fast.io/resources/ai-agent-delegation-patterns) — fast.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Agentic Identity and Access Management](https://www.coalitionforsecureai.org/wp-content/uploads/2026/04/agentic-identity-and-access-control.pdf) — coalitionforsecureai.org (2026-05-08): New implementation patterns for AI agent identity (updated May 6, 2026) highlight the convergence of the Model Context Protocol (MCP) for agent-server handshakes and OAuth 2.1 with Dynamic Client Registration (DCR) for runtime credential issuance. A key pattern is the use of 'dis --- ## TNC202 — Accelerate Your Cloud Journey with AWS Transform URL: https://aws-summit-2026-kb.pages.dev/sessions/TNC202 Level: intermediate Type: Lightning talk Category: Other Topics: Agentic AI; Compute: EC2, Graviton & Nitro; Migration & Modernization; Retrieval Augmented Generation (RAG) Embark on a faster, smoother cloud transformation with agentic AI and integrated solutions. This session reveals how AWS Transform accelerates your cloud journey, addressing migration and modernization challenges through intelligent automation. Through real-world examples, discover how to leverage this powerful integration to fast-track your cloud adoption and transformation efforts. With the specialized AI agents of AWS Transform, customers can migrate VMware workloads to Amazon EC2, modernize .NET applications to cross-platform .NET, and modernize IBM z/OS mainframe applications, delivering transformation projects up to 4x faster. ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## DEV312 — Strands Agents on Lambda: Observability With Powertools & X-Ray URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV312 Level: advanced Category: Developer Tools Topics: Agentic AI; Observability & Monitoring; Serverless: Lambda & Step Functions When a Strands Agent fails across five Lambda log streams with no correlation, debugging takes 20 minutes minimum. This session demonstrates a structured observability layer that reduces diagnosis to under two minutes. You'll learn how Lambda Powertools Tracer wraps Strands tool invocations as X-Ray subsegments, how Powertools Logger injects AgentCore session correlation IDs across invocations, and how Powertools Metrics surfaces tool retry frequency as CloudWatch alarms — before timeouts occur. The session covers three production failure classes — tool timeout, reasoning loop, and retry storm — and delivers a reusable CDK construct providing full instrumentation for any Strands Agent Lambda deployment. ### Live monitored sources - [AI Agent Token Budget Enforcement [2026]](https://www.waxell.ai/blog/ai-agent-token-budget-enforcement) — axell.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [The Operator Vault Publishes Free OpenClaw API ...](http://usatoday.com/press-release/story/27628/the-operator-vault-publishes-free-openclaw-api-database-for-ai-agent-builders) — usatoday.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [43,750% Surge! BNB Chain is Crushing It with 150,000 AI Agents Blasting Through the Track | 小机构集团 on Binance Square](http://binance.com/en/square/post/316311861525314) — binance.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. --- ## ISV301 — Rolling to Scale: Roller's Multi-Tenant SaaS platform on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV301 Level: advanced Type: Lightning talk Category: ISV & Partners Topics: Security, Identity & Compliance; Startups & Innovation; Retrieval Augmented Generation (RAG) Learn how Roller Software grew from an Australian startup into a global venue management platform serving 3,000 venues across 30 countries and delivering 120 million experiences annually. Using AWS multi-tenant architecture, Roller maintains 99.99% uptime while processing $4 billion in transactions each year through their modern monolith application. This session covers practical strategies for tenant isolation, infrastructure scaling, and enterprise-grade security. Youll discover how to leverage AWSs native multi-tenant capabilities and get a proven roadmap for scaling your SaaS business from startup to enterprise while keeping costs efficient and operations excellent. ### Live monitored sources - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Tekst Raises $13.5 Million Series A | The SaaS News](http://thesaasnews.com/news/tekst-raises-13-5-million-series-a) — thesaasnews.com (2026-05-11): China's Ministry of Industry and Information Technology (MIIT) released a national standard for AI terminal intelligence grading, which includes the implementation of the smart cockpit level 3 standard. --- ## STP211 — Authenticating AI Agents: How Kinde Navigates Agentic Identity URL: https://aws-summit-2026-kb.pages.dev/sessions/STP211 Level: intermediate Category: Startups Topics: Agentic AI; Security, Identity & Compliance AI agents are no longer just answering questions - they're booking flights, managing infrastructure, and calling APIs on our behalf. But when an agent acts autonomously, who's really knocking on the door This talk explores how Kinde is rethinking authentication and authorisation for a world where your users aren't always human, covering machine-to-machine identity, delegated scopes, and why traditional auth flows break down when agents enter the chat. ### Live monitored sources - [Meow Technologies launches the first agentic banking ...](http://thenextweb.com/news/meow-technologies-agentic-banking-ai-agents) — thenextweb.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [AI Agent Identity and MCP: Authenticating Non-Human Identities](https://guptadeepak.com/ciam-compass/guides/ai-agent-identity-mcp) — guptadeepak.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [Experian Announces Agent Trust to Power Trusted AI ...](http://businesswire.com/news/home/20260430719198/en/Experian-Announces-Agent-Trust-to-Power-Trusted-AI-Driven-Commerce) — businesswire.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Payment rails built for AI agents - primary.vc](https://www.primary.vc/articles/payment-rails-built-for-ai-agents) — primary.vc (2026-05-06): Solana Foundation President Lily Liu announced at Consensus Miami 2026 that Solana is building the payment rails for the 'AI machine economy,' emphasizing that traditional credit card networks are structurally unable to support the micropayments necessary for autonomous AI agent --- ## AIM203 — Prompt Engineering to Learning Systems: Woodside's Agentic Ecosystem URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM203 Level: intermediate Type: Breakout session Category: AI & Machine Learning Topics: Agentic AI; Security, Identity & Compliance Woodsides Agentic Maintenance Framework connects frontline execution to longhorizon strategy, turning each job into fuel for continuous improvement. The approach uses governed evidence and multiagent AI to assemble the right context at the decision pointimproving request quality, planning accuracy, and execution readinesswhile capturing planvsactual signals that strengthen backlog quality, scheduling confidence, and longterm maintenance strategies. The result is a closed loop where execution improves strategy, and strategy improves execution, all within existing governance and systems of record. In this talk, well share practical lessons from designing the tactical layer (Maint Assist) and the strategic layer (Maint Intel), show how evidence is created once and reused across the lifecycle, and outline a maturity path from prompts to agentic orchestrationfocused on safety, reliability, and efficiency. ### Live monitored sources - [Agentic RAG Explained: AI Agents + RAG in 2026](https://freeacademy.ai/blog/agentic-rag-ai-agents-supercharge-retrieval-2026) — freeacademy.ai (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [The State of AI Agent Memory in 2026: What the Research ...](https://dev.to/vektor_memory_43f51a32376/the-state-of-ai-agent-memory-in-2026-what-the-research-actually-shows-3aja) — dev.to (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [AI Agent Memory Systems Cut Costs 60% with Long-Term Context 2026](https://iterathon.tech/blog/ai-agent-memory-systems-implementation-guide-2026) — iterathon.tech (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [AI Agent Context Window Cost: Why Bills Multiply [2026]](https://www.waxell.ai/blog/ai-agent-context-window-cost) — axell.ai (2026-05-05): Waxell published an analysis on the compounding cost of AI agent context windows, detailing how naive history management leads to 3x-5x budget underestimation. They proposed a runtime enforcement architecture (Waxell Runtime) that operates in the execution path to enforce hard to --- ## ARC304 — Demystifying Agent Identity URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC304 Level: advanced Type: Breakout session Category: Architecture Topics: Agentic AI; Security, Identity & Compliance; Manufacturing & Industry 4.0 Confused by inbound vs. outbound authentication for agents You're not alone. This Level 300 session demystifies OAuth flows and agent identity patterns through the lens of a practitioner's learning journey. Explore the differences between SPA (single-page web app) and agent authentication, then dive into AgentCore's inbound/outbound auth with Runtime and Gateway. Through live code demonstrations of 3-legged OAuth flows, you'll see exactly how agents authorize actions on behalf of users. Leave with working code examples from aws-samples and practical implementation insights to accelerate your agent development. Part of the AgentCore session track. ### Live monitored sources - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Token Security Introduces Intent-Based Security for AI Agents](http://finance.yahoo.com/news/token-security-introduces-intent-based-130200458.html) — finance.yahoo.com (2026-05-08): New implementation patterns for AI agent identity (updated May 6, 2026) highlight the convergence of the Model Context Protocol (MCP) for agent-server handshakes and OAuth 2.1 with Dynamic Client Registration (DCR) for runtime credential issuance. A key pattern is the use of 'dis - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [draft-klrc-aiagent-auth-01 - AI Agent Authentication and ...](https://datatracker.ietf.org/doc/draft-klrc-aiagent-auth/) — datatracker.ietf.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## COP302 — Applying AI for FinOps and FinOps for AI URL: https://aws-summit-2026-kb.pages.dev/sessions/COP302 Level: advanced Type: Breakout session Category: Other Topics: Cost Optimization & FinOps; Kiro & Spec-Driven Development; Generative AI & Foundation Models; Compute: EC2, Graviton & Nitro; Machine Learning & SageMaker Explore the intersection of AI and FinOps in this advanced session. First, discover how Kiro CLI can simplify AWS cost management by analyzing trends, explaining spend, and recommending optimizations like rightsizing and Savings Plans. Then, dive into FinOps for AI- learn how to track and control generative AI costs across Amazon EC2, Amazon SageMaker, Amazon Bedrock, and more. We'll share architecture patterns, cost-saving strategies, and real-world examples to help you build scalable, production-ready AI solutions while staying on budget. Whether you're optimizing existing workloads or launching new AI initiatives, you'll leave with practical tools to maximize value. ### Live monitored sources - [Morgan Stanley warns an AI breakthrough Is coming in 2026](https://finance.yahoo.com/news/morgan-stanley-warns-ai-breakthrough-072000084.html) — finance.yahoo.com (2026-05-09): Mother Ventures announced the close of its first $10 million early-stage venture capital fund (Fund I), targeting companies where mothers are the primary consumer. - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [Devin AI Guide 2026: Features, Pricing, How to Use & Complete ...](https://aitoolsdevpro.com/ai-tools/devin-guide/) — aitoolsdevpro.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [What's New in Cursor — Latest Updates & Release Notes](http://cursor.com/changelog) — cursor.com (2026-05-11): Devin updated its pricing structure to include several tiers: Free (limited usage, Devin Review, DeepWiki), Pro ($20/month with usage quotas and integrations), Max ($200/month with increased quotas), Teams ($80/month with unlimited members and centralized billing), and Enterprise - [GitHub - microsoft/autogen: A programming framework for ...](https://github.com/microsoft/autogen) — github.com (2026-05-07): CrewAI released pre-release version 1.14.5a3 on 2026-05-06. Key changes include: - Refactored the CLI into a standalone `crewai-cli` package. - Fixed the status endpoint path from `/{kickoff_id}/status` to `/status/{kickoff_id}`. - Updated the `gitpython` dependency to version >= --- ## DAT402 — Deep dive into database integrations with AWS Zero-ETL URL: https://aws-summit-2026-kb.pages.dev/sessions/DAT402 Level: expert Type: Breakout session Category: Databases Topics: Streaming & Real-Time Data; Observability & Monitoring; Voice & Conversational AI; Databases & Aurora; Analytics, Redshift & Generative BI; OpenSearch & Vector Search; Machine Learning & SageMaker Learn how AWS zero-ETL integrations eliminate complex data movement pipelines across multiple database engines, enabling data engineers, architects, and DBAs to reduce maintenance overhead while ensuring near real-time data availability for analytics and ML workloads. Examine the underlying architecture for supported zero-ETL integrations between Amazon Aurora, Amazon DynamoDB, and Amazon RDS sources to Amazon Redshift, Amazon SageMaker, and Amazon OpenSearch Service targets. Explore data movement options, tunable settings, and monitoring capabilities for ongoing data replicationall without traditional ETL complexity. ### Live monitored sources - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Gartner 2026 Confirms It: The Context Graph Is the Missing ...](https://thecontextgraph.co/memos/gartner-2026-ai-agents-decision-intelligence-sales) — thecontextgraph.co (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi --- ## DEV201 — How Flybuys Built AI Governance to Accelerate Adoption at Scale URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV201 Level: intermediate Type: Breakout session Category: Developer Tools Topics: Kiro & Spec-Driven Development; Security, Identity & Compliance; Databases & Aurora; Machine Learning & SageMaker; Retrieval Augmented Generation (RAG) Scaling AI successfully isnt just about moving fast — its about building the right foundations first. In this session, learn how Flybuys focused early on AI governance, steering documents, and engineering standards to enable smooth, secure AI adoption at scale. Well explore how upfront investment in guardrails, training, and approval processes allowed teams to deploy AI capabilities faster and with confidence. Youll hear how Flybuys is embedding governance and security expectations into engineering workflows using Kiro, including standardised steering patterns, approval pathways, and controlled rollout of AI capabilities such as Powers. Attendees will gain practical insights into how slowing down early can unlock faster, safer AI delivery across the organisation. ### Live monitored sources - [MCP Security Gateway - Agent Governance Toolkit](https://microsoft.github.io/agent-governance-toolkit/tutorials/07-mcp-security-gateway/) — microsoft.github.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [What’s New in Agent 365: May 2026 | Microsoft Community Hub](https://techcommunity.microsoft.com/blog/agent-365-blog/what%E2%80%99s-new-in-agent-365-may-2026/4516340) — techcommunity.microsoft.com (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [MCP Governance (2026): Policy Gates for MCP Servers](https://cordum.io/blog/mcp-governance-servers) — cordum.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding --- ## DEV305 — Agents in the enterprise: Best practices with Amazon Bedrock AgentCore URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV305 Level: advanced Type: Breakout session Category: Developer Tools Topics: Model Context Protocol (MCP); Observability & Monitoring; Security, Identity & Compliance; Generative AI & Foundation Models; Agentic AI As organizations scale AI agent development, robust enterprise architecture patterns become essential. In this advanced session, we'll explore how Amazon Bedrock AgentCore enables teams to build modular systems using their preferred frameworks while sharing tools through MCP gateways. Learn about A2A collaboration, shared memory, identity-based access controls, and integrated observability. Discover practical strategies for secure runtime deployment, standardized tool integration, evaluation frameworks, and end-to-end monitoring. Leave with actionable insights to build secure, scalable agent infrastructures that balance centralized governance with team autonomy. ### Live monitored sources - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Guild Raises $44M to Build the Agent Control Plane](https://www.guild.ai/knowledge/guild-raises-44m-agent-control-plane) — guild.ai (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [MCP Governance (2026): Policy Gates for MCP Servers](https://cordum.io/blog/mcp-governance-servers) — cordum.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding --- ## SEC301 — Inside the Attack Chain: Emerging Threat Actor Tactics and Techniques URL: https://aws-summit-2026-kb.pages.dev/sessions/SEC301 Level: advanced Type: Breakout session Category: Security, Identity & Compliance Topics: Industry Spotlight: Public Sector & Government Get up to speed on emerging threat actor tactics and techniques across the attack chain. In this session, we'll expound upon methods of Initial Access, Persistence, Defense Evasion, Lateral Movement, and Impact. You'll gain practical knowledge of specific configurations to detect and mitigate these threats, and learn common misconfigurations to avoid. --- ## DEV401 — Build Intelligent Memory Systems for AI Agents URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV401 Level: expert Category: Developer Tools Topics: Agentic AI; Generative AI & Foundation Models Explore how to build production-grade AI agents with persistent, context-aware memory using Amazon Bedrock AgentCore. This expert session covers architectural patterns for implementing four memory typesepisodic, semantic, preference, and summaryenabling agents that recall past interactions, recognize cross-session patterns, and maintain conversation context. You'll examine real implementation techniques for long-term memory management, workflow orchestration, and retrieval strategies using Amazon Bedrock. Leave with practical skills to design intelligent agents that deliver faster, more accurate responses in high-stakes applications. Ideal for architects and engineers ready to move beyond stateless AI toward genuinely intelligent, memory-enabled systems. ### Live monitored sources - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [AI Agent Memory Systems Cut Costs 60% with Long-Term Context 2026](https://iterathon.tech/blog/ai-agent-memory-systems-implementation-guide-2026) — iterathon.tech (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Agentic RAG Explained: AI Agents + RAG in 2026](https://freeacademy.ai/blog/agentic-rag-ai-agents-supercharge-retrieval-2026) — freeacademy.ai (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [The State of AI Agent Memory in 2026: What the Research ...](https://dev.to/vektor_memory_43f51a32376/the-state-of-ai-agent-memory-in-2026-what-the-research-actually-shows-3aja) — dev.to (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys --- ## ISV205 — AWS Graviton: The best price performance for your AWS workloads URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV205 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Security, Identity & Compliance; Migration & Modernization; Compute: EC2, Graviton & Nitro AWS Graviton-based Amazon EC2 instances provide the best price performance for workloads in Amazon EC2. In this session, dive deep into the AWS Graviton processor and learn about its workload performance, energy efficiency, and software offerings. Hear from Atlassian as they share their Graviton adoption journey and practical tips for migration success. Learn about common use cases, best practices to optimize your workloads across various applications, customer success stories and how you can accelerate your AWS Graviton journey. ### Live monitored sources - [CISA, US and International Partners Release Guide to Secure ...](https://www.cisa.gov/news-events/news/cisa-us-and-international-partners-release-guide-secure-adoption-agentic-ai) — cisa.gov (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Comment and Control: GitHub AI Agents as Credential ...](https://labs.cloudsecurityalliance.org/research/csa-research-note-comment-control-github-prompt-injection-20/) — labs.cloudsecurityalliance.org (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against --- ## STP208 — NextAI's LegalScout: A Data Foundation for Private Legal AI URL: https://aws-summit-2026-kb.pages.dev/sessions/STP208 Level: intermediate Category: Startups Topics: Storage: S3, EBS & EFS; Security, Identity & Compliance; Generative AI & Foundation Models; Manufacturing & Industry 4.0; Data Governance & Privacy; Retrieval Augmented Generation (RAG) LegalScout helps Australian SME law firms turn Generative AI into a competitive advantage by securely leveraging their own client data and confidential matters to work smarter, not harder. Built with Australian lawyers on AWS using Amazon Bedrock for inference and Amazon S3Vectors for secure document searches, it automates repetitive work, streamlines workflows, and improves drafting, contract review, and research to boost productivity, reduce costs, and lift accuracy while maintaining strict privacy and compliance. ### Live monitored sources - [The State of AI Agent Memory in 2026: What the Research ...](https://dev.to/vektor_memory_43f51a32376/the-state-of-ai-agent-memory-in-2026-what-the-research-actually-shows-3aja) — dev.to (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [title: State of AI Agent Memory 2026 description: "The state of AI agent memory in 2026: benchmark data across 10 approaches, 21 integrations, and the architectural shifts that matter." published: "Ma](https://mem0.ai/blog/state-of-ai-agent-memory-2026) — mem0.ai (2026-05-08): Mem0 published a comprehensive report, 'State of AI Agent Memory 2026,' benchmarking ten distinct memory approaches using the LOCOMO dataset. The study found that while full-context passing is most accurate, Graph-Enhanced Memory (Mem0g) provides a superior balance of accuracy (6 - [AI Agent Memory Systems Cut Costs 60% with Long-Term Context 2026](https://iterathon.tech/blog/ai-agent-memory-systems-implementation-guide-2026) — iterathon.tech (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [Agentic RAG Explained: AI Agents + RAG in 2026](https://freeacademy.ai/blog/agentic-rag-ai-agents-supercharge-retrieval-2026) — freeacademy.ai (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [ec.europa.eu](https://ec.europa.eu/transparency/documents-register/api/files/SWD(2026)123?ersIds=090166e52cccdfa5) — ec.europa.eu (2026-04-29): European Commission published staff working document SWD(2026)123 (28 April 2026) — a review/report that discusses digital markets, cloud services and mentions AI/agent considerations (interoperability, data portability, and the potential extension of certain obligations to AI se --- ## GHJ301 — R2 — AWS Game Day : Secret Agent Unicorns URL: https://aws-summit-2026-kb.pages.dev/sessions/GHJ301 Level: advanced Type: Gamified learning Category: Other Topics: Agentic AI; Gaming & Interactive Media AWS Game Day : Secret Agent UnicornsAWS GameDay is a gamified learning event that challenges participants to use AWS solutions to solve real-world technical problems in a risk-free setting. As a new hire at Unicorn.Rentals, the worlds largest mythical creature rental company, youll test your AWS knowledge in an interactive, team-based, risk-free environment! The Secret Agentic Unicorns GameDay covers the different components of AgentCore, guiding participants through agent creating using Strands, A2A, and other AgentCore technologies. Youll get real world experience creating and learning about AgentCore agents, and have fun along the way. ### Playbook (editorial commentary) **The concept.** Hands-on AgentCore lab — Strands SDK, agent-to-agent (A2A) protocols, agent building under simulated incident pressure. **Why it matters.** You don't learn agents from slides. You learn by debugging them at 2 AM when something is broken and the demo gods have fled. **The hard parts.** GameDays are time-boxed. Trade-offs and shortcuts mirror real life — that's the point. **Playbook moves.** (1) Send 2–3 engineers, not one. The team dynamics are part of the lesson. (2) Debrief afterwards as a written internal post-mortem. (3) Pick people who'll bring lessons back to their teams, not just enthusiasts. **The surprise.** GameDay's hidden value is meeting *other practitioners*. The networking IS the curriculum. Hallway conversations during breaks generate more "huh, we should try that" than the labs themselves. Optimise for that. --- ### Live monitored sources - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## TNC201 — Explore the Agentic Capabilities of Amazon Quick Suite URL: https://aws-summit-2026-kb.pages.dev/sessions/TNC201 Level: intermediate Type: Lightning talk Category: Other Topics: Agentic AI; Amazon Q & AI Assistants; Analytics, Redshift & Generative BI; Generative AI & Foundation Models Discover the latest features of Amazon Quick Suite, a generative AI-powered business intelligence platform transforming organizational data workflows. Explore the newest capabilities including Quick Sight for interactive visualizations, Quick Flows for workflow creation, Quick Automate for intelligent automation, and Quick Research for comprehensive analysis. Learn how custom chat agents, knowledge spaces, and workplace extensions integrate seamlessly to enhance productivity through natural language interactions across your organization. ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Edge Delta Makes All Telemetry Pipelines Data ...](http://prnewswire.com/news-releases/edge-delta-makes-all-telemetry-pipelines-data-throughput-limitless-and-free-for-all-customers-302736808.html) — prnewswire.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [How to Build an Agentic AI Strategy With Process Intelligence](http://skan.ai/blogs/process-intelligence-for-agentic-ai-enterprise-automation) — skan.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## DEV209 — CI/CD Guardrails for Agentic Coding Workflows URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV209 Level: intermediate Category: Developer Tools Topics: Agentic AI; DevOps, CI/CD & DevSecOps; Containers: EKS, ECS & Fargate; Code Generation & AI-Assisted Development AI coding agents introduce failure modes traditional CI/CD pipelines weren't built to catch — deleted tests, weakened type constraints, silent cross-service regressions. This session examines practical pipeline-level guardrails for agentic workflows running on ECS Fargate and distributed CI environments. You'll learn which failure patterns agents introduce that humans rarely do, which automated checks reliably catch them, and how to structure pipelines that apply appropriate scrutiny to agent-generated code without blocking developer velocity. Leave with concrete, implementable patterns covering test integrity enforcement, type safety validation, and cross-service regression detection — applicable whether you're managing one agent or coordinating many across multiple repositories. ### Live monitored sources - [Scaling Autonomous Agent Swarms with Distributed Task ...](https://martinuke0.github.io/posts/2026-03-31-scaling-autonomous-agent-swarms-with-distributed-task-orchestration-and-low-latency-communication-protocols/) — martinuke0.github.io (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv - [Announcing Agent Control: The Open Source Control Plane for ...](https://galileo.ai/blog/announcing-agent-control) — galileo.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [80% of Fortune 500 use active AI Agents: Observability ...](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/) — microsoft.com (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 --- ## ISV304 — Managing AI Agents with Confidence and Control using Kasada & AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV304 Level: advanced Type: Lightning talk Category: ISV & Partners Topics: Agentic AI; Voice & Conversational AI; Manufacturing & Industry 4.0 AI agents are powerful but riskythey can access sensitive data, trigger workflows, and make autonomous decisions. Kasada and AWS are helping enterprises adopt agents with confidence through comprehensive AI agent trust management that protects legitimate AI agents while detecting and blocking malicious automated visits. Kasada's platform, integrated with AWS services, enables organizations to distinguish trusted agent traffic from sophisticated bot threats, monitor for anomalous behavior, and maintain agent integrity against evolving AI-powered attacks. Join Kasada and AWS experts to explore a practical framework for managing agent trust: how AI and agentic traffic are being abused today, where risks appear across discovery and checkout, how teams decide when to allow or block agents, what new protocols like Web Bot Auth do and where they fall short, and what Kasada has built for agent traffic visibilityall while maintaining seamless customer experience. ### Live monitored sources - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Stripe introduces Link, a digital wallet that autonomous AI ...](https://postofday.com/2026/05/01/stripe-introduces-link-a-digital-wallet-that-autonomous-ai-agents-can-use-too/) — postofday.com (2026-05-05): Reports indicate that MoonPay and the Agentic Experience Protocol (AXP) have launched functional agent payment infrastructure (April-May 2026), with AXP extending the Universal Commerce Protocol (UCP) to support unified agentic commerce experiences and rich product data. - [What’s New in Agent 365: May 2026 | Microsoft Community Hub](https://techcommunity.microsoft.com/blog/agent-365-blog/what%E2%80%99s-new-in-agent-365-may-2026/4516340) — techcommunity.microsoft.com (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding --- ## STP302 — Unleash Live: Cloud-Powered Vision for Infrastructure URL: https://aws-summit-2026-kb.pages.dev/sessions/STP302 Level: advanced Category: Startups Topics: Media & Entertainment; Security, Identity & Compliance; Analytics, Redshift & Generative BI; Compute: EC2, Graviton & Nitro What happens when live video meets AI and the scalability of AWS This session explores how Unleash live harnesses AWS to deliver real-time vision analytics, moving from ingestion to insight in milliseconds. We detail the architecture of cloud-native pipelines that process live streams at scale and apply custom computer vision models across the energy, security, and infrastructure sectors. By combining edge connectivity with AWSs elastic infrastructure, Unleash live transforms drone and CCTV feeds into actionable intelligence. Attendees will gain insights into key design decisions and learn how cloud-based AI optimises operations, reduces risk, and unlocks the speed that modern physical AI demands. ### Live monitored sources - [Experian Announces Agent Trust to Power Trusted AI ...](http://businesswire.com/news/home/20260430719198/en/Experian-Announces-Agent-Trust-to-Power-Trusted-AI-Driven-Commerce) — businesswire.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Cohesity and ServiceNow Deliver Real-Time Recovery for ...](http://cohesity.com/newsroom/press/cohesity-servicenow-deliver-resilience-for-enterprise-ai-agents) — cohesity.com (2026-05-10): Research Publication: The Cloud Security Alliance (CSA) released a research note on May 8, 2026, formalizing 'Promptware'—a class of prompt injection attacks that function as malware to create Agentic Command and Control (C2) infrastructure. Risk: The research highlights that L - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [proofpoint.com](https://www.proofpoint.com/us/products/ai-mcp-security) — proofpoint.com (2026-05-01): Proofpoint provides an MCP Security Platform to secure AI connectivity at scale by routing MCP traffic through a central gateway. The platform enables centralized discovery and risk classification of 'shadow' MCP servers, enforces authentication via OAuth 2.0, controls user and a --- ## CMP501 — Nitro Isolation Engine: Formally Verifying Confidentiality URL: https://aws-summit-2026-kb.pages.dev/sessions/CMP501 Type: Chalk talk Category: Compute Topics: Compute: EC2, Graviton & Nitro What does it mean for the data of a virtual machine to be confidential Answering takes us on a journey through low-level systems and high-level mathematics. At re:Invent 2025, Graviton5 was introduced with the AWS Nitro Isolation Engine, a new software component enforcing isolation between virtual machines that was designed from the beginning with formal verification as a first-class consideration. You will learn about the hardware and software that isolate guest virtual machines, our mathematical definition of confidentiality, and the proofs used to establish this property for the Nitro Isolation Engine. No background in virtualization or formal methods is assumed. --- ## INO201 — Build and scale AI: from reliable agents to transformative systems URL: https://aws-summit-2026-kb.pages.dev/sessions/INO201 Level: intermediate Type: Breakout session Category: Other Topics: Agentic AI; Generative AI & Foundation Models Many teams move fast with agentic AI prototypes that impress in demos but stall in productionblocked by gaps in reliability, accuracy, and safety. In this session, AWS agentic AI technical leaders will help builders rethink how to build and scale production-grade, trustworthy agentic AI. Learn proven patterns to build and deploy agents that earn trust in the real world. See AWS agentic AI platform Amazon Bedrock AgentCore in action. Discover how AWS customers move fast from sparks of experiment to scaled AI-driven innovation with trust at the core, transforming industries. ### Live monitored sources - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## SEC401 — Advanced AI Security: Architecting Defense-in-Depth for AI Workloads URL: https://aws-summit-2026-kb.pages.dev/sessions/SEC401 Level: expert Type: Breakout session Category: Security, Identity & Compliance Topics: Agentic AI; Industry Spotlight: Public Sector & Government; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) Dive deep into advanced security architectures for AI workloads, exploring how to protect your workload against sophisticated attack vectors. Through technical examples, we'll implement secure architectures for AI workloads, covering identity, fine-grained access policies, and secure foundation model deployment patterns. Learn how to harden generative and agentic AI applications using AWS security capabilities, implementing least-privilege controls, and building secure architectures at scale. ### Live monitored sources - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-05): The CSAI Foundation (Cloud Security Alliance) announced milestones to secure the agentic control plane, including the strategic acquisition/stewardship of two foundational specifications: the Autonomous Action Runtime Management (AARM) specification (for securing AI-driven action - [AI Agent Delegation Patterns: Four Best Architectures for 2026 | Fastio](https://fast.io/resources/ai-agent-delegation-patterns) — fast.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Agentic Identity and Access Management](https://www.coalitionforsecureai.org/wp-content/uploads/2026/04/agentic-identity-and-access-control.pdf) — coalitionforsecureai.org (2026-05-08): New implementation patterns for AI agent identity (updated May 6, 2026) highlight the convergence of the Model Context Protocol (MCP) for agent-server handshakes and OAuth 2.1 with Dynamic Client Registration (DCR) for runtime credential issuance. A key pattern is the use of 'dis --- ## AIM301 — Commbank pioneering AI-driven DevSecOps with AWS DevOps Agent URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM301 Level: advanced Type: Breakout session Category: AI & Machine Learning Topics: DevOps, CI/CD & DevSecOps; Media & Entertainment; Resilience & Disaster Recovery CBA is achieving operational excellence by harnessing the power of the AWS DevOps Agent, part of AWS's new Frontier Agents. In this session, discover how CBA is using AI-driven automation to streamline incident response, reduce operational friction, and strengthen resilience across critical systems. We'll discuss CBA's cloud transformation journey and operational challenges, explore the DevOps Agent implementation including architecture, integration, and user journeys, and share results and business impact with real-world metrics. You'll see how automated remediation, and proactive insights are helping teams move faster with greater confidence. Join us to discover how CBA is shaping a future where operations are smarter, safer, and built for scale. ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [IBM Consulting Expands AI Capabilities to Accelerate Enterprise Transformation](https://newsroom.ibm.com/2026-05-06-ibm-consulting-expands-ai-capabilities-to-accelerate-enterprise-transformation) — newsroom.ibm.com (2026-05-08): IBM announced an expansion of its AI capabilities through 'IBM Enterprise Advantage' and 'IBM Consulting Advantage,' including the 'Agent2Agent (A2A)' interoperability standard to allow multi-agent orchestration across enterprise ecosystems (e.g., watsonx Orchestrate and SAP's Jo - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## ARC303 — Unlock GenAI inference anywhere with Amazon EKS Hybrid Nodes URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC303 Level: advanced Type: Breakout session Category: Architecture Topics: Streaming & Real-Time Data; Containers: EKS, ECS & Fargate; Generative AI & Foundation Models Join this session to explore how Amazon EKS Hybrid Nodes enables GenAI inference anywhere. We'll discuss reference architectures for adding on-prem GPUs to your EKS hybrid cluster, and for running real-time data capture and processing at the edge. You'll learn how EKS Hybrid Nodes enables seamless integration between the cloud and your on-prem or edge environments. Well also walk through a real-world example, showcasing how to accelerate GenAI inference at the edge using Amazon EKS Hybrid Nodes with NVIDIA DGX platform. ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## ARC401 — The Art of Managing Trade-Offs for your Network Design with Megaport URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC401 Level: expert Type: Breakout session Category: Architecture Topics: Networking & Edge; Security, Identity & Compliance Every network design decision involves careful trade-off considerations. In this session, master the art of making critical network design decisions through real-world scenarios that showcase Megaport's innovative NaaS platform for AWS connectivity. Learn how to evaluate key trade-offs between centralized and distributed architectures while balancing security, performance, and cost requirements. Explore how Megaport's on-demand bandwidth, private connectivity to AWS Direct Connect, and multi-cloud capabilities transform network design decisions. Whether you're building a global network, managing multi-account environments, or implementing hybrid connectivity to on-premises locations, attendees will leave with actionable insights and practical decision-making tools to optimize network infrastructure architecture designs with Megaport and AWS ### Live monitored sources - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [MCP Security Gateway - Agent Governance Toolkit](https://microsoft.github.io/agent-governance-toolkit/tutorials/07-mcp-security-gateway/) — microsoft.github.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [AI Agent Delegation Patterns: Four Best Architectures for 2026 | Fastio](https://fast.io/resources/ai-agent-delegation-patterns) — fast.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [MCP Governance (2026): Policy Gates for MCP Servers](https://cordum.io/blog/mcp-governance-servers) — cordum.io (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding --- ## DAT301 — Powering your Agentic AI experience with AWS Streaming and Messaging URL: https://aws-summit-2026-kb.pages.dev/sessions/DAT301 Level: advanced Type: Breakout session Category: Databases Topics: Agentic AI; Streaming & Real-Time Data; Generative AI & Foundation Models; Serverless: Lambda & Step Functions Powering your Agentic AI experience with AWS Streaming and MessagingOrganizations are accelerating innovation with generative AI and agentic AI use cases. This session explores how AWS streaming and messaging services such as Amazon Managed Streaming for Apache Kafka, Kinesis Data Streams, Amazon Managed Service for Apache Flink, and Amazon SQS build intelligent, responsive applications. Discover how streaming supports real-time data ingestion and processing, while messaging ensures reliable coordination between AI agents, orchestrates workflows, and delivers critical information at scale. Learn architectural patterns that highlight how a unified approach acts on data as fast as needed, providing the reliability and scale to grow for your next generation of AI. ### Live monitored sources - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [How to Build an Agentic AI Strategy With Process Intelligence](http://skan.ai/blogs/process-intelligence-for-agentic-ai-enterprise-automation) — skan.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## DAT401 — Real-Time DataLakes with Apache Iceberg, Amazon MSK, and Amazon S3 URL: https://aws-summit-2026-kb.pages.dev/sessions/DAT401 Level: expert Type: Breakout session Category: Databases Topics: Cost Optimization & FinOps; Streaming & Real-Time Data; Storage: S3, EBS & EFS; Data Lakes, Lakehouse & AI-Ready Data; Analytics, Redshift & Generative BI Learn how to optimize Apache Iceberg data lakes on Amazon S3 for cost-effectiveness while enabling real-time analytics. This session explores S3 Tables deployments, focusing on streaming data from Apache Kafka via Amazon MSK into Iceberg format. Discover practical approaches for real-time table maintenance, metadata optimization for high-velocity writes, and data compaction strategies. Implement cost-effective retention policies using S3 Lifecycle configurations while maintaining sub-minute data freshness. See how MSK's native Iceberg integration eliminates pipeline overhead, reducing latency and operational costs. Gain actionable insights for balancing streaming performance with cost optimization at scale. ### Live monitored sources - [Talent Harbor | Sales Recruitment as a Service (RaaS)](http://talentharbor.com/) — talentharbor.com (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Outcome-based pricing for AI Agents - Sierra](http://sierra.ai/blog/outcome-based-pricing-for-ai-agents) — sierra.ai (2026-05-11): Sierra announced an outcome-based pricing model for its AI agents, ensuring that the company is only paid when its AI agents drive real, tangible results for the business, aligning cost directly with success. - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent --- ## DEV311 — Serverless Developer Experience: Day in a life of builder URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV311 Level: advanced Type: Breakout session Category: Developer Tools Topics: Serverless: Lambda & Step Functions; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) What does it mean to be a serverless developer in the era of GenAI What disciplines do you need to master to build cloud-native, serverless solutions today In this session, we'll walk through a day in the life of a serverless developer and explore the core principles, architecture patterns, frameworks, and how to leverage GenAI tools to build your next-generation serverless application. ### Live monitored sources - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Red Hat adds support for agentic AI development | CIO](https://www.cio.com/article/4169833/red-hats-message-to-enterprises-you-dont-need-to-re-platform-for-ai-agents-2.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. --- ## MAM304 — Modernize SQL Server & .NET Together with AWS Transform's New AI Agent URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM304 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Agentic AI Want to cut your Microsoft licensing costs by up to 70% while modernizing SQL Server workloads Join us to explore AWS Transform's groundbreaking ### Live monitored sources - [AI Agent Memory Systems Cut Costs 60% with Long-Term Context 2026](https://iterathon.tech/blog/ai-agent-memory-systems-implementation-guide-2026) — iterathon.tech (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [ServiceNow expands AI agent governance through deeper ...](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-expands-AI-agent-governance-through-deeper-integration-with-Microsoft/default.aspx) — newsroom.servicenow.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Guild Raises $44M to Build the Agent Control Plane](https://www.guild.ai/knowledge/guild-raises-44m-agent-control-plane) — guild.ai (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Agentic Identity and Access Management](https://www.coalitionforsecureai.org/wp-content/uploads/2026/04/agentic-identity-and-access-control.pdf) — coalitionforsecureai.org (2026-05-08): New implementation patterns for AI agent identity (updated May 6, 2026) highlight the convergence of the Model Context Protocol (MCP) for agent-server handshakes and OAuth 2.1 with Dynamic Client Registration (DCR) for runtime credential issuance. A key pattern is the use of 'dis --- ## SEC302 — Leap ahead in Cloud Operations with AWS DevOps Agent URL: https://aws-summit-2026-kb.pages.dev/sessions/SEC302 Level: advanced Type: Breakout session Category: Security, Identity & Compliance Topics: Observability & Monitoring; DevOps, CI/CD & DevSecOps; Manufacturing & Industry 4.0 Downtime costs revenue. Alert fatigue burns out your best engineers. Manual incident investigation wastes hours that could be spent building. Every cloud team faces these operational challenges, yet most still rely on tribal knowledge and context-switching across multiple tools to diagnose issues. In this session, we demonstrate how AWS DevOps Agent transforms incident response from hours of manual investigation to minutes of autonomous analysis. Watch as the agent automatically correlates data across your observability tools, identifies root causes, and delivers actionable mitigation plans freeing your team to build instead of firefight. ### Live monitored sources - [Gartner 2026 Confirms It: The Context Graph Is the Missing ...](https://thecontextgraph.co/memos/gartner-2026-ai-agents-decision-intelligence-sales) — thecontextgraph.co (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [GitHub - Siddhant-K-code/agent-trace: strace for AI agents. Capture and replay every tool call, prompt, and response from Claude Code, Cursor, Gemini CLI or any MCP client · GitHub](https://github.com/Siddhant-K-code/agent-trace) — github.com (2026-05-04): The 'agent-trace' developer tool (GitHub: Siddhant-K-code/agent-trace) has launched significant new monitoring and control features: 1) A 'watch' mode that automatically terminates agents (using SIGSTOP or SIGTERM) when specific rules in a .watch-rules.json file are triggered, su - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [AI Agent Data Governance: Enterprise Playbook for 2026](https://promethium.ai/guides/ai-agent-data-governance-enterprise-playbook-2026/) — promethium.ai (2026-05-02): Core idea: contextgraph.tech serves as a resource for 'Context Graphs,' defined as living records of decision traces that capture the specific reasoning, exceptions, and context (the 'why') behind AI decisions. It identifies TrustGraph and Graphiti (by Zep) as leading open-source --- ## DEV210 — AI-Driven Incident Triage: From Slack Alert to Root Cause URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV210 Level: intermediate Category: Developer Tools Topics: Observability & Monitoring; Containers: EKS, ECS & Fargate; Analytics, Redshift & Generative BI; Manufacturing & Industry 4.0 Modern AWS environments generate more alerts than teams can realistically investigate. This session demonstrates a proof-of-concept that transforms Slack alerts into automated investigation workflows using AI.Learn how to trigger parallel queries across CloudWatch, Amazon EKS, Prometheus, and deployment history when an alert fires — returning correlated summaries with probable causes and dashboard links directly in Slack.You'll leave understanding practical integration patterns for AI-assisted triage, telemetry hygiene requirements, and guardrails for safely introducing AI into production incident response. Discover how AI augments — rather than replaces — your existing observability stack, meaningfully reducing time-to-insight during incidents. ### Live monitored sources - [AI Agent Context Window Cost: Why Bills Multiply [2026]](https://www.waxell.ai/blog/ai-agent-context-window-cost) — axell.ai (2026-05-05): Waxell published an analysis on the compounding cost of AI agent context windows, detailing how naive history management leads to 3x-5x budget underestimation. They proposed a runtime enforcement architecture (Waxell Runtime) that operates in the execution path to enforce hard to - [Scaling Autonomous Agent Swarms with Distributed Task ...](https://martinuke0.github.io/posts/2026-03-31-scaling-autonomous-agent-swarms-with-distributed-task-orchestration-and-low-latency-communication-protocols/) — martinuke0.github.io (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [datadoghq.com](https://www.datadoghq.com/blog/agentic-ai-llm-observability-bedrock-agentcore) — datadoghq.com (2026-04-20): Datadog — Blog: 'Operating agentic AI with Amazon Bedrock AgentCore and Datadog LLM Observability' (NTT DATA guest post). Published April 7, 2026. Describes a validated integration pattern combining Amazon Bedrock AgentCore for execution and Datadog LLM Observability for tracing, - [pwc.com](https://www.pwc.com/us/en/technology/alliances/library/deploying-agentic-ai-at-enterprise-scale-with-amazon-bedrock-agentcore.html) — pwc.com (2026-04-17): PwC article that had been removed was restored. The page now hosts the full case study 'Deploying agentic AI at enterprise scale with Amazon Bedrock AgentCore' (Apr 10, 2026) describing a production multi-agent deployment: supervisor-led routing, isolated runtimes per agent, tool - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## ISV206 — Scaling RAG to Millions of Vectors: The Squiz Story URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV206 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Serverless: Lambda & Step Functions; Media & Entertainment; Storage: S3, EBS & EFS; Retrieval Augmented Generation (RAG) Squiz, a global Digital Experience Platform provider, is transforming how organizations deliver conversational search experiences. By adopting Amazon S3 Vectors, Squiz reimagined its ingestion pipeline — increasing data processing speed by 50% and shifting from bespoke, always-on infrastructure to a scalable serverless model. This allows Squiz to seamlessly scale from 25,000 to millions of vectors per client, while significantly reducing costs. Hear how this shift freed engineering teams to focus on RAG innovation rather than infrastructure management, and how it powers smart video search capabilities across their platform. ### Live monitored sources - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Agentic RAG Explained: AI Agents + RAG in 2026](https://freeacademy.ai/blog/agentic-rag-ai-agents-supercharge-retrieval-2026) — freeacademy.ai (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [Agentic RAG: The 2026 Production Guide | MarsDevs](https://www.marsdevs.com/guides/agentic-rag-2026-guide) — marsdevs.com (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## STP203 — Build, Evaluate and Scale Production ready Agents with AWS Containers URL: https://aws-summit-2026-kb.pages.dev/sessions/STP203 Level: intermediate Category: Startups Topics: Containers: EKS, ECS & Fargate Building an agent that works once is easy; building an agent that works reliably for thousands of users is an architectural challenge. This session bridges the gap between experimental notebooks and deployed systems, focusing on the specific engineering disciplines needed for success. Join us to learn practical strategies for: 1. System Design: architecting decoupled, scalable agent backends from day one. 2. Continuous Evaluation: moving beyond "vibes-based" testing to metrics-driven evaluation suites that ensure reliability. 3. DevEx & Tooling: streamlining the developer experience to tighten feedback loops and ship improvements faster using open-source frameworks. ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Red Hat adds support for agentic AI development | CIO](https://www.cio.com/article/4169833/red-hats-message-to-enterprises-you-dont-need-to-re-platform-for-ai-agents-2.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs --- ## TNC101 — Exam Prep: AWS Certified AI Practitioner URL: https://aws-summit-2026-kb.pages.dev/sessions/TNC101 Level: foundational Type: Breakout session Category: Other Prepare for the AWS Certified AI Practitioner exam with this study session. Review questions from the official practice question set. Ask questions of the subject matter expert leading the study session. Continue preparing for your exam with confidence following the four-step plan on AWS Skill Builder. --- ## ARC201 — Building on AWS resilience: Innovations for critical success URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC201 Level: intermediate Type: Breakout session Category: Architecture Topics: Industry Spotlight: Public Sector & Government; Resilience & Disaster Recovery Essential services that power global economies and critical infrastructure demand exceptional resilience. Through nearly two decades of focused innovation, AWS has developed core engineering practices and operational approaches that power critical workloads worldwide. Explore how AWS's architectural innovations and organizational practices help customers build robust services that maintain resilience during severe disruptions. Learn how AWS's continued investment in resilience provides the foundation for delivering essential services across governments, economies, and critical infrastructure. ### Live monitored sources - [Red Hat adds support for agentic AI development | CIO](https://www.cio.com/article/4169833/red-hats-message-to-enterprises-you-dont-need-to-re-platform-for-ai-agents-2.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs - [ServiceNow’s AI control tower offers hazy view of spend | CIO](https://www.cio.com/article/4169954/servicenows-ai-control-tower-offers-hazy-view-of-spend.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [How enterprises are building AI agents in 2026 | Claude](https://claude.com/blog/how-enterprises-are-building-ai-agents-in-2026) — claude.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach --- ## ARC302 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC302 Level: advanced Type: Breakout session Category: Architecture Topics: Serverless: Lambda & Step Functions; Compute: EC2, Graviton & Nitro; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) In this session, learn about AWS Lambda's execution environment lifecycle, diving deep into how the service manages isolation at the function level, and understanding the security implications of environment reuse patterns. Learn about traditional patterns for compute isolation in multi-tenant environments, as well as explore Lambda's tenant isolation mode - a new powerful capability that enables tenant-level compute separation without operational overhead. Explore how to implement robust tenant isolation strategies, manage state across executions, and leverage Lambda's security boundaries effectively. Whether building new SaaS applications or enhancing existing ones, leave with practical knowledge to implement secure multi-tenant architectures at scale. ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AI Agent Rate Limiting Strategies & Best Practices](https://fast.io/resources/ai-agent-rate-limiting/) — fast.io (2026-05-10): Arcjet introduced 'Guards,' a runtime security service for AI agent workflows that enables enforcement of per-user token budgets and spend limits inside agent loops and can detect prompt injection in tool results. - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## ARC307 — AI Powered Resilience Lifecycle URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC307 Level: advanced Type: Breakout session Category: Architecture Topics: Voice & Conversational AI; Security, Identity & Compliance; Resilience & Disaster Recovery; Generative AI & Foundation Models; Manufacturing & Industry 4.0; Retrieval Augmented Generation (RAG) Not all disaster recovery strategies can address the complex, dynamic nature of modern cloud infrastructures, leading to gaps in system resilience and compliance adherence. Discover how to enhance resilience and disaster recovery on AWS empowered by AI. This approach bridges infrastructure insights and application-level testing, enabling more effective disaster recovery preparation. You will learn how to leverage Large Language Models (LLMs) with AWS Resilience Hub and AWS Systems Manager to modernize testing, analyze infrastructure, and generate targeted AWS Fault Injection Service experiments and recovery runbooks. Walk away with practical examples of automated test generation with templates and learn to design prompts. ### Live monitored sources - [When prompts become shells: RCE vulnerabilities in AI agent ...](https://www.microsoft.com/en-us/security/blog/2026/05/07/prompts-become-shells-rce-vulnerabilities-ai-agent-frameworks) — microsoft.com (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [AI Agent Rate Limiting Strategies & Best Practices](https://fast.io/resources/ai-agent-rate-limiting/) — fast.io (2026-05-10): Arcjet introduced 'Guards,' a runtime security service for AI agent workflows that enables enforcement of per-user token budgets and spend limits inside agent loops and can detect prompt injection in tool results. - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [[2602.21012] International AI Safety Report 2026 - arXiv.org](https://arxiv.org/abs/2602.21012) — arxiv.org (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. --- ## ARC402 — DynamoDB: Resilience & lessons from the Oct 2025 service disruption URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC402 Level: expert Type: Breakout session Category: Architecture Topics: Resilience & Disaster Recovery; Security, Identity & Compliance; Databases & Aurora In this session, we will walk through the architecture for the Amazon DynamoDB DNS management system that triggered the service disruption on October 20, 2025. We will share the lessons that the DynamoDB team learned from this event and explain how we are using these insights to improve both DynamoDB and AWS. You will walk away with actionable knowledge that you can apply to the systems you build. ### Live monitored sources - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [AI Agent Context Window Cost: Why Bills Multiply [2026]](https://www.waxell.ai/blog/ai-agent-context-window-cost) — axell.ai (2026-05-05): Waxell published an analysis on the compounding cost of AI agent context windows, detailing how naive history management leads to 3x-5x budget underestimation. They proposed a runtime enforcement architecture (Waxell Runtime) that operates in the execution path to enforce hard to --- ## ARC403 — Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep Dive URL: https://aws-summit-2026-kb.pages.dev/sessions/ARC403 Level: expert Type: Breakout session Category: Architecture Topics: Serverless: Lambda & Step Functions; Compute: EC2, Graviton & Nitro; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) Secure Multi-tenant SaaS with AWS Lambda: A Tenant Isolation Deep DiveIn this session, learn about AWS Lambda's execution environment lifecycle, diving deep into how the service manages isolation at the function level, and understanding the security implications of environment reuse patterns. Learn about traditional patterns for compute isolation in multi-tenant environments, as well as explore Lambda's tenant isolation mode - a new powerful capability that enables tenant-level compute separation without operational overhead. Explore how to implement robust tenant isolation strategies, manage state across executions, and leverage Lambda's security boundaries effectively. Whether building new SaaS applications or enhancing existing ones, leave with practical knowledge to implement secure multi-tenant architectures at scale. ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AI Agent Rate Limiting Strategies & Best Practices](https://fast.io/resources/ai-agent-rate-limiting/) — fast.io (2026-05-10): Arcjet introduced 'Guards,' a runtime security service for AI agent workflows that enables enforcement of per-user token budgets and spend limits inside agent loops and can detect prompt injection in tool results. - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## DAT201 — Scaling Data Analytics: Easygo's Modern Lakehouse Journey on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/DAT201 Level: intermediate Type: Breakout session Category: Databases Topics: Data Lakes, Lakehouse & AI-Ready Data; Streaming & Real-Time Data; Databases & Aurora; Analytics, Redshift & Generative BI Discover how Melbourne-based Easygo, powering Stake and Kick.com, transformed their data analytics infrastructure to process over 600,000 daily transactions and tens of millions of streaming events. Learn about their implementation of a modern lakehouse architecture combining Amazon Aurora Zero-ETL integration with Amazon Redshift, Amazon Kinesis with AWS Glue streaming, and Apache Iceberg on Amazon S3. Results include 95% faster queries, 80% fewer ingestion incidents, 9 hours weekly maintenance savings, and accelerated global expansion. Explore practical strategies for building scalable, secure data foundations delivering near real-time analytics with robust governance across regulated markets. ### Live monitored sources - [The AI Agent Infrastructure Landscape in 2026: A Practitioner ...](https://chenagent.dev/articles/ai-agent-infrastructure-landscape-2026) — chenagent.dev (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. - [How to Secure Vector Stores for AI Agents in 2025 | Fastio](https://fast.io/resources/ai-agent-vector-store-security/) — fast.io (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Notion launches its first AI agents for data analysis and task ... - MLQ.ai](http://mlq.ai/news/notion-launches-its-first-ai-agents-for-data-analysis-and-task-automation) — mlq.ai (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. --- ## DAT303 — Explore whats new in data and AI governance with SageMaker Catalog URL: https://aws-summit-2026-kb.pages.dev/sessions/DAT303 Level: advanced Type: Breakout session Category: Databases Topics: Machine Learning & SageMaker; Analytics, Redshift & Generative BI; Security, Identity & Compliance; Databases & Aurora Join this session to learn about the latest capabilities in Amazon SageMaker Catalog that help organizations govern data and AI more effectively. We will walk through new features that make it easier to discover, govern, and securely share structured and unstructured data, models, business intelligence dashboards, and applications. Youll hear how customers are using these capabilities to improve data discovery and access, streamline compliance, and support AI initiatives. ### Live monitored sources - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [draft-klrc-aiagent-auth-01 - AI Agent Authentication and ...](https://datatracker.ietf.org/doc/draft-klrc-aiagent-auth/) — datatracker.ietf.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## MAM305 — Legacy App modernization and reverse engineering using Kiro URL: https://aws-summit-2026-kb.pages.dev/sessions/MAM305 Level: advanced Type: Breakout session Category: Migration & Modernization Topics: Agentic AI; Migration & Modernization; Kiro & Spec-Driven Development This session demonstrates how agentic AI systems revolutionize legacy application modernization through intelligent reverse engineering workflows. We showcase a multi-agent architecture that autonomously analyzes legacy codebases, generates comprehensive business specifications, and rebuilds cloud-native applications while preserving critical business logic. Our agentic framework orchestrates specialized AI agentsAnalyzer, Generator, Evaluator, and Refinerworking collaboratively to extract institutional knowledge from decades-old systems and validate modernized applications against original code. Drawing from real-world enterprise implementations including Commonwealth Bank of Australia's transformation, we present practical patterns for deploying autonomous AI agents that achieve 70% efficiency gains in modernization cycles. ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Meow Technologies launches the first agentic banking ...](http://thenextweb.com/news/meow-technologies-agentic-banking-ai-agents) — thenextweb.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits --- ## AIM101 — AI League Championship | 14-May | 08:00 - 16:00 URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM101 Level: foundational Type: Lightning talk Category: AI & Machine Learning Topics: Agentic AI; Voice & Conversational AI; Generative AI & Foundation Models Experience the ultimate AI showdown from your theater seat as finalists from the AWS AI League: Agentic AI challenge battle head-to-head in real-time. Watch top performers compete in a live, tournament-style competition where their AI agents navigate complex challenges, demonstrating the power of intelligent automation and decision-making under pressure. This electrifying viewing experience offers AI practitioners and enthusiasts unique insights into orchestrating intelligent agents using Amazon Bedrock AgentCore. See firsthand how agents handle real-world scenarios including code execution, web browsing, content moderation, and multi-step problem solvingall while competing for championship glory. The 1st place winner will represent Australia at the 2026 AWS AI League Championship at re:Invent 2026. ### Playbook (editorial commentary) **The concept.** All-day live tournament: AWS AI League finalists run agents head-to-head against complex challenges (code execution, web browsing, content moderation, multi-step problem solving). Winner represents Australia at re:Invent 2026. **Why it matters.** Demos lie. Tournaments don't. Watching agents fail under tournament pressure teaches more than 100 vendor decks. **The hard parts.** Tournament metrics may not match your business metrics. "Wins championship" ≠ "ships in your environment." Don't confuse benchmark wins with deployment fitness. **Playbook moves.** (1) Send people who'll watch *how* agents fail, not just who wins. (2) Capture the design patterns the leaders use — those usually preview AWS's reference architectures. (3) Note the tasks where *all* finalists struggle; those are the unsolved problems where your team can't expect easy wins either. **The surprise.** The most useful observation in any AI tournament isn't who wins — it's the *variance* in approach. When the same problem yields radically different agent designs, the design space is still wide open and your team's bets remain differentiable. When approaches converge, the pattern is solved and commoditising. --- ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. --- ## KEY002 — Innovation Day Keynote URL: https://aws-summit-2026-kb.pages.dev/sessions/KEY002 Type: Keynote Category: Keynote Innovation Day Keynote ### Playbook (editorial commentary) **The concept.** Major announcements + roadmap signaling. **Why it matters.** Keynotes telegraph AWS investment direction. Where they put capex hints at where the platform is going next. **The hard parts.** Keynote announcements often arrive months ahead of GA. Don't architect on slides. **Playbook moves.** (1) Note announcements; check actual availability before committing. (2) Budget 6–12 months from keynote to enterprise-ready. (3) Watch the deprecation announcements — they affect this quarter's spend. **The surprise.** The most actionable keynote signal is usually the *price changes* and *deprecations*, not the new shiny things. Those move budgets immediately. New features move budgets next year. --- --- ## AIM403 — AI League URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM403 Level: expert Type: Gamified learning Category: AI & Machine Learning Topics: Agentic AI; Gaming & Interactive Media; Databases & Aurora; Generative AI & Foundation Models Building intelligent AI Agents with Amazon Bedrock AgentCore Dive into the future of Al in this competitive workshop where you'll create intelligent agents using Amazon Bedrock AgentCore. Using our no-code Ul interface and generative AI you will develop agents and agentic tools that tackle real-world technical challenges. Experience hands-on learning through dynamic gameplay with live leaderboards while mastering prompt engineering, safety guardrails, tool creation, and performance optimization. Whether you are an avid Al developer or an architect eager to gain practical experience with production-grade agentic Al, this session will give you a new hands-on experience. ### Playbook (editorial commentary) **The concept.** Competitive 2-hour workshop: build agents and tools using AgentCore's no-code UI + generative AI. Live leaderboards. Prompt engineering, safety guardrails, tool creation, performance optimisation. **Why it matters.** Hands-on AgentCore experience compresses weeks of self-learning into one afternoon. **The hard parts.** No-code UIs hide important details. What you build in the workshop won't translate 1:1 to production code paths. **Playbook moves.** (1) Send a builder, not an observer. (2) Capture the underlying tool/prompt patterns rather than just the artifacts. (3) Compare your design to the leaderboard's top entries — divergence is the lesson. **The surprise.** The workshop's hidden value is seeing *what other teams build*. Compete-and-compare reveals the design patterns AWS expects you to adopt. Sometimes those patterns are the actual product roadmap signal six months early. --- ### Live monitored sources - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [43,750% Surge! BNB Chain is Crushing It with 150,000 AI Agents Blasting Through the Track | 小机构集团 on Binance Square](http://binance.com/en/square/post/316311861525314) — binance.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Amazon Builds AI Agent Payments With Coinbase and Stripe](https://thedefiant.io/news/infrastructure/amazon-builds-ai-agent-payments-with-coinbase-and-stripe) — thedefiant.io (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. --- ## FSI206 — Agentic AI Transforming Quality at Cloud Speed URL: https://aws-summit-2026-kb.pages.dev/sessions/FSI206 Level: intermediate Type: Breakout session Category: Financial Services Topics: Agentic AI; AWS Well-Architected Framework; Security, Identity & Compliance; Industry Spotlight: Financial Services Agentic AI Transforming Quality at Cloud SpeedFinancial services cloud transformations unlock tremendous opportunity - and quality assurance is the catalyst. The Quality Validation Agent Framework deploys specialised AI agents across the delivery lifecycle, collaborating through the AgentCore Runtime Platform to continuously monitor, validate, and accelerate work in real-time. Covering Assessment, Transformation, Testing, and Deployment Intelligence, these autonomous agents shift FSI organisations from reactive checkpoints to proactive validation - eliminating rework, detecting defects early, and achieving AWS Well-Architected compliance with confidence. Come and join us to discover how agentic AI turns quality assurance into your fastest path to production in banking and financial services! ### Playbook (editorial commentary) **The concept.** Specialised AI agents collaborating across the FSI delivery lifecycle (Assessment → Transformation → Testing → Deployment Intelligence). Continuous validation rather than checkpoint-based QA. **Why it matters.** Reactive quality gates are the FSI delivery bottleneck. Continuous validation ships faster without raising risk. **The hard parts.** "Continuous" requires agents to access privileged systems. Compliance teams will want a log of every agent decision. **Playbook moves.** (1) Start with one phase (testing is usually safest). (2) Prove the audit trail before extending to other phases. (3) Don't skip change control — agents inside change control is the goal, not agents outside it. **The surprise.** Most QA ROI claims focus on speed. The real win in FSI is *evidence*. Auditable decision logs from agents are paradoxically easier to produce than from humans. That's a regulatory advantage, not just a velocity one. --- ### Live monitored sources - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [43,750% Surge! BNB Chain is Crushing It with 150,000 AI Agents Blasting Through the Track | 小机构集团 on Binance Square](http://binance.com/en/square/post/316311861525314) — binance.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. --- ## IDE301 — Diversity In Tech - AI Literacy Skills - Rapid prototyping with Kiro URL: https://aws-summit-2026-kb.pages.dev/sessions/IDE301 Level: advanced Type: Workshop Category: Diversity, Equity & Inclusion Topics: Diversity, Equity & Inclusion in Tech; Kiro & Spec-Driven Development; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) In this workshop you will be shown how to build functional prototypes using our proven techniques. Learn how to leverage Kiro to go beyond "vibe-coding", and transform ideas into fully functional prototypes while validating technical approaches. Discover effective spec-driven development and prompt engineering techniques combining generative AI capabilities with AWS services for rapid iteration and refinement. If participants would like to follow along please bring your own laptop. ### Playbook (editorial commentary) **The concept.** Build functional prototypes using Kiro and spec-driven development. Move beyond "vibe-coding." **Why it matters.** Fast prototyping separates AI-native teams from AI-curious ones. **The hard parts.** Prototypes that work aren't products. The gap between "demo works" and "passes security review" is wider than ever with AI in the mix. **Playbook moves.** (1) Treat prototype output as a *validated spec*, not as production code. (2) Rewrite for production explicitly; resist the temptation to "harden" the prototype. (3) Capture the spec; throw away the code. **The surprise.** Spec-driven development's lasting value isn't speed — it's that the spec becomes the durable artifact. Code is increasingly disposable; specs are the IP. Adjust your IP policies and code-review priorities accordingly. --- ### Live monitored sources - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Top Tech News Today, May 7, 2026 - Tech Startups](https://techstartups.com/2026/05/07/top-tech-news-today-may-7-2026/) — techstartups.com (2026-05-08): Google shut down Project Mariner, its experimental AI browser agent, and folded its agentic AI capabilities into Gemini Agent and AI Mode. - [Fetched web page](https://mem0.ai/blog/6-techniques-to-cut-ai-agent-memory-cost-beyond-basic-retrieval) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin - [The 2026 Token Optimization Playbook: Cut AI Agent Memory Costs 3–4X](https://mem0.ai/blog/the-2026-token-optimization-playbook-cut-ai-agent-memory-costs-3–4x) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin --- ## ISV201 — MCP on EKS: Xero's AI-Driven Developer Experience URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV201 Level: intermediate Type: Breakout session Category: ISV & Partners Topics: Model Context Protocol (MCP); Containers: EKS, ECS & Fargate; Kiro & Spec-Driven Development; Code Generation & AI-Assisted Development; Retrieval Augmented Generation (RAG) AI coding agents are transforming how developers build and operate modern cloud-native applications. With tools such as Kiro CLI, Kiro IDE, or any MCP-compatible AI coding assistant, developers are embracing AI to move faster and scale smarter. This session explores how MCP servers help developers streamline code generation, deployment, and debugging by embedding infrastructure awareness directly into the AI assistant. Learn how Xero is leveraging MCP to speed up development, simplify operations, and deliver more reliable containerized apps at scale. Xero will also share their success story using Kiro CLI, Prometheus MCP, EKS MCP, and AWS Knowledge Base MCP to identify and resolve Prometheus cost spikesslashing costs by 40%. ### Playbook (editorial commentary) **The concept.** MCP servers embed infrastructure awareness directly into AI coding assistants. Xero's stack: Kiro CLI + Prometheus MCP + EKS MCP + AWS Knowledge Base MCP. Cut Prometheus costs by 40%. **Why it matters.** AI assistants without infrastructure context produce generic answers. With MCP, they reason about *your* deployment. **The hard parts.** MCP server quality varies. A poorly-configured MCP gives the agent confidently wrong context — worse than no context. **Playbook moves.** (1) Inventory the MCP servers your teams use. (2) Treat them like tools — versioned, owned, documented. (3) Audit what context they expose; over-broad MCPs leak. **The surprise.** The 40% Prometheus cost cut came from agents finding *cardinality bombs* humans missed. The gain wasn't AI being smarter — it was AI being patient enough to read every metric label. Use AI for tedious-but-tractable problems; that's where the 80% of the wins live. --- ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [GitHub - agentgateway/agentgateway: Next Generation Agentic Proxy for AI Agents and MCP servers · GitHub](https://github.com/agentgateway/agentgateway) — github.com (2026-05-07): Agentgateway released version v1.2.0-alpha.1, continuing the development of its open-source AI-native proxy for agent-to-agent and agent-to-tool communication. The project maintains approximately 2.6k GitHub stars and is part of the Linux Foundation. --- ## MAE203 — 27 Faster: How Service Stream Automated Work Order Verification with AI on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/MAE203 Level: intermediate Type: Breakout session Category: Media & Entertainment Every month, Service Stream's field operations generate over 1 million images requiring verification — proof that infrastructure maintenance work has been completed to contract specification. Manually reviewing this volume was resource-intensive, error-prone, and created payment bottlenecks that slowed contractor payment cycles and strained client relationships. To solve this, Service Stream partnered with AWS to build an AI-powered work order verification system. The system now processes up to 800 work orders per hour — 27 times faster than manual review — with improved accuracy, reduced subjectivity, and no increase in headcount. In this session, learn how Service Stream used AWS AI and cloud services to automatically check field images against contract requirements — delivering a production-ready solution in just three months. We'll walk through the business drivers, solution design, and real-world outcomes — including how eliminating manual bottlenecks improved ServiceS tream's ability to meet client commitments, reduce payment delays, and build trust across their contractor network. Whether you're in field services, utilities, infrastructure, or any operations-heavy industry, this session will show you how intelligent automation can transform high-volume verification workflows and deliver tangible business value. ### Playbook (editorial commentary) **The concept.** 1M+ images monthly checked against contract specs. AI replaces manual review. 800 work orders/hour. Production in 3 months. **Why it matters.** Image-verification at scale is one of the highest-ROI AI use cases in operations-heavy industries. The pattern generalises across utilities, infrastructure, field services. **The hard parts.** False negatives (missed defects) have legal and contractual consequences. Confidence threshold calibration is where the work lives. **Playbook moves.** (1) Define cost-per-error explicitly — false positive vs. false negative are not equal. (2) Tune thresholds against business cost, not just F1 score. (3) Keep a human-review tier for ambiguous cases. **The surprise.** The biggest payoff in this kind of system isn't the speed gain — it's *consistency*. Human reviewers vary by mood, fatigue, time of day. The AI doesn't. That consistency is what fixes payment disputes with contractors, not throughput. Frame the business case on consistency, not speed. --- --- ## PRT105-S — 3 examples of context-aware agents at work URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT105-S Level: foundational Type: Lightning talk Category: Partner Showcase As AI moves from assistants to agents, success depends on contextwhat the system attends to and acts on. In this lightning talk, well show three examples of context-aware agents in the enterprise through quick demos: meeting coordination, engineering issue resolution, and sales forecasting, highlighting how the right context enables reliable execution at work. ### Playbook (editorial commentary) **The concept.** Three quick demos: meeting coordination, engineering issue resolution, sales forecasting. Context determines reliable execution. **Why it matters.** Agents without your team's context produce generic answers. Generic answers are useless. **The hard parts.** Context aggregation across SaaS tools is a permission and freshness problem. Stale context is worse than no context. **Playbook moves.** (1) Audit which corpora the agent has access to. (2) Refresh frequency matters more than corpus breadth. (3) Build relevance filtering before broader access. **The surprise.** The hardest part of "context-aware" agents isn't access — it's *relevance*. Most enterprise data is noise to most queries. Relevance filtering is where the engineering work hides. --- --- ## PRT107-S — From Reactive to Preventative - The Path to Autonomous Operations URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT107-S Level: foundational Type: Lightning talk Category: Partner Showcase Technology leaders need operations that evolve from reactive to autonomous. Enable teams to resolve incidents faster with intelligent systems that handle routine issues automatically. Shift from reactive responses to prevention, tool sprawl to coordination, manual work to automation. See how an AWS-powered platform learns from every incident to deflect toil, protect revenue, and scale. ### Playbook (editorial commentary) **The concept.** Operations evolution: reactive → preventive → autonomous. AI handles routine incidents. Tool consolidation, automation. **Why it matters.** Tier-1 toil dominates ops cost. Automating it is the largest available lever. **The hard parts.** "Autonomous" requires write access. Most orgs are nowhere near comfortable with that. **Playbook moves.** (1) Stage by blast radius. Start with read-only diagnostics. (2) Earn write privileges per playbook, not all at once. (3) Track which decisions humans override — that's your model's edge. **The surprise.** The fastest path to reducing alert fatigue isn't more AI — it's killing alerts that nobody acts on. Audit your last 1,000 pages; deprecate the noise. AI on a clean alert pipeline outperforms AI on a dirty one by a wide margin. --- --- ## PRT210-S — Charting the CX Frontier: A Cohesive, AI-Enabled Engagement Platform URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT210-S Level: intermediate Type: Lightning talk Category: Partner Showcase Topics: Voice & Conversational AI; Retrieval Augmented Generation (RAG) A forward-looking view of how enterprises are transforming customer experience through unified, AI-driven architecture. By moving away from fragmented solutions to an integrated platform spanning AI-powered engagement, automation, and orchestration, organisations can deliver more consistent, scalable, and outcome-focused experiences in an AI-first era. ### Playbook (editorial commentary) **The concept.** Unified, AI-driven CX architecture. Engagement + automation + orchestration in one platform. **Why it matters.** Fragmented CX tools create fragmented customer experiences. Customers feel the seams between systems. **The hard parts.** Platform consolidation usually loses depth. The integrated platform may be weaker than best-of-breed in any single area. **Playbook moves.** (1) List your top 5 CX moments. (2) Score the integrated platform on each. (3) Don't consolidate on average score; consolidate on critical-moment score. **The surprise.** CX platform success correlates more with *operational excellence* (the team running it) than with platform features. Tools amplify team competence — they don't substitute for it. The buying decision matters less than the staffing decision. --- ### Live monitored sources - [The 2026 Token Optimization Playbook: Cut AI Agent Memory Costs 3–4X](https://mem0.ai/blog/the-2026-token-optimization-playbook-cut-ai-agent-memory-costs-3–4x) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin --- ## SMB202 — PMY Delivers Realtime Crowd Analytics at the F1 Australian Grand Prix URL: https://aws-summit-2026-kb.pages.dev/sessions/SMB202 Level: intermediate Type: Breakout session Category: Small & Medium Business Topics: Analytics, Redshift & Generative BI; Retrieval Augmented Generation (RAG) Major events produce fragmented data across CCTV, sensors, ticketing, and venue systems. PMY Group will show how Optic, built on AWS, brings these sources together to create real-time crowd intelligence. Using the Australian Grand Prix as a case study, this session explores how operators gained live visibility into movement and congestion to support faster operational decisions. It also highlights how the same AWS foundation can support scalable analytics and broader unified data outcomes across venues and events. ### Playbook (editorial commentary) **The concept.** Optic platform unifies CCTV + sensors + ticketing + venue systems. Real-time crowd intelligence at major events. **Why it matters.** Major events generate fragmented data; live operational decisions need unified views. **The hard parts.** Multi-source synchronisation at sub-second granularity. CCTV and sensor latency vary; alignment is non-trivial. **Playbook moves.** (1) Define the operational decision *before* the data architecture. (2) Start with one decision type (congestion routing); add more once it works. (3) Plan the post-event analytics from day one. **The surprise.** The real value of unified event analytics is post-event, not live. The same data informs next year's layout, staffing, and pricing. Most teams build for live and then have to retrofit for the more valuable post-event use case. --- ### Live monitored sources - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [How UKG taps workforce intelligence with the Agentic Data Cloud | Google Cloud Blog](https://cloud.google.com/blog/products/databases/how-ukg-taps-workforce-intelligence-with-the-agentic-data-cloud) — cloud.google.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Introducing Agent Relay](https://x.com/mattshumer_/status/2027605370470867280) — x.com (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Notion launches its first AI agents for data analysis and task ... - MLQ.ai](http://mlq.ai/news/notion-launches-its-first-ai-agents-for-data-analysis-and-task-automation) — mlq.ai (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. --- ## TNC203 — Structured Approach to AI coding with Spec-Driven Development on Kiro URL: https://aws-summit-2026-kb.pages.dev/sessions/TNC203 Level: intermediate Type: Lightning talk Category: Other Topics: Agentic AI; Code Generation & AI-Assisted Development; Kiro & Spec-Driven Development; Retrieval Augmented Generation (RAG) This session demonstrates how Kiro brings discipline and clarity to AI-assisted software development, ensuring generated code aligns with intended functionality and architecture. Explore Kiro's innovative spec-driven development approach for AI coding. Learn how to leverage structured specifications as a single source of truth, contrasting with unstructured 'vibe coding'. Discover how Kiro uses AI to generate detailed requirements, design, and task documents, guiding AI agents in code creation. Experience a workflow that enhances collaboration, maintainability, and documentation accuracy. ### Playbook (editorial commentary) **The concept.** Discipline + clarity for AI-assisted development. Specifications as single source of truth. Direct contrast with unstructured "vibe coding." **Why it matters.** Vibe coding doesn't scale. Specs make AI-generated code reviewable, maintainable, auditable. **The hard parts.** Writing good specs is harder than writing good code. Most engineers haven't been trained for it. **Playbook moves.** (1) Pair junior + senior on spec authoring initially. (2) Review specs as code. (3) Version them in the repo alongside the artefacts they generate. **The surprise.** Spec-driven development surfaces spec quality immediately — bad specs produce visibly bad code. That's a feature. Use it as a forcing function for product clarity rather than an inconvenience. --- ### Live monitored sources - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## WPS203 — Optimising Outpatient Waitlists with ML at Gold Coast Health URL: https://aws-summit-2026-kb.pages.dev/sessions/WPS203 Level: intermediate Type: Breakout session Category: Public Sector Topics: Industry Spotlight: Financial Services; Observability & Monitoring; Manufacturing & Industry 4.0; Data Governance & Privacy; Machine Learning & SageMaker; Industry Spotlight: Healthcare & Life Sciences Deploying ML in high-stakes environments demands enterprise readiness, governance, and continuous monitoring. In this session, you'll learn how Gold Coast Health moved from pilot to production with a predictive model identifying patients unlikely to attend procedures — achieving 33% precision, doubling the 15% manual baseline — while ensuring fairness across cohorts. The session covers real-world ML architecture on Amazon SageMaker Pipelines, production monitoring including data quality, pipeline health, and drift detection, plus navigating AI governance through bias analysis and impact assessment. Whether you're in healthcare, financial services, or any regulated industry, walk away with actionable patterns for deploying responsible ML at scale. ### Playbook (editorial commentary) **The concept.** ML predicting outpatient no-shows. 33% precision, doubling the 15% manual baseline. Production-grade with fairness analysis and drift detection. **Why it matters.** No-shows are expensive (idle clinical capacity); even modest improvement compounds across thousands of slots. **The hard parts.** Bias is a real concern. "Likely to no-show" must not correlate with protected attributes. Fairness analysis is non-optional in healthcare. **Playbook moves.** (1) Bias analysis at every model release. (2) Document the impact assessment. (3) Make the model interpretable to clinicians who'll act on it. **The surprise.** 33% precision means 67% false positives. Deployment success depends on what you *do* with the prediction — calling patients vs. removing slots vs. double-booking. The model is only as good as the workflow around it. Build the intervention design before chasing higher precision. --- ### Live monitored sources - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [Anthropic deepens Wall Street push with new AI agents, and ...](https://fortune.com/2026/05/05/anthropic-wall-street-financial-services-agents-jamie-dimon/) — fortune.com (2026-05-05): Anthropic has deployed production AI agents using Claude Opus 4.7 at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa for high-stakes financial workflows including KYC, underwriting, and insurance claims. AIG reported that these agents scored 88% as accurate as human experts on - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Enterprise AI Agents Move From Pilot to Production: What 2026 ...](https://insights.reinventing.ai/articles/ai-agents-enterprise-production-2026-02-25) — insights.reinventing.ai (2026-05-05): Microsoft's 2026 Data Security Index reports that more than 80% of Fortune 500 companies are now running active AI agents in production, integrated across sales, finance, customer service, and security workflows. --- ## PRT106-S — The AI Challenge You Don't Yet Know About - Software Supply Chain URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT106-S Level: foundational Type: Lightning talk Category: Partner Showcase Topics: DevOps, CI/CD & DevSecOps; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) Most teams have "done DevOps" but still face low platform adoption, rising cloud costs, lagging security, and fragile incident response. This session explores the "After" state: treating your platform as a product, with golden paths, built-in security, AI-driven simplicity, and reliability by defaultusing Harness as the model for a governed, cost-aware, AI-native developer experience. ### Playbook (editorial commentary) **The concept.** "After DevOps" state. Platform as a product, golden paths, built-in security, AI-driven simplicity, reliability by default. **Why it matters.** "Done DevOps" left many orgs with low platform adoption, rising cloud costs, security lag, fragile incident response. **The hard parts.** "Platform as a product" requires platform engineers who can talk to PMs, not just admins. That's a hiring problem, not a tooling problem. **Playbook moves.** (1) Treat platform adoption as the team's KPI. (2) Survey internal users like external customers — NPS, satisfaction, usage. (3) Kill features no one uses. **The surprise.** The most successful internal platforms have NPS scores in the 60s. If yours has angry users, no amount of new features fixes that — fix the relationship first. New features on a low-NPS platform often make NPS *worse*. --- ### Live monitored sources - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [GitHub - Siddhant-K-code/agent-trace: strace for AI agents. Capture and replay every tool call, prompt, and response from Claude Code, Cursor, Gemini CLI or any MCP client · GitHub](https://github.com/Siddhant-K-code/agent-trace) — github.com (2026-05-04): The 'agent-trace' developer tool (GitHub: Siddhant-K-code/agent-trace) has launched significant new monitoring and control features: 1) A 'watch' mode that automatically terminates agents (using SIGSTOP or SIGTERM) when specific rules in a .watch-rules.json file are triggered, su - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca --- ## PRT111-S — From Risk to Resilience - How Mimecast Works with AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT111-S Level: foundational Type: Lightning talk Category: Partner Showcase Topics: Resilience & Disaster Recovery; Security, Identity & Compliance; Analytics, Redshift & Generative BI Human risk is a critical layer of any security strategy. Human risk management addresses how employee behaviorfrom accidental sharing to shadow AI usecreates organizational exposure. Discover how Mimecast, on AWS, helps identify risky behavior, protect critical data and account access, and support compliance. Real-world insights. Behavioural analytics. Adaptive controls. Measurable ROI. ### Playbook (editorial commentary) **The concept.** Human behaviour as a security layer. Behavioural analytics + adaptive controls covering accidental sharing, account compromise, and shadow AI use. **Why it matters.** Most breaches start with human action. Detecting risky behaviour proactively reduces incidents. **The hard parts.** Behavioural analytics tread close to surveillance. Cultural acceptance matters as much as technical capability. **Playbook moves.** (1) Be transparent internally about what's monitored. (2) Tie controls to outcomes (block phishing) rather than surveillance (track who clicks). (3) Use behavioural data to drive training, not punishment. **The surprise.** "Shadow AI use" — employees feeding company data into public LLMs — is the new shadow IT. It's already everywhere; your DLP probably can't see most of it. Plan for it as a category, not a one-off incident. --- ### Live monitored sources - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [Experian Announces Agent Trust to Power Trusted AI ...](http://businesswire.com/news/home/20260430719198/en/Experian-Announces-Agent-Trust-to-Power-Trusted-AI-Driven-Commerce) — businesswire.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [AI Agent Authentication & Authorization in 2026: What Works ...](https://api.aport.io/blog/best-ai-agent-authentication-authorization-2026) — api.aport.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [proofpoint.com](https://www.proofpoint.com/us/products/ai-mcp-security) — proofpoint.com (2026-05-01): Proofpoint provides an MCP Security Platform to secure AI connectivity at scale by routing MCP traffic through a central gateway. The platform enables centralized discovery and risk classification of 'shadow' MCP servers, enforces authentication via OAuth 2.0, controls user and a - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. --- ## FSI201 — BELIEVE: The Impossible Migration That Transformed Australian Banking URL: https://aws-summit-2026-kb.pages.dev/sessions/FSI201 Level: intermediate Type: Breakout session Category: Financial Services Topics: Migration & Modernization; Resilience & Disaster Recovery; Industry Spotlight: Financial Services Commonwealth Bank migrated the world's largest SAP core banking deployment to AWS in 18 months: the system behind 40% of Australia's payments, 15 million customers, running 247. This isn't a lift-and-shift story, it's a reinvention of how the bank runs critical systems - from architecture and resilience engineering to replacing siloed operational teams with full-stack automation, and the cultural shift this required. Join us to hear how CBA's critical financial infrastructure was modernised with AWS, and what this unlocks for their AI-enabled future. If you're building foundations for regulated, mission-critical workloads, this is the session you don't want to miss. ### Playbook (editorial commentary) **The concept.** World's largest SAP core banking migration to AWS in 18 months. 40% of Australia's payments. 15M customers. 24×7. Full reinvention — not lift-and-shift. **Why it matters.** This is the proof that critical regulated workloads can be migrated. The case study unlocks board-level confidence at every other Australian bank. **The hard parts.** The cultural shift from siloed operational teams to full-stack automation. The org change is the harder problem; the tech is downstream. **Playbook moves.** (1) Plan the org redesign *before* the migration plan. (2) Invest in resilience engineering as a discipline, not a checkbox. (3) Treat 24×7 cutover as a multi-quarter program with explicit reversibility. **The surprise.** 18 months for a SAP core banking migration sounds impossible because it usually *is* impossible — except CBA also redesigned their team structure simultaneously. The migration was an *outcome* of the org change, not the cause of it. Most failed migrations try to lift-and-shift the org alongside the workloads. That doesn't work. --- ### Live monitored sources - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Meow Technologies launches the first agentic banking ...](http://thenextweb.com/news/meow-technologies-agentic-banking-ai-agents) — thenextweb.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [ServiceNow’s AI control tower offers hazy view of spend | CIO](https://www.cio.com/article/4169954/servicenows-ai-control-tower-offers-hazy-view-of-spend.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs - [How Agentic AI Will Reshape Payments - IMF](https://www.imf.org/en/publications/imf-notes/issues/2026/04/22/how-agentic-ai-will-reshape-payments-575560) — imf.org (2026-05-06): Solana Foundation President Lily Liu announced at Consensus Miami 2026 that Solana is building the payment rails for the 'AI machine economy,' emphasizing that traditional credit card networks are structurally unable to support the micropayments necessary for autonomous AI agent --- ## FSI207 — From enterprise data mesh to AI with Amazon SageMaker Unified Studio URL: https://aws-summit-2026-kb.pages.dev/sessions/FSI207 Level: intermediate Type: Breakout session Category: Financial Services Topics: Agentic AI; Machine Learning & SageMaker; Data Governance & Privacy; Voice & Conversational AI From enterprise data mesh to AI with Amazon SageMaker Unified StudioFinancial institutions are unlocking enormous value with AI agents — from personalised customer experiences to better risk decision making. But to deliver on that promise, agents need data they can find, understand, and trust. This session shows how a data mesh architecture on Amazon SageMaker Unified Studio builds that foundation: discoverable data across lines of business, business context that grounds agent responses in real meaning, quality signals that build confidence in every answer, and governed access that keeps you compliant by design. We cover domain ownership, multi-account strategies, data contracts, business glossaries, data quality, and cross-domain governance — and demonstrate how this foundation empowers agentic AI that delivers trusted, accurate results at enterprise scale. ### Playbook (editorial commentary) **The concept.** Data mesh on SageMaker Unified Studio as the foundation for trustable agentic AI. Domain ownership, data contracts, business glossaries, cross-domain governance. **Why it matters.** Agents need data they can find, understand, and trust. Data mesh provides discoverability + ownership in a way central lakes can't. **The hard parts.** Data mesh requires domain teams to own their data products. That's a culture and skills shift that platform teams can't impose. **Playbook moves.** (1) Pick 2–3 domains to pilot. (2) Get them publishing data products before extending mesh-wide. (3) Make data contracts the artefact, not the platform. **The surprise.** Most data mesh failures aren't technical — they're domain teams refusing to own their output. The CDO who can convince domain VPs to accept ownership is worth more than the platform itself. Hire for influence, not just engineering. --- ### Live monitored sources - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Gartner 2026 Confirms It: The Context Graph Is the Missing ...](https://thecontextgraph.co/memos/gartner-2026-ai-agents-decision-intelligence-sales) — thecontextgraph.co (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [The Context Graph Revolution: Why Enterprise AI ... - Medium](https://medium.com/@thanapong_18619/the-context-graph-revolution-why-enterprise-ai-needs-decision-lineage-c01d90fd1db4) — medium.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [Arango News & Press - ArangoDB](http://arango.ai/news-press) — arango.ai (2026-05-08): XMPro introduced the concept of 'operational memory' powered by decision traces, which capture the reasoning behind specific actions (including exceptions and human judgment) rather than just general rules. This is implemented via a decision trace layer in the orchestration path - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## ISV202 — Architecting for growth and resilience: Cell based design deep dive URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV202 Level: intermediate Type: Breakout session Category: ISV & Partners Topics: Security, Identity & Compliance; Resilience & Disaster Recovery; Retrieval Augmented Generation (RAG) As business demands evolve, architectural patterns must evolve too. SafetyCulture and Buildkite implemented cell-based architectures driven by distinct business objectivesscaling for hypergrowth and enhancing infrastructure resilience. SafetyCulture's expansion plans required proactive architectural evolution to unlock unlimited scaling capacity. Buildkite needed to meet stringent security isolation requirements while achieving scale through repeatable deployment units. This session shares real-world experiences as both companies designed and implemented cell-based architectures for their SaaS platforms. Discover how SafetyCulture identified bottlenecks, redesigned systems for isolation and resilience, and aligned technical capabilities with business growth targets. Learn how Buildkite leveraged cell-based design to achieve both scale and security isolation. Walk away with actionable patterns for building resilient, scalable architectures. ### Playbook (editorial commentary) **The concept.** Cell-based architecture: SafetyCulture for hypergrowth scaling, Buildkite for security isolation + scale. Repeatable deployment units. **Why it matters.** Cells solve scaling and blast-radius simultaneously. Avoid both single-cluster bottlenecks and single-point-of-failure risks. **The hard parts.** Migrating an existing system to cells is brutal. The pattern works best when designed in from day one. **Playbook moves.** (1) Identify cell boundaries early. Common axes: tenant size, region, security tier. Don't mix axes. (2) Migrate progressively; don't try to flip a whole estate. (3) Standardise the cell — heterogeneous cells defeat the purpose. **The surprise.** Cells aren't just an architecture pattern — they're an *organisational* pattern. The team structure must mirror the cells, or operational complexity explodes. Most cell migrations fail because the org didn't move with the architecture. Conway's Law applies in reverse here too. --- ### Live monitored sources - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## MAE202 — Seven's AWS Journey: Streaming Premium Content at the Speed of Innovation URL: https://aws-summit-2026-kb.pages.dev/sessions/MAE202 Level: intermediate Category: Media & Entertainment Topics: Streaming & Real-Time Data; Media & Entertainment; Retrieval Augmented Generation (RAG) Join Tim Sheridan, Director of Product & Technology at Seven West Media, as he shares how Seven is leveraging cloud and AI to maximise the return on their most valuable asset — premium live content. With marquee events like the AFL Grand Final and The Ashes cricket series, the stakes couldn't be higher: massive concurrent audiences, critical advertising revenue, and zero tolerance for failure. Tim shares how they leaned on AI-powered developer and business tools to accelerate delivery, de-risk high-profile events, and maximise the return on its premium content investments. Discover how Seven's team transformed their approach to innovation — using cloud-native architecture and AI to achieve speed to market, audience experience, and advertising revenue. ### Playbook (editorial commentary) **The concept.** AFL Grand Final, The Ashes — massive concurrent streaming, zero failure tolerance. AI-powered dev tools accelerated delivery. **Why it matters.** Premium live content is high-stakes; failure is brand-damaging. The patterns generalise to any high-stakes peak event. **The hard parts.** Burst capacity for concurrent peak (millions watching the same moment). Cost optimisation vs. headroom is a constant tension. **Playbook moves.** (1) Pre-event load tests at 1.5× expected peak. (2) Have rollback ready. (3) Keep AI tools out of the critical path until proven; use them for build-time, not run-time, on first deployments. **The surprise.** The hardest part of premium live streaming isn't streaming — it's the *ad insertion* at peak. Ad systems break first under burst load. Test those harder than the video pipeline; that's the actual fragile part. --- --- ## STP213 — AI-Powered Farming: How Halter's ML Models Transform Dairy Operations URL: https://aws-summit-2026-kb.pages.dev/sessions/STP213 Level: intermediate Type: Breakout session Category: Startups Topics: Machine Learning & SageMaker; Startups & Innovation; Retrieval Augmented Generation (RAG) New Zealand Unicorn agritech startup Halter is revolutionizing dairy farming with AI-powered smart collars that predict critical livestock events. Their machine learning models enable heat detection, calving prediction, pasture optimization, and animal behavior classification, processing data from thousands of GPS-enabled collars across remote farms. By leveraging AWS infrastructure, Halter's engineering team built scalable ML pipelines that help farmers make data-driven decisions, reduce labor costs, and improve animal welfare. Learn how Halter developed production ML models for agriculture, overcame challenges of training on livestock data, and their journey toward managed ML services. ### Playbook (editorial commentary) **The concept.** GPS collars on cattle, ML for heat detection, calving prediction, pasture optimisation, animal behaviour classification. Production ML across thousands of remote farms. **Why it matters.** Agritech proves AI's reach beyond knowledge work. Real-world physical optimisation at agricultural scale. **The hard parts.** Training on livestock data is hard — labels are sparse, ground truth requires veterinary observation, deployment loops are slow. **Playbook moves.** (1) Active learning to prioritise labelling effort. (2) Treat farmer feedback as a structured signal channel. (3) Build slow iteration cycles into the project plan; animals don't iterate fast. **The surprise.** The hardest engineering problem in agritech ML isn't the model — it's *connectivity*. Cellular dead zones in rural farms are everywhere. Edge inference + delayed sync is the operating reality. Most cloud-first ML architectures don't survive contact with rural Australia. --- ### Live monitored sources - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [Talent Harbor | Sales Recruitment as a Service (RaaS)](http://talentharbor.com/) — talentharbor.com (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [2026 - TechCrunch](https://techcrunch.com/2026/) — techcrunch.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [Meta bought some help in its quest for humanoid robots](https://www.businessinsider.com/meta-acquires-assured-robot-intelligence-humanoid-robotics-2026-5) — businessinsider.com (2026-05-02): Meta has acquired Assured Robot Intelligence, a startup specializing in AI models for robots, to advance its humanoid robot technology. - [Basata Raises $21M Series A to Expand AI Healthcare ...](https://www.citybiz.co/article/844227/basata-raises-21m-series-a-to-expand-ai-healthcare-operations-platform) — citybiz.co (2026-05-09): Phoenix-based healthcare AI startup Basata raised $21 million in Series A funding to expand its AI-driven healthcare operations automation tools. --- ## WPS301 — AWS for healthcare analytics: accelerating time to insights URL: https://aws-summit-2026-kb.pages.dev/sessions/WPS301 Level: advanced Type: Chalk talk Category: Public Sector Topics: Streaming & Real-Time Data; Analytics, Redshift & Generative BI; Industry Spotlight: Healthcare & Life Sciences In today's data-driven healthcare landscape, organisations must rapidly transform diverse data sources into actionable insights that improve patient outcomes and accelerate operational efficiency. This session showcases how AWS' integrated analytics capabilities can deliver unmatched price-performance for every analytics workload, from data processing and SQL analytics to streaming and business intelligence. Through real-world healthcare examples, learn how AWS' built-in governance and scalability enable organisations to build secure, efficient analytics pipelines that accelerate time-to-insight. Ideal for data practitioners, IT decision-makers, and executives evaluating enterprise analytics platforms to drive their data-driven transformation. ### Playbook (editorial commentary) **The concept.** Integrated analytics for healthcare data. Data processing → SQL → streaming → BI. Built-in governance and scalability. **Why it matters.** Healthcare data is heavy on governance. Integrated platforms reduce compliance overhead substantially. **The hard parts.** Healthcare data formats (HL7, FHIR) require domain expertise that generic data engineers lack. **Playbook moves.** (1) Hire or partner with healthcare data experts. (2) Don't outsource schema design. (3) Treat governance as a feature, not a cost. **The surprise.** Most "healthcare analytics" wins come from joining clinical and operational data — currently siloed in most hospitals. Bringing them together unlocks insights neither team has individually. The technical work is medium; the political work is hard. --- ### Live monitored sources - [Agentic AI Enterprise Use Cases — 30+ Real Deployments (2026)](https://www.ampcome.com/post/post-agentic-ai-enterprise-use-cases) — ampcome.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [Enterprise AI Agents Move From Pilot to Production: What 2026 ...](https://insights.reinventing.ai/articles/ai-agents-enterprise-production-2026-02-25) — insights.reinventing.ai (2026-05-05): Microsoft's 2026 Data Security Index reports that more than 80% of Fortune 500 companies are now running active AI agents in production, integrated across sales, finance, customer service, and security workflows. - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [80% of Fortune 500 use active AI Agents: Observability ...](https://www.microsoft.com/en-us/security/blog/2026/02/10/80-of-fortune-500-use-active-ai-agents-observability-governance-and-security-shape-the-new-frontier/) — microsoft.com (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [How enterprises are building AI agents in 2026 | Claude](https://claude.com/blog/how-enterprises-are-building-ai-agents-in-2026) — claude.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach --- ## DEV306 — Taming Legacy Code: Multi-Agent AI in Brownfield Systems URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV306 Level: advanced Category: Developer Tools Topics: Kiro & Spec-Driven Development; Voice & Conversational AI; Manufacturing & Industry 4.0; Code Generation & AI-Assisted Development; Agentic AI Real engineering happens in legacy codebases, not blank canvases. This session explores deploying multi-agent AI workflows using Kiro against brownfield production systems with tangled dependencies and accumulated technical debt. Learn how to orchestrate specialised agents for system mapping, dependency navigation, code generation, and validation within complex existing architectures. We'll examine practical strategies for providing sufficient context to agents, implementing guardrails to prevent regressions, and coordinating multiple agents toward shared goals. Walk away with actionable techniques for applying agentic AI to real-world codebases, understanding where automation delivers value and where human judgment remains irreplaceable. ### Playbook (editorial commentary) **The concept.** Multi-agent workflows on legacy codebases. Specialised agents for system mapping, dependency navigation, code generation, and validation. **Why it matters.** Brownfield is where real engineering happens. AI tools that only work on greenfield are useless for most enterprises. **The hard parts.** Providing sufficient context to agents in tangled codebases. Coordinating multiple agents toward shared goals without thrashing. **Playbook moves.** (1) Start with read-only mapping agents. (2) Add modification agents only after the map is verified. (3) Limit specialist count — coordination overhead beats specialisation gains beyond a small number. **The surprise.** The right number of agents is rarely the obvious one. Two specialists usually outperform five. Multi-agent designs that look organised in diagrams often thrash in production. Default to fewer until you can name the specific reason you need more. --- ### Live monitored sources - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## ISV211 — Scaling Conversation Intelligence with Agentic AI on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV211 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Security, Identity & Compliance; Databases & Aurora; Manufacturing & Industry 4.0; Analytics, Redshift & Generative BI; Agentic AI Businesses capture millions of conversations daily, sales calls, support interactions and compliance discussions, yet most of this intelligence remains locked away. Standard dashboards and predefined reports cannot address every customer's unique questions. Dubber, a world leader in conversation capture and intelligence, built Insight Agent on AWS, enabling users to ask bespoke, natural language questions across conversations and structured data to receive context-aware answers in seconds. Learn how Dubber innovated from static dashboards to surfacing business value moments, and now to agentic AI that compresses time to value, making conversation intelligence accessible, scalable and viable. ### Playbook (editorial commentary) **The concept.** Insight Agent on AWS. Natural language queries across captured conversations + structured data. Replaces static dashboards. **Why it matters.** Dashboard-centric BI can't answer custom questions. Agentic interfaces can. **The hard parts.** Conversation data is sensitive — compliance, privacy, retention rules. Access controls per conversation type matter. **Playbook moves.** (1) Define data access tiers (internal / customer / regulated). (2) Apply at retrieval time, not at agent layer. (3) Build retention-aware querying. **The surprise.** The biggest unlock from natural-language BI isn't asking *new* questions — it's the *speed* of asking questions. When questions are cheap, hypothesis-testing changes. Teams stop deferring to "I'll check the dashboard" and start exploring. That cultural shift is the actual ROI. --- ### Live monitored sources - [How Fortune 500 Companies Are Scaling Agentic AI to Production](https://ai2.work/blog/how-fortune-500-companies-are-scaling-agentic-ai-to-production) — ai2.work (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [Gartner 2026 Confirms It: The Context Graph Is the Missing ...](https://thecontextgraph.co/memos/gartner-2026-ai-agents-decision-intelligence-sales) — thecontextgraph.co (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi --- ## STP204 — How Heidi Health Fine-Tunes Speech-to-Text Models on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/STP204 Level: intermediate Category: Startups Topics: Voice & Conversational AI; Security, Identity & Compliance; Generative AI & Foundation Models; Machine Learning & SageMaker; Industry Spotlight: Healthcare & Life Sciences Join Heidi Health and AWS's Generative AI Innovation Center (GenAIIC) for a behind-the-scenes look at building and deploying custom speech-to-text AI for healthcare. Learn hard-won lessons and a practical blueprint: curating domain-specific training data, fine-tuning open-weight models, validating non-deterministic outputs at scale, and shipping to production with optimized inference. Both teams share how AWS services reduced infrastructure complexity, accelerated iteration cycles, and scaled custom models across diverse real-world use cases — all while maintaining security and cost efficiency. Ideal for ML engineers, data scientists, and technical leaders exploring fine-tuning and production ML on AWS. ### Playbook (editorial commentary) **The concept.** Fine-tuning open-weight speech-to-text models for healthcare. Domain-specific training data, validating non-deterministic outputs at scale, optimised inference. **Why it matters.** Generic STT misses medical terminology. Fine-tuning closes the gap; it's a textbook fine-tune use case. **The hard parts.** Validating non-deterministic STT output at scale is statistical, not anecdotal. You need a test set, not vibes. **Playbook moves.** (1) Build a labelled gold-standard test set per medical specialty. (2) Re-evaluate monthly. (3) Track word-error rate by accent, specialty, and recording quality separately. **The surprise.** The dominant accuracy issue in healthcare STT in Australia isn't medical jargon — it's *accents and code-switching*. Patient cohorts are linguistically diverse; clinicians switch registers. Train accordingly; English-only test sets miss most of the failure cases. --- ### Live monitored sources - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 - [How enterprises are building AI agents in 2026 | Claude](https://claude.com/blog/how-enterprises-are-building-ai-agents-in-2026) — claude.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [Agentic AI at Scale: The Rackspace Story](http://rackspace.com/resources/agentic-ai-scale-rackspace-story) — rackspace.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM - [Top Tech News Today, May 1, 2026 - Tech Startups](https://techstartups.com/2026/05/01/top-tech-news-today-may-1-2026/) — techstartups.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [AI Agent Identity and MCP: Authenticating Non-Human Identities](https://guptadeepak.com/ciam-compass/guides/ai-agent-identity-mcp) — guptadeepak.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho --- ## DEV307 — Active-Active Global Architecture with CloudFront and Route 53 URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV307 Level: advanced Category: Developer Tools Topics: Networking & Edge; Containers: EKS, ECS & Fargate In this lightning talk, we'll walk through a real-world architectural pattern used in production: combining AWS CloudFront with Route 53 latency-based routing to make your ECS-backed services truly global. Starting with the problem of slow response times for APAC users, we'll build up a practical active-active architecture step by step. You'll see how CloudFront sits in front of your regional ALBs, how WAF is woven into the design from the start rather than bolted on later, and why getting your domain configuration right — distinguishing between your ALB origin domain and your public-facing CloudFront alternate domain — is critical to making this pattern work correctly. ### Playbook (editorial commentary) **The concept.** CloudFront in front of regional ALBs. Route 53 latency-based routing. ECS-backed services made truly global. WAF designed in from the start. **Why it matters.** Global users demand sub-200ms responses. Single-region architectures don't deliver. **The hard parts.** Domain configuration. Distinguishing the ALB origin domain from the public-facing CloudFront alternate domain. Easy to misconfigure. **Playbook moves.** (1) Build the domain map explicitly — write it down. (2) Test failover before launch. (3) Use staging that mirrors prod's DNS structure exactly. **The surprise.** WAF retrofit costs are 2–3× WAF design-from-day-one costs. The "we'll add security later" pattern is more expensive than building it in. Don't defer. --- ### Live monitored sources - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Scaling Autonomous Agent Swarms with Distributed Task ...](https://martinuke0.github.io/posts/2026-03-31-scaling-autonomous-agent-swarms-with-distributed-task-orchestration-and-low-latency-communication-protocols/) — martinuke0.github.io (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. --- ## ISV209 — From dev tools to customer value: BGL's agentic AI journey URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV209 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Agentic AI; Security, Identity & Compliance; Generative AI & Foundation Models; Industry Spotlight: Financial Services Australian fintech leader BGL demonstrates their systematic approach to scaling agentic AI. Starting with Claude Code on Amazon Bedrock for developer productivity, BGL built OpsGorilla, a Slack-integrated agent serving 200+ employees. BGL extended AI agents to customer-facing innovation. BGL's Roni AI autonomously drafts annual compliance work for Self-Managed Superannuation Funds, reducing accountants' workload from days to hours. Built on Claude Agent SDK and Amazon Bedrock AgentCore, BGL's implementation showcases practical architectural patterns for organizations scaling AI agents from developer tools through enterprise operations to customer solutions. ### Playbook (editorial commentary) **The concept.** Three-stage agent rollout. Claude Code on Bedrock for devs → OpsGorilla (200+ employees, Slack-integrated) → Roni AI (customer-facing SMSF compliance, days → hours). **Why it matters.** A repeatable pattern: dev tools → internal ops → customer-facing. Each stage banks confidence for the next. **The hard parts.** Customer-facing agents have compliance and liability stakes that internal tools don't. The leap is real, not incremental. **Playbook moves.** (1) Bank confidence with internal tools first. (2) Customer-facing requires legal and compliance review as a gate. (3) Run dual-key approval for first 90 days of customer-facing. **The surprise.** Roni AI's "days to hours" for SMSF compliance is the kind of vertical-specific gain where fine-tuned, domain-aware approaches beat general agents by an order of magnitude. Generic agents wouldn't get there. Vertical specialisation beats horizontal scale here — and probably in most regulated workflows. --- ### Live monitored sources - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Edge Delta Makes All Telemetry Pipelines Data ...](http://prnewswire.com/news-releases/edge-delta-makes-all-telemetry-pipelines-data-throughput-limitless-and-free-for-all-customers-302736808.html) — prnewswire.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca --- ## STP209 — How Cartesian Turns AI Agents from SaaS Killer to SaaS Moat URL: https://aws-summit-2026-kb.pages.dev/sessions/STP209 Level: intermediate Category: Startups Topics: Agentic AI; Data Governance & Privacy The agents invasion into the software market is a fact of life now. Agents are changing how we are consuming software, services and information. But just like any technological inflection point, theres a redistribution of power with and SaaS platforms are struggling to find their centre of gravity in this new world. In this talk we will explore how, Cartesian is helping platforms lean in to their strategic assets like access to customers and privacy and find their moat in the agentic age by distributing and monetizing 3rd party agents. ### Playbook (editorial commentary) **The concept.** Agents are redistributing power. SaaS platforms can lean into customer access and privacy as moats. Distribute and monetise third-party agents. **Why it matters.** If your customer's agent shops competitors' APIs, your moat erodes overnight. Strategic question for every SaaS CEO. **The hard parts.** Re-architecting around agent-native workflows means your existing UI is no longer the differentiator. Some teams will resist that hard. **Playbook moves.** (1) Treat customer agents as a first-class API consumer. (2) Charge for them differently from human users. (3) Build agent governance — who can call what, when, with what audit. **The surprise.** The SaaS moat in the agentic era is *agent governance* — not features. Who decides which agents touch your customer's data, in what order, with what audit trail? That's not a feature you build; it's a position you claim. The first mover in each vertical will own it. --- ### Live monitored sources - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [The agent control plane becomes the new enterprise buying surface](https://www.linkedin.com/pulse/agent-control-plane-becomes-new-enterprise-buying-andrew-mcpherson-gcd4f) — linkedin.com (2026-05-07): At Cloud Next 2026, Google committed $750 million to a partner fund designed to accelerate the development of agentic AI builds, supporting partners like Accenture and KPMG in scaling AI agent deployment. - [Agentic Benchmarks 2026: Tool Use, Browsing, Computer Use | BenchLM.ai](https://benchlm.ai/agentic) — benchlm.ai (2026-05-12): BenchLM.ai updated its Agentic Benchmarks leaderboard on 2026-05-11. The update introduced two new benchmarks: 1) OpenHands Index, a holistic coding-agent benchmark covering issue resolution, frontend work, greenfield development, testing, and information gathering; and 2) SWE-At - [Google launches $750M partner fund at Cloud Next 2026 to ...](https://thenextweb.com/news/google-cloud-750m-partner-fund-agentic-ai) — thenextweb.com (2026-05-07): At Cloud Next 2026, Google committed $750 million to a partner fund designed to accelerate the development of agentic AI builds, supporting partners like Accenture and KPMG in scaling AI agent deployment. --- ## FSI204 — Agentic AI in Financial Services: Architectural Patterns That Work URL: https://aws-summit-2026-kb.pages.dev/sessions/FSI204 Level: intermediate Type: Breakout session Category: Financial Services Topics: AWS Well-Architected Framework; Voice & Conversational AI; Industry Spotlight: Financial Services; Security, Identity & Compliance; Agentic AI Getting agentic AI right in financial services means balancing innovation with the realities of compliance, risk, and auditability. This session cuts through the hype — exploring proven architectural patterns from reactive agents to multi-agent topologies, and how FSI organisations are using them to transform customer experience and automate operations. Leave with actionable guidance on building the business case, avoiding enterprise-scale pitfalls, and putting well-architected agents into production. ### Playbook (editorial commentary) **The concept.** Patterns from reactive agents to multi-agent topologies. Compliance, risk, auditability balanced with innovation. **Why it matters.** FSI has the highest scrutiny + the highest stakes. Patterns proven there generalise to other regulated industries. **The hard parts.** Multi-agent topologies multiply audit complexity. Each agent's decisions need separate provenance. **Playbook moves.** (1) Document agent boundaries explicitly. (2) Audit logs per agent, not just per system. (3) Make the human override path obvious in the architecture diagram. **The surprise.** The architectural pattern most FSI organisations underuse is the *human-in-the-loop circuit breaker* — an explicit point where the agent stops and waits for approval, not just review. That's the auditable artefact regulators care about. Review-after-the-fact looks the same on paper but isn't. --- ### Live monitored sources - [Agentic AI at Scale: The Rackspace Story](http://rackspace.com/resources/agentic-ai-scale-rackspace-story) — rackspace.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Enterprise AI Agents 2026: Mid-Year Report on What's Working](https://www.ampcome.com/post/enterprise-ai-agents-2026-mid-year-report) — ampcome.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## IND202 — How Zuru Uses AI to Analyze TikTok Trends for Rapid Content Creation URL: https://aws-summit-2026-kb.pages.dev/sessions/IND202 Level: intermediate Type: Breakout session Category: Other Topics: Media & Entertainment; Containers: EKS, ECS & Fargate; Generative AI & Foundation Models For modern consumer brands, winning means moving at the speed of culture. Zuru uses Amazon Bedrock and Twelve Labs to analyze up to 10,000 TikTok videos a day, rapidly identifying viral trends, emotional cues, and content patterns to create creator briefs in hours instead of weeks. Join this session to see how AWS gives Zuru a measurable edge, from 30 million organic views in seven days to 50x faster content creation, with industry peers now looking to replicate its speed-to-market advantage. ### Playbook (editorial commentary) **The concept.** 10,000 TikTok videos/day analysed via Bedrock + Twelve Labs. Viral trend identification, emotional cues, content patterns. Creator briefs in hours instead of weeks. 50× faster content creation. **Why it matters.** Speed-to-culture is the new speed-to-market for consumer brands. **The hard parts.** TikTok's API and content access are unstable. Compliance with platform terms can shift overnight. **Playbook moves.** (1) Build redundancy into video sources. Don't depend on one platform's API. (2) Cache aggressively where terms allow. (3) Treat platform-policy risk as a real budget line. **The surprise.** 50× faster content creation isn't only about AI speed — it's about removing the *approval bottleneck*. When briefs come from data instead of from creative directors, the chain shortens dramatically. The AI is the cover for an org-design change. --- ### Live monitored sources - [Mother Ventures Closes $10 Million Debut Fund to Back ...](https://www.rutlandherald.com/news/business/mother-ventures-closes-10-million-debut-fund-to-back-companies-serving-the-2-4-trillion/article_f6a35f35-9da2-550a-b896-fa927824bf32.html) — rutlandherald.com (2026-05-09): Mother Ventures announced the close of its first $10 million early-stage venture capital fund (Fund I), targeting companies where mothers are the primary consumer. - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [AI News May 2026: Monthly Digest - toolscompare.ai](https://toolscompare.ai/news/may-2026) — toolscompare.ai (2026-05-11): China's Ministry of Industry and Information Technology (MIIT) released a national standard for AI terminal intelligence grading, which includes the implementation of the smart cockpit level 3 standard. - [Morgan Stanley warns an AI breakthrough Is coming in 2026](https://finance.yahoo.com/news/morgan-stanley-warns-ai-breakthrough-072000084.html) — finance.yahoo.com (2026-05-09): Mother Ventures announced the close of its first $10 million early-stage venture capital fund (Fund I), targeting companies where mothers are the primary consumer. - [Tekst Raises $13.5 Million Series A | The SaaS News](http://thesaasnews.com/news/tekst-raises-13-5-million-series-a) — thesaasnews.com (2026-05-11): China's Ministry of Industry and Information Technology (MIIT) released a national standard for AI terminal intelligence grading, which includes the implementation of the smart cockpit level 3 standard. --- ## IND204 — How Transurban Transformed Customer Experience with AI Agents on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/IND204 Level: intermediate Type: Breakout session Category: Other Topics: Agentic AI; Voice & Conversational AI; Security, Identity & Compliance; Manufacturing & Industry 4.0 Every month, Transurban handles 5.5 million customer interactions across its Linkt brand — and is reimagining every one of them. Built on Amazon Connect, Transurban's AI-powered customer service platform has evolved from simple chatbots to multi-turn conversational AI and personalised experiences that are already lifting bot containment and freeing agents for higher-value work. Join this session to hear how Transurban is aligning people, process, and AI to transform customer service, what's coming next with Amazon Connect Cases and Email, and the hard-won lessons from scaling AI in a complex enterprise. ### Playbook (editorial commentary) **The concept.** 5.5M monthly customer interactions through Linkt. Multi-turn conversational AI on Amazon Connect. Bot containment up; agents freed for higher-value work. **Why it matters.** At Transurban scale, even single-percent containment improvements compound to material savings. **The hard parts.** Toll-related conversations are emotional (disputed charges, fines). Botched bots make things measurably worse — CSAT can fall while containment rises. **Playbook moves.** (1) Tier conversations by emotion. Route emotional ones to humans early. (2) Bots handle transactional flows. (3) Track CSAT *per channel*, not aggregate. **The surprise.** The metric to watch isn't bot containment — it's CSAT *per channel*. Containment can rise while CSAT falls; they're not the same goal. Most teams optimise the easier metric and damage the harder one. --- ### Live monitored sources - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Google launches $750M partner fund at Cloud Next 2026 to ...](https://thenextweb.com/news/google-cloud-750m-partner-fund-agentic-ai) — thenextweb.com (2026-05-07): At Cloud Next 2026, Google committed $750 million to a partner fund designed to accelerate the development of agentic AI builds, supporting partners like Accenture and KPMG in scaling AI agent deployment. --- ## SMB201 — The AI-Driven Development Lifecycle: How Skyjed Shipped in 48 hours URL: https://aws-summit-2026-kb.pages.dev/sessions/SMB201 Level: intermediate Type: Breakout session Category: Small & Medium Business Topics: Kiro & Spec-Driven Development Traditional AI-assisted development treats AI as a tool. The AI-Driven Development Lifecycle (AI-DLC) treats it as a central collaborator throughout the development process. In this session, join Skyjed CTO Stephen Brown, to share how his team applied AI-DLC to compress a 4-6 week implementation into 48 hours. Discover how AI-DLC's three phases — Inception, Construction, and Operations — reimagines the software lifecycle through AI-powered execution with human oversight and dynamic team collaboration. Drawing on Skyjed's firsthand experience on integrating AI-assisted development into their existing software development lifecycle, this talk explores how team collaboration, Kiro-powered workflows, and enterprise-grade code development practices were enhanced rather than replaced. You'll walk away with practical steps to achieve 1520x accelerationwithout sacrificing quality, governance, or control. ### Playbook (editorial commentary) **The concept.** AI-DLC's three phases: Inception, Construction, Operations. Kiro-powered workflows. 15–20× acceleration claimed against pre-AI baseline. **Why it matters.** This roughly matches Xero's claim from Day 1. The pattern is now validated by multiple companies. **The hard parts.** 15–20× is achievable on right-fit work. Wrong-fit work doesn't compress meaningfully. **Playbook moves.** (1) Identify work classes that suit AI-DLC: well-spec'd features, brownfield with clear context. (2) Avoid: greenfield with uncertain requirements. (3) Track actual compression ratio per project type. **The surprise.** AI-DLC's three-phase split (Inception / Construction / Operations) is the tell that the methodology has matured. Earlier "AI coding" framings collapsed everything into "code faster." This separates concerns properly. The phase boundaries are where the discipline lives. --- ### Live monitored sources - [GitHub Copilot in Visual Studio Code, April releases](http://github.blog/changelog/2026-05-06-github-copilot-in-visual-studio-code-april-releases) — github.blog (2026-05-11): Devin introduced an update to its 'Auto-fix with Devin' feature on pull requests, which now includes failing CI check names in the prompt alongside review findings to provide more context for resolving issues. - [See what’s new with GitHub Copilot](https://github.com/features/copilot/whats-new) — github.com (2026-05-05): Cursor released new Enterprise admin controls providing granular model access (allow/block lists at the provider and model level), soft spend limits with automated alerts at 50%, 80%, and 100% of the limit, and enhanced usage analytics that allow admins to filter consumption by s --- ## TNC301 — Using Tools and Agents in Generative AI applications URL: https://aws-summit-2026-kb.pages.dev/sessions/TNC301 Level: advanced Type: Lightning talk Category: Other Topics: Agentic AI; Model Context Protocol (MCP); Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) Join us for an engaging session on AI Agents and Tools in AWS, where well explore how to build intelligent, autonomous systems using Amazon Bedrock and open-source frameworks. Learn about function calling, ReAct patterns, and AWSs comprehensive agent platforms. Well dive into practical demonstrations using Strands and CrewAI, and discover how to leverage protocols like MCP and A2A for seamless tool integration and agent collaboration. Perfect for developers looking to create production-ready AI solutions. ### Playbook (editorial commentary) **The concept.** Function calling, ReAct patterns, Strands, CrewAI, MCP, A2A protocols. Foundational concepts. **Why it matters.** Pattern proliferation is confusing. Most teams pick one and over-rotate without seeing alternatives. **The hard parts.** Choosing among framework options. Each has its own assumptions, limitations, ecosystem. **Playbook moves.** (1) Pick one orchestration framework and master it before evaluating alternatives. (2) Pattern hopping costs months. (3) Document why you chose what you chose; future you will need that. **The surprise.** ReAct (reasoning + acting) is the boring default that beats fancy multi-agent topologies for many tasks. Don't overcomplicate when ReAct will do. Most "we need multi-agent" assertions don't survive a careful look at whether ReAct would suffice. --- ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [GitHub - agentgateway/agentgateway: Next Generation Agentic Proxy for AI Agents and MCP servers · GitHub](https://github.com/agentgateway/agentgateway) — github.com (2026-05-07): Agentgateway released version v1.2.0-alpha.1, continuing the development of its open-source AI-native proxy for agent-to-agent and agent-to-tool communication. The project maintains approximately 2.6k GitHub stars and is part of the Linux Foundation. --- ## WPS204 — Safe Transport Victoria's Migration to AWS Cloud URL: https://aws-summit-2026-kb.pages.dev/sessions/WPS204 Level: intermediate Type: Breakout session Category: Public Sector Topics: Migration & Modernization; Resilience & Disaster Recovery Join us for an in-depth case study on Safe Transport Victoria's successful use of Cloud to modernise, streamline and save costs while moving from on-premises infrastructure to AWS Cloud. This session will demonstrate how a small Regulator has added resilience to their safety outcomes, including to those Victorians with accessibility and mobility needs achieved modernization while maintaining service continuity and reliability. ### Playbook (editorial commentary) **The concept.** Small regulator migrating from on-prem to AWS. Resilience, accessibility, mobility-needs continuity for vulnerable Victorians. **Why it matters.** Government migrations are usually told from large-agency perspectives. Small-agency stories are more relevant for most. **The hard parts.** Small teams, fixed budgets, public scrutiny, accessibility requirements. The constraints are real. **Playbook moves.** (1) Modernise *while* migrating, not after. (2) The "lift then improve" path stalls in government environments. (3) Bake accessibility in from the architecture, not as a UX layer. **The surprise.** Cloud platforms now provide accessibility primitives (live captioning, screen-reader-friendly APIs, language translation) that on-prem rarely had. The migration unlocks accessibility, not just scale. Frame the business case accordingly when selling to government boards. --- ### Live monitored sources - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [Rate Limiting Controls | GitHub Agentic Workflows](https://github.github.com/gh-aw/reference/rate-limiting-controls/) — github.github.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits - [empathic.dev](http://empathic.dev/) — empathic.dev (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [AI Agent Token Budget Enforcement [2026]](https://www.waxell.ai/blog/ai-agent-token-budget-enforcement) — axell.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios --- ## DEV203 — Decisions Over Diagrams: How Bell Financial Group Architects on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV203 Level: intermediate Category: Developer Tools Topics: Containers: EKS, ECS & Fargate; Industry Spotlight: Financial Services; Security, Identity & Compliance; Databases & Aurora; Compute: EC2, Graviton & Nitro; Serverless: Lambda & Step Functions Architecture diagrams show what you built. They don't explain why. At Bell Financial Group, every major technology choice — from landing zone design to compute platform to database engine — is captured in an Architecture Decision Document that forces honest evaluation of trade-offs. In this talk, the Head of Engineering at Bell Financial Group walks through the real decisions behind their AWS platform: why ECS Fargate beat EKS, when DynamoDB wins over relational databases, why the entire infrastructure is written in TypeScript CDK, and the deliberate constraints they place on Lambda usage. No slides full of boxes and arrows — just the reasoning, the trade-offs, and the lessons learned building a regulated financial services platform on AWS. ### Playbook (editorial commentary) **The concept.** Every architecture decision captured in an Architecture Decision Document. ECS Fargate over EKS. DynamoDB over relational where appropriate. TypeScript CDK everywhere. Deliberate constraints on Lambda usage. **Why it matters.** Diagrams show what you built. ADRs explain *why*. Without ADRs, every new hire re-litigates settled decisions. **The hard parts.** ADRs require discipline that most teams don't have. They feel like overhead until you need them. **Playbook moves.** (1) Start now. Backfill ADRs for the 10 biggest decisions of the last year. (2) Make new ADRs mandatory for major changes. (3) Review ADRs annually — circumstances change. **The surprise.** ADRs are the cheapest insurance against tech-debt litigation. When the next CTO asks "why did we pick X?", a good ADR settles it in 5 minutes. Without one, the question reopens forever — and your team burns hours each time. Start the discipline this quarter. --- ### Live monitored sources - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Stripe introduces Link, a digital wallet that autonomous AI ...](https://postofday.com/2026/05/01/stripe-introduces-link-a-digital-wallet-that-autonomous-ai-agents-can-use-too/) — postofday.com (2026-05-05): Reports indicate that MoonPay and the Agentic Experience Protocol (AXP) have launched functional agent payment infrastructure (April-May 2026), with AXP extending the Universal Commerce Protocol (UCP) to support unified agentic commerce experiences and rich product data. - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Everything we announced at Sessions 2026 - Stripe](https://stripe.com/blog/everything-we-announced-at-sessions-2026) — stripe.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. --- ## ISV103 — Working With AI: Lessons They Don't Put in the Demo URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV103 Level: foundational Type: Lightning talk Category: ISV & Partners Topics: Security, Identity & Compliance We're past the hype phase and into the "figuring it out" phase. AI is reshaping how we build, hire, and collaborate, but the reality on the ground doesn't always match the pitch. In this session, Pushpay shares what they are actually seeing across teams and organizations: the common traps, the unexpected side effects, and the coping strategies that are actually working. Come ready for an honest, practical conversation about thriving with AI rather than just surviving it. ### Playbook (editorial commentary) **The concept.** Past hype phase, into "figuring it out" phase. Honest read on AI's effects on building, hiring, collaborating. Common traps, side effects, coping strategies. **Why it matters.** Hype-cycle consensus is wrong. Ground truth from teams in the trenches matters more than vendor decks. **The hard parts.** AI's side effects show up months later: skill atrophy in juniors, debugging gaps, code-review fatigue. **Playbook moves.** (1) Track AI usage's *second-order effects*. (2) Survey juniors specifically — they're affected differently from seniors. (3) Make first-principles debugging part of the team's training cadence. **The surprise.** One of the most common unexpected AI side effects is junior engineers losing the ability to debug systems they didn't write. AI gets them through; the deeper understanding doesn't form. Plan for it explicitly. The juniors you hire today will be your seniors in 3 years; what they're learning now matters. --- ### Live monitored sources - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg --- ## STP301 — AI-Native Remediation with Pleri: Your Security Engineer That Ships URL: https://aws-summit-2026-kb.pages.dev/sessions/STP301 Level: advanced Category: Startups Topics: Media & Entertainment; Security, Identity & Compliance; Generative AI & Foundation Models; Manufacturing & Industry 4.0 Most security tools find the problem and hand it to a human. Plerion closes the loop. In this talk, we'll show how Pleri, our AI security engineer powered by Amazon Bedrock, takes a critical cloud risk from detection to remediation without the alert-ticket-backlog cycle. Watch a top risk get prioritized, a ticket filed, a PR opened, and code-level remediation land in your environment. Re-define what it means to have an AI teammate that does the work, not just alerts and reporting. ### Playbook (editorial commentary) **The concept.** AI security agent that goes from detection to remediation. Prioritisation → ticket → PR → code-level fix in your environment. Closes the alert-to-action loop. **Why it matters.** Most security tools find problems and hand them off. The handoff is where things rot. Closing the loop is the unlock. **The hard parts.** Auto-remediation requires write access to production code. That's a trust gate, not a tech gate. **Playbook moves.** (1) Stage trust. Start with PR proposals, not auto-merge. (2) Track agent suggestion quality before granting more authority. (3) Keep humans in the loop on critical-path changes indefinitely. **The surprise.** The risk most teams worry about (the AI ships a bad fix) is rarer than the risk they ignore (humans ignore PR queues that pile up). Auto-remediation works because it forces the queue to drain. Stalled queues are the bigger threat than occasional bad fixes. --- ### Live monitored sources - [Prompt Injection in Production Agents: 2026 Taxonomy](https://www.digitalapplied.com/blog/prompt-injection-production-agents-2026-taxonomy) — digitalapplied.com (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [When prompts become shells: RCE vulnerabilities in AI agent ...](https://www.microsoft.com/en-us/security/blog/2026/05/07/prompts-become-shells-rce-vulnerabilities-ai-agent-frameworks) — microsoft.com (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [Prompt Injection Attack to Tool Selection in LLM Agents](https://www.ndss-symposium.org/wp-content/uploads/2026-s675-paper.pdf) — ndss-symposium.org (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [[2602.21012] International AI Safety Report 2026 - arXiv.org](https://arxiv.org/abs/2602.21012) — arxiv.org (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf --- ## IDE101 — From principles to practice: Scaling AI responsibly URL: https://aws-summit-2026-kb.pages.dev/sessions/IDE101 Level: foundational Type: Breakout session Category: Diversity, Equity & Inclusion Topics: Data Governance & Privacy; Security, Identity & Compliance; Manufacturing & Industry 4.0; Retrieval Augmented Generation (RAG) Building AI applications that customers trust requires more than technical excellenceit demands a deliberate approach to managing risk across every stage of the AI lifecycle. As organizations scale their AI initiatives, the challenge of balancing innovation speed with responsible AI practices across dimensions like privacy, security, fairness, safety, and explainability becomes increasingly critical. Join our panelists for a 30-minute discussion where they will explore: Practical approaches to embedding responsible AI principles into AI application development without slowing down innovation, key considerations across privacy, security, fairness, safety, and explainability that organizations should prioritize, lessons learned from building AI applications that earn and maintain customer trust, and strategies for navigating the evolving responsible AI landscape and managing risk at scale. Whether you are a technical leader building AI solutions, a business decision-maker shaping your organization's AI strategy, or a practitioner looking to deepen your understanding of responsible AI, this session will provide actionable insights to help you build AI applications that are not only innovative but also trustworthy. ### Playbook (editorial commentary) **The concept.** Embedding responsible AI across privacy, security, fairness, safety, and explainability without slowing innovation. Lessons from building trustworthy AI applications at scale. **Why it matters.** Trust is durable; speed without trust collapses. Responsible AI is a competitive moat once you can deliver it. **The hard parts.** "Responsible AI" gets framed as a brake. The good versions are accelerators (clearer specs, better tests, fewer rollbacks). **Playbook moves.** (1) Define responsible AI as a quality bar, not a separate process. (2) Bake it into release criteria. (3) Make explainability a first-class requirement, not a nice-to-have. **The surprise.** The orgs that deploy responsible AI fastest are the ones that already had strong product safety review processes — they're extending an existing muscle. Orgs without that muscle have to build it first; the schedule is real and underestimated. Plan for 6–12 months of muscle-building if you're starting cold. --- ### Live monitored sources - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [Fetched web page](https://mem0.ai/blog/6-techniques-to-cut-ai-agent-memory-cost-beyond-basic-retrieval) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin - [The 2026 Token Optimization Playbook: Cut AI Agent Memory Costs 3–4X](https://mem0.ai/blog/the-2026-token-optimization-playbook-cut-ai-agent-memory-costs-3–4x) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. --- ## DEV308 — AI Blast-Radius Reviews for AWS Changes Using Amazon Bedrock URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV308 Level: advanced Category: Developer Tools Topics: Security, Identity & Compliance; Generative AI & Foundation Models Cloud infrastructure changes carry hidden risks that manual reviews often miss. This session demonstrates how to build a pre-deployment blast-radius reviewer using Amazon Bedrock that analyzes infrastructure-as-code diffs, IAM policy changes, and deployment metadata to produce structured risk assessments in seconds. Attendees will learn how to design grounded AI workflows that identify affected services, security gaps, cost implications, and rollback considerations, and how to apply Bedrock guardrails for consistent, safe outputs. The session includes a live demonstration and covers practical patterns for integrating AI-generated assessments into existing engineering approval and governance processes without sacrificing speed or reliability. ### Playbook (editorial commentary) **The concept.** Pre-deployment AI review of IaC diffs, IAM policy changes, deployment metadata. Structured risk assessments in seconds. Bedrock guardrails for safe outputs. **Why it matters.** Manual change reviews miss subtleties. AI catches what reviewers gloss over — especially in IAM and network diffs. **The hard parts.** AI can hallucinate risk assessments. False alarms erode trust fast. **Playbook moves.** (1) Validate AI suggestions against historical incidents. (2) Tune prompts on your environment, not generic templates. (3) Treat AI assessments as input to review, not the review itself. **The surprise.** The best AI-generated change review isn't one that lists risks — it's one that lists *what would need to be true* for the change to be safe. That framing forces the human to verify those assumptions, not rubber-stamp. Reframe your prompt accordingly. --- ### Live monitored sources - [Token Security Introduces Intent-Based Security for AI Agents](http://finance.yahoo.com/news/token-security-introduces-intent-based-130200458.html) — finance.yahoo.com (2026-05-08): New implementation patterns for AI agent identity (updated May 6, 2026) highlight the convergence of the Model Context Protocol (MCP) for agent-server handshakes and OAuth 2.1 with Dynamic Client Registration (DCR) for runtime credential issuance. A key pattern is the use of 'dis - [aws.amazon.com](https://aws.amazon.com/about-aws/whats-new/2026/04/bedrock-openai-models-codex-managed-agents/) — aws.amazon.com (2026-04-29): Amazon Bedrock (AWS) now offers OpenAI models, Codex, and Managed Agents (Limited Preview) — announced 2026-04-28. What changed: OpenAI models and Managed Agents are available inside AWS Bedrock limited preview, letting AWS customers run OpenAI models and managed agent capabiliti - [GitHub - microsoft/autogen: A programming framework for ...](https://github.com/microsoft/autogen) — github.com (2026-05-07): CrewAI released pre-release version 1.14.5a3 on 2026-05-06. Key changes include: - Refactored the CLI into a standalone `crewai-cli` package. - Fixed the status endpoint path from `/{kickoff_id}/status` to `/status/{kickoff_id}`. - Updated the `gitpython` dependency to version >= - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes - [Releases · crewAIInc/crewAI · GitHub](https://github.com/crewAIInc/crewAI/releases) — github.com (2026-05-12): CrewAI (crewAIInc/crewAI) published release 1.14.5a4 (pre-release) on 2026-05-08. Highlights: updated LLM listings and dependency adjustments (moved textual dependency into crewai-cli and added certifi), several bug fixes and changelog/documentation updates. Migration implication --- ## ISV102 — From documents to voice - building AI products on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV102 Level: foundational Type: Lightning talk Category: ISV & Partners Topics: Containers: EKS, ECS & Fargate; Industry Spotlight: Financial Services; Voice & Conversational AI; Generative AI & Foundation Models; Agentic AI; Machine Learning & SageMaker; Retrieval Augmented Generation (RAG) How Affinda leverages Amazon Bedrock (Claude), SageMaker, EKS & CloudFormation to deliver intelligent document processing at enterprise scale, cutting setup time and costs by 90% with 95%+ accuracy. This session will demonstrate how Affinda powers real-world AI product development from Affinda's Intelligfent Document Processing platform to Pathfindr's (acquired by Affinda) custom AI agents. The session will showcase the complete journey of building Honey Insurance's voice agent - Australia's first voice agent in financial services, and how the Affinda-AWS partnership enables rapid AI product development for Enterprises. ### Playbook (editorial commentary) **The concept.** Bedrock + SageMaker + EKS + CloudFormation for intelligent document processing. 90% setup time/cost reduction, 95%+ accuracy. Honey Insurance voice agent — Australia's first in financial services. **Why it matters.** Document processing is one of the most universal enterprise needs. Voice agents are crossing into regulated industries. **The hard parts.** 95% accuracy still leaves 5% errors. In finance, those errors matter materially. **Playbook moves.** (1) Tier documents by risk. Auto-process low-risk; human-in-the-loop for high-stakes (legal, financial, medical). (2) Track accuracy per document type, not aggregate. (3) Plan an error-handling UX from day one. **The surprise.** Voice agents in finance face an unexpected hurdle — they must *legally* identify themselves as AI in some jurisdictions. Design for the disclosure from day one; retrofitting it sounds easy but breaks the conversational flow. --- ### Live monitored sources - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Everything we announced at Sessions 2026 - Stripe](https://stripe.com/blog/everything-we-announced-at-sessions-2026) — stripe.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. --- ## STP212 — How Apate AI uses Amazon Bedrock and voice AI to catch scammers URL: https://aws-summit-2026-kb.pages.dev/sessions/STP212 Level: intermediate Category: Startups Topics: Streaming & Real-Time Data; Voice & Conversational AI; Generative AI & Foundation Models; Agentic AI; Machine Learning & SageMaker Scams are a global epidemic costing businesses and consumers trillions. Apate AI turns the tables on fraudsters by deploying lifelike conversational AI agents, powered by Amazon Bedrock and speech models on Amazon SageMaker bidirectional streaming, that engage scammers in real time to detect, divert, disrupt, and decode their tactics. In this session, learn how Apate AI converts every scam interaction into actionable intelligence and how to build your own voice AI agents on AWS. ### Playbook (editorial commentary) **The concept.** Lifelike conversational AI agents that engage scammers in real-time on Bedrock + SageMaker bidirectional streaming. Detect, divert, disrupt, decode tactics. Convert each interaction into intelligence. **Why it matters.** Defensive AI is a growing category. Counter-deception is one of the fascinating use cases of the year. **The hard parts.** Scammers adapt fast. Models must evolve continuously. **Playbook moves.** (1) Treat the scam corpus as adversarial training data. (2) Update continuously — weekly cycles, not quarterly. (3) Capture every interaction's metadata for downstream intel. **The surprise.** The intelligence value of engaging scammers (their tactics, scripts, escalation patterns) is *bigger* than the disruption value. The product underneath Apate is intelligence-as-a-service to law enforcement, with disruption as the lead-in. Counter-fraud AI is going to look more like an intelligence operation than a defence tool. --- ### Live monitored sources - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## IDE102 — Power of Possibility: Leading Through Innovation and Connection URL: https://aws-summit-2026-kb.pages.dev/sessions/IDE102 Level: foundational Type: Workshop Category: Diversity, Equity & Inclusion Topics: Voice & Conversational AI; Security, Identity & Compliance; Mobile & Cross-Platform Development; Networking & Edge; Retrieval Augmented Generation (RAG) As AI reshapes every industry, professionals across all roles and backgrounds are navigating an unprecedented pace of changebringing new opportunities but also rising burnout, blurred boundaries, and pressure to continuously adapt. This moment of disruption presents a powerful opportunity to not only deliver innovation but to redesign how we lead, build culture, and sustain meaningful careers in more equitable ways. Join accomplished AWS leaders and peers for this immersive session that combines strategic leadership frameworks, emotional intelligence, and interactive roundtable discussion to accelerate your impact in tech. Together, we will explore practical strategies for claiming visible technical leadership, activating professional networks, setting sustainable boundaries in hybrid work, and championing responsible AI adoption without amplifying existing inequalities. Participants will share lived experiences, tactics that worked and lessons learned, and build meaningful connections through guided speed networking in a collaborative, supportive environment. This session empowers professionals from all backgrounds, with particular focus on amplifying diverse voices and fostering inclusive innovation. Leave with actionable strategies to strengthen your leadership presence, leverage emotional intelligence as a career accelerator, build psychologically safe and inclusive team environments, and navigate the challenges shaping your future in tech. ### Playbook (editorial commentary) **The concept.** Strategic leadership frameworks + emotional intelligence + roundtable discussion. Sustainable careers, equitable adoption, psychological safety. **Why it matters.** AI-driven change creates burnout. Leaders need new frameworks for sustaining team performance through extended disruption. **The hard parts.** Inclusion + speed get framed as opposites. Good leaders make them complementary, not competing. **Playbook moves.** (1) Audit team well-being explicitly and quantitatively. (2) Make sustainability a measurable leadership KPI. (3) Be deliberate about which teams you *don't* push to AI-native — adoption inequity compounds. **The surprise.** AI's biggest org-design risk is amplifying existing inequalities. The engineers already best at "managing up" benefit most from AI assistance; the ones who could most benefit may be the slowest to adopt. Plan for equity in adoption explicitly — it doesn't happen on its own. --- ### Live monitored sources - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Fetched web page](https://mem0.ai/blog/6-techniques-to-cut-ai-agent-memory-cost-beyond-basic-retrieval) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin - [The 2026 Token Optimization Playbook: Cut AI Agent Memory Costs 3–4X](https://mem0.ai/blog/the-2026-token-optimization-playbook-cut-ai-agent-memory-costs-3–4x) — mem0.ai (2026-05-08): Mem0 released technical guides on optimizing AI agent memory costs to reduce the 'token tax.' Key strategies include moving from naive injection to retrieval-based architectures (reducing prompt tokens by ~72%), implementing token budgeting, hierarchical summarization, and 'Ebbin - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv --- ## IND301 — Stockland Empowers People with a GenAI Assistant Built on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/IND301 Level: advanced Type: Breakout session Category: Other Topics: Agentic AI; Security, Identity & Compliance; Generative AI & Foundation Models Discover how Stockland, one of Australia's largest diversified property groups, built an intelligent AI assistant on AWS that puts enterprise knowledge at everyone's fingertips. This session explores a multi-agent system powered by Amazon Bedrock & Strands SDK and embedded in Microsoft Teams, that enables teams to streamline vendor management, automate routine tasks and accelerate decision-making. Learn how Stockland built intelligent workflows that transform procurement data into actionable insights, reducing manual effort while improving accuracy and compliance. ### Playbook (editorial commentary) **The concept.** Multi-agent system on Bedrock + Strands. Embedded in Microsoft Teams for vendor management, routine tasks, decision-making. Procurement data → actionable insights. **Why it matters.** Property is logistics-heavy. Agents inside Teams put intelligence where work actually happens. **The hard parts.** Teams bot integrations have UX constraints. Agents must be glanceable, not deep-dive. **Playbook moves.** (1) Optimise for the 30-second interaction. (2) Anything longer should kick to a different surface. (3) Make handoffs to human owners explicit and easy. **The surprise.** The biggest engagement driver in chat-based agents isn't capability — it's *response time*. Sub-second feels magical; 5+ seconds feels broken. Latency budget matters more than feature count for chat-embedded agents. Optimise inference time before adding tools. --- ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [The Context Graph Revolution: Why Enterprise AI ... - Medium](https://medium.com/@thanapong_18619/the-context-graph-revolution-why-enterprise-ai-needs-decision-lineage-c01d90fd1db4) — medium.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. --- ## INO203 — Behind the curtain: How Amazons AI innovations are powered by AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/INO203 Level: intermediate Type: Breakout session Category: Other Topics: Mobile & Cross-Platform Development; Media & Entertainment; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) Discover how Amazon leaders across Zoox, Prime Video, and Amazon Stores are leveraging AI to power their next-generation innovations with AWS and create better experiences for customers. Through three key stories discover how customer needs sparked transformative experiences: fan feedback revolutionising sports broadcasting with AI-powered highlights, shopping patterns evolving Amazons mobile commerce platform, and robotics and automation delivering advances in supply chain optimisation and fulfillment. Learn the best practices Amazon has applied using AWS that can help scale innovation in your organisation. ### Playbook (editorial commentary) **The concept.** Inside view of Zoox, Prime Video, and Amazon Stores AI. Sports highlights, mobile commerce evolution, robotics fulfillment. **Why it matters.** Amazon's internal use cases are the canary for AWS feature direction. **The hard parts.** Amazon's scale is unrepresentative for most companies. Patterns may not translate down without modification. **Playbook moves.** (1) Watch the patterns, not the scale. (2) The engineering principles translate; the volume doesn't. (3) Note which AWS services Amazon retail uses — those usually become productised within 6–12 months. **The surprise.** Most "Amazon does X" announcements are AWS feature roadmap leaks in disguise. If Amazon retail is doing it on AWS, AWS will productise the underlying capability for everyone within a year. Track these to anticipate platform direction. --- ### Live monitored sources - [L'Oréal Invests in Future of AI-Powered Commerce and the ...](http://prnewswire.com/apac/news-releases/loreal-invests-in-future-of-ai-powered-commerce-and-the-creator-economy-with-tech-startups-from-south-asia-pacific-middle-east-and-north-africa-302769056.html) — prnewswire.com (2026-05-12): Google streamed the Android Show on May 12, 2026, providing previews of Android 17, updates to Gemini's agentic capabilities, and details regarding the new Aluminium OS. - [Oracle Unveils AI Database Agentic Innovations for Business Data](https://www.oracle.com/news/announcement/oracle-unveils-ai-database-agentic-innovations-for-business-data-2026-03-24/) — oracle.com (2026-05-11): Understanding Data published a detailed blueprint for an 'Event Sourcing for Agents' storage pattern, describing a log-based architecture that stores agent state as an append-only sequence of events to enable deterministic replay, time-travel debugging, and audit trails for produ - [2026 - TechCrunch](https://techcrunch.com/2026/) — techcrunch.com (2026-05-02): KKR & Co. launched Helix Digital Infrastructure, a $10 billion company led by former AWS CEO Adam Selipsky, focused on building AI data centers and power infrastructure. - [Meta bought some help in its quest for humanoid robots](https://www.businessinsider.com/meta-acquires-assured-robot-intelligence-humanoid-robotics-2026-5) — businessinsider.com (2026-05-02): Meta has acquired Assured Robot Intelligence, a startup specializing in AI models for robots, to advance its humanoid robot technology. - [lasvegassun.com](https://lasvegassun.com/news/2026/apr/21/ai-startup-anthropic-commits-100-billion-to-amazon/) — lasvegassun.com (2026-04-22): AI startup Anthropic has committed $100 billion to Amazon's AWS over the next 10 years. --- ## ISV204 — AWS Networking Fundamentals: Connect, secure and scale URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV204 Level: intermediate Type: Breakout session Category: ISV & Partners Topics: Networking & Edge Dive into AWS networking fundamentals. Learn essential skills configuring your network in a single region before scaling to global, multi-Region architectures. In this session, explore various VPC connectivity methods, hybrid network integration, and secure traffic management techniques. Whether new to AWS or wanting to expand networking building blocks, this session delivers practical insights into best practices and proven architecture patterns for establishing global connectivity. ### Playbook (editorial commentary) **The concept.** VPC connectivity, hybrid network integration, secure traffic. Single → multi-region. Best practices and proven architecture patterns. **Why it matters.** Networking is the load-bearing layer most teams under-invest in. Multi-region without networking discipline is multi-region with bugs. **The hard parts.** VPC peering vs. Transit Gateway vs. PrivateLink. Picking right has cost and operability implications. **Playbook moves.** (1) Transit Gateway for >3 VPCs. Don't accumulate peering connections. (2) Plan egress costs into the architecture explicitly. (3) Document the network topology — most orgs can't draw their own. **The surprise.** The most expensive networking mistake at scale is data transfer costs you didn't plan for. NAT gateway egress, cross-AZ traffic, inter-region replication add up surprisingly fast. Cost-model your traffic *before* architecting; reverse-engineering is brutal. --- --- ## MAE205 — AI at Speed of News: Unlocking Value from Media with Generative AI URL: https://aws-summit-2026-kb.pages.dev/sessions/MAE205 Level: intermediate Type: Breakout session Category: Media & Entertainment Topics: Media & Entertainment; Databases & Aurora; Generative AI & Foundation Models; OpenSearch & Vector Search; Agentic AI For media and communications organizations, the ability to rapidly discover, repurpose, and distribute content across platforms directly impacts revenue and audience engagement. This session examines how Generative AI is transforming content operations through intelligent metadata extraction, semantic search, and automated workflow orchestration. Using a case study from a global media organization managing 13 petabytes of content growing at 3,000 hours daily, we'll explore practical implementations using Amazon OpenSearch for multimodal retrieval, Amazon Neptune for knowledge graphs, and agentic AI for content assembly. Learn how organizations are achieving faster time-to-market, improved content monetization, and enhanced audience experiences through AI-powered content discovery and recommendation systems ### Playbook (editorial commentary) **The concept.** 13PB content, 3,000 hours/day growth. Metadata extraction, semantic search, automated workflow. OpenSearch + Neptune knowledge graphs + agentic content assembly. **Why it matters.** Media monetisation scales with discoverability. Generative AI removes the metadata bottleneck that's plagued the industry for decades. **The hard parts.** Knowledge graphs decay without maintenance. Stale graphs produce confidently wrong recommendations. **Playbook moves.** (1) Plan for graph maintenance from day one. (2) Schema evolution is the tax — budget for it. (3) Tier content by freshness; old content can use older metadata. **The surprise.** Multimodal retrieval (vision + text + audio together) is finally good enough for production media use cases. The unlock is *search-by-vibe* ("find me clips like this one"), not just keywords. That changes editorial workflows fundamentally — editors become curators, not searchers. --- ### Live monitored sources - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [A2A Protocol Security: Authenticating Agent-to- ...](http://securew2.com/blog/a2a-protocol-security) — securew2.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Decision Traces: Essential AI Infrastructure for Enterprise Scale](https://atlan.com/know/what-are-decision-traces-for-ai-agents/) — atlan.com (2026-05-06): Core idea: The Context Graph resource guide defines context graphs as a 'living record of decision traces' used for execution validation and temporal reasoning. It distinguishes them from knowledge graphs (static entities) and vector databases (semantic similarity) by their nativ - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [WebMCP: How Browsers Are Becoming Native Platforms for AI Agents | Kassebaum Engineering](http://kassebaumengineering.com/insights/webmcp-ai-agents-browser-interaction) — kassebaumengineering.com (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## STP101 — Driving Profitable Growth with Generative AI: From Prompt to Product URL: https://aws-summit-2026-kb.pages.dev/sessions/STP101 Level: foundational Type: Workshop Category: Startups Topics: Startups & Innovation; Generative AI & Foundation Models; Manufacturing & Industry 4.0 Generative AI is the largest disruption to software company business models since the emergence of SaaS. In this workshop we'll cover best practices software companies use to take AI-native products from pilot to production, including identifying use cases that drive business value, features that accelerate adoption and pricing strategies that result in profitable growth. This will include an interactive session - so bring your ideas and collaborate with other startups to help find generative AI features that add value to your product ### Playbook (editorial commentary) **The concept.** Best practices for AI-native software companies: identifying value-driving use cases, features that drive adoption, pricing for profitable growth. Interactive workshop format. **Why it matters.** AI is the largest software business model disruption since SaaS. Companies that don't redesign their economics get crushed. **The hard parts.** Cost structures don't match SaaS. Per-user pricing leaks margin badly. Profit margins compress 10–30 points without intervention. **Playbook moves.** (1) Cost-model every AI feature *before* pricing it. (2) Outcome-based pricing where possible; hybrid (base + usage) where not. (3) Re-price quarterly until you understand the cost curve. **The surprise.** Many AI startups underestimate their Day-1 cost structure by 3–5×. Inference costs scale with engagement; engagement scales with success. Successful adoption can break unit economics if pricing is wrong. Pricing is now an engineering concern, not just a finance one. --- ### Live monitored sources - [Outcome-based pricing for AI Agents - Sierra](http://sierra.ai/blog/outcome-based-pricing-for-ai-agents) — sierra.ai (2026-05-11): Sierra announced an outcome-based pricing model for its AI agents, ensuring that the company is only paid when its AI agents drive real, tangible results for the business, aligning cost directly with success. - [About Us](http://anyway.sh/about-us) — anyway.sh (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [Releases · microsoft/autogen · GitHub](https://github.com/microsoft/autogen/releases) — github.com (2026-05-07): CrewAI released pre-release version 1.14.5a3 on 2026-05-06. Key changes include: - Refactored the CLI into a standalone `crewai-cli` package. - Fixed the status endpoint path from `/{kickoff_id}/status` to `/status/{kickoff_id}`. - Updated the `gitpython` dependency to version >= - [Talent Harbor | Sales Recruitment as a Service (RaaS)](http://talentharbor.com/) — talentharbor.com (2026-05-11): Anyway introduced an outcome-based agentic payment platform that allows AI agent developers to charge based on actual value delivered rather than subscriptions or token usage. Operationally, it integrates agent payment rails with LLM-powered optimization to lower model costs and - [Meta bought some help in its quest for humanoid robots](https://www.businessinsider.com/meta-acquires-assured-robot-intelligence-humanoid-robotics-2026-5) — businessinsider.com (2026-05-02): Meta has acquired Assured Robot Intelligence, a startup specializing in AI models for robots, to advance its humanoid robot technology. --- ## TNC204 — Exam Prep: AWS Solutions Architect Associate URL: https://aws-summit-2026-kb.pages.dev/sessions/TNC204 Level: intermediate Type: Breakout session Category: Other Topics: Security, Identity & Compliance Prepare for the AWS Certified Solutions Architect — Associate exam with this interactive study session. Review the AWS Certified Solutions Architect — Associate Official Practice Question Set, and get your questions answered by the subject matter expert leading the study session. ### Playbook (editorial commentary) **The concept.** Interactive exam study session. Official practice questions; subject matter expert answers. **Why it matters.** Certifications are baseline credentials. Useful but not differentiating. **The hard parts.** Exam content evolves. Third-party prep lags AWS's official material. **Playbook moves.** (1) Use AWS official practice questions. (2) Avoid third-party prep that's >12 months old. (3) For senior engineers, push beyond Associate to Specialty certs. **The surprise.** The certs that matter for senior engineers are no longer Associate-level. SAA is table stakes now. Specialty certs (security, ML, networking, advanced networking) are where market differentiation actually happens for individual contributors. --- --- ## PRT108-S — From Experiment to Production: Unlock AI Deployment Bottlenecks URL: https://aws-summit-2026-kb.pages.dev/sessions/PRT108-S Level: foundational Type: Lightning talk Category: Partner Showcase Topics: Industry Spotlight: Public Sector & Government How fast can AI drive results Banks seek competitive advantage while staying compliant, and the public sector aims to enhance citizen services and boost productivity. In this session, we demonstrate how Synthetic Data and robust AI Governance operationalise AI with confidence. Learn how SAS Viya on AWS enables effective deployment, unlocking faster, better decisions through a proven AI framework. ### Playbook (editorial commentary) **The concept.** Synthetic data + AI governance for compliant AI deployment. Banking and public sector focus. **Why it matters.** Data scarcity and privacy regulations make synthetic data essential for some use cases — particularly testing and edge-case coverage. **The hard parts.** Synthetic data quality varies. Bad synthetic data poisons models in subtle, hard-to-detect ways. **Playbook moves.** (1) Validate synthetic data against real holdouts. (2) Don't trust unvalidated synthetic data, especially for training. (3) Document the generation process for audit. **The surprise.** Synthetic data is most valuable not for training but for *testing*. Edge cases that don't occur in real data can be synthesised to stress-test models. That's the underrated use case — and it solves problems training-set augmentation doesn't. --- ### Live monitored sources - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Enterprise AI Agents Move From Pilot to Production: What 2026 ...](https://insights.reinventing.ai/articles/ai-agents-enterprise-production-2026-02-25) — insights.reinventing.ai (2026-05-05): Microsoft's 2026 Data Security Index reports that more than 80% of Fortune 500 companies are now running active AI agents in production, integrated across sales, finance, customer service, and security workflows. - [Agentic AI Enterprise Use Cases — 30+ Real Deployments (2026)](https://www.ampcome.com/post/post-agentic-ai-enterprise-use-cases) — ampcome.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [How enterprises are building AI agents in 2026 | Claude](https://claude.com/blog/how-enterprises-are-building-ai-agents-in-2026) — claude.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [How Fortune 500 Companies Are Scaling Agentic AI to Production](https://ai2.work/blog/how-fortune-500-companies-are-scaling-agentic-ai-to-production) — ai2.work (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach --- ## DEV206 — AI Isnt Just for Developers: Using Kiro CLI & AWS MCP for Cloud Ops URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV206 Level: intermediate Category: Developer Tools Topics: Model Context Protocol (MCP); Kiro & Spec-Driven Development; Observability & Monitoring; Media & Entertainment; Security, Identity & Compliance; Manufacturing & Industry 4.0 You cant turn your head sideways without seeing a slew of articles, blogs, or videos about AI, and most of them focus on developer tooling and writing code. But AI isnt just for developers. Its an incredibly powerful tool for operations folks, too.In this lightning talk, Ill share how I use Kiro CLI and the Kiro console with AWS Model Context Protocol (MCP) integrations for day-to-day cloud operations. From information gathering and log analysis to reporting and IAM policy interpretation, these tools help reduce cognitive load and speed up your output when working with AWS environments.Ill also discuss how I used Kiros spec-driven development approach to build a Python-based reporting tool, despite not being a software developer.This session is designed to make AI tooling feel approachable and practical for anyone working in AWS — not just developers. ### Playbook (editorial commentary) **The concept.** AI tools for ops, not just devs. Kiro CLI + AWS MCP integrations for log analysis, IAM policy interpretation, reporting. Spec-driven development for ops tooling. **Why it matters.** Ops engineers were the forgotten audience in early AI tooling. Cloud ops is high-leverage AI territory because tasks are well-bounded. **The hard parts.** Ops tasks span systems. Single-MCP tools miss context across services. **Playbook moves.** (1) Inventory your ops MCPs. (2) Combine where useful (e.g., logs + metrics + IAM). (3) Build internal ops tooling using spec-driven development — even non-engineers can ship. **The surprise.** Ops engineers using AI tools become more *capable* than developers using AI tools — because ops tasks are highly structured and well-bounded. The AI productivity ceiling is higher in ops than in feature development. Underrated career bet for engineers in 2026. --- ### Live monitored sources - [The best new AI agents in 2026 - Product Hunt](https://www.producthunt.com/categories/ai-agents?order=recent_launches&page=1) — producthunt.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [AI MCP OAuth2 - Plugin - Kong Docs](http://developer.konghq.com/plugins/ai-mcp-oauth2) — developer.konghq.com (2026-05-08): Azure.Mcp.Server released version 3.0.0-beta.10, which improves tool validation to ensure the complete registered tool set is recognized at runtime and enhances MSAL error handling with PII-safe telemetry and more accurate exception mapping. - [GitHub - modelcontextprotocol/ext-apps: Official repo for spec & SDK of MCP Apps protocol - standard for UIs embedded AI chatbots, served by MCP servers · GitHub](https://github.com/modelcontextprotocol/ext-apps) — github.com (2026-05-02): The `modelcontextprotocol/ext-apps` repository added 15 new example MCP server implementations, including tools for Map, Three.js, ShaderToy, Sheet Music, Wiki Explorer, Cohort Heatmap, Scenario Modeler, Budget Allocator, Customer Segmentation, System Monitor, Transcript, Video R - [AI News — 2026-05-05 | SkillsLLM](https://skillsllm.com/news/ai-news-2026-05-05) — skillsllm.com (2026-05-08): Azure.Mcp.Server released version 3.0.0-beta.10, which improves tool validation to ensure the complete registered tool set is recognized at runtime and enhances MSAL error handling with PII-safe telemetry and more accurate exception mapping. --- ## ISV101 — How AI is Transforming Pharmacy Care with Amazon Nova:MedAdvisor Story URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV101 Level: foundational Type: Lightning talk Category: ISV & Partners Topics: Voice & Conversational AI; Industry Spotlight: Healthcare & Life Sciences; Generative AI & Foundation Models MedAdvisor, Australia's leading medication management platform connecting over 90% of community pharmacies, faced a critical challenge: pharmacists were losing hours daily to manual documentation, diverting time from patient care. To solve this, MedAdvisor built an AI-powered Scribe using Amazon Nova on Amazon Bedrock. The tool listens to pharmacy consultations in real time, transcribes conversations, and automatically generates structured clinical notes. Through iterative prompt engineering, output quality improved from 3.5 to nearly 4.5 out of 5, surpassing market alternatives. Now in beta across Australia, the solution went from concept to production-grade clinical documentation in under six months. ### Playbook (editorial commentary) **The concept.** Real-time transcription of pharmacy consultations on Amazon Nova via Bedrock. Auto-generated structured clinical notes. Quality 3.5 → 4.5 through iterative prompt engineering. Beta across Australia in <6 months. **Why it matters.** Pharmacist time is expensive and clinical. Removing documentation toil is high-ROI in regulated healthcare. **The hard parts.** Pharmacy consultations are semi-private. Patient consent and data handling matter — and vary by jurisdiction. **Playbook moves.** (1) Consent flows first. (2) Don't deploy STT in consultation rooms without patient knowledge. (3) Treat the prompt as load-bearing IP — version, review, test it. **The surprise.** Iterative prompt engineering moved quality from 3.5 to 4.5 — meaningful, but more importantly a sign that prompts are now *load-bearing IP*. Treat them like code: versioned, code-reviewed, regression-tested. Most orgs still treat prompts as folklore. --- ### Live monitored sources - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Agentic AI Enterprise Use Cases — 30+ Real Deployments (2026)](https://www.ampcome.com/post/post-agentic-ai-enterprise-use-cases) — ampcome.com (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [How Fortune 500 Companies Are Scaling Agentic AI to Production](https://ai2.work/blog/how-fortune-500-companies-are-scaling-agentic-ai-to-production) — ai2.work (2026-05-07): Ampcome has published a report detailing 30+ production AI agent deployments across 8 industries. Key deployments include: Smart Grid analytics for 25+ cities (150m people), a retail chain with 700+ stores, a multinational logistics firm, and a global teacher community (1m+ teach - [How Fortune 500 Companies Are Moving Agentic AI Into Production](https://ai2.work/blog/how-fortune-500-companies-are-moving-agentic-ai-into-production) — ai2.work (2026-05-05): Anthropic has deployed production AI agents using Claude Opus 4.7 at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa for high-stakes financial workflows including KYC, underwriting, and insurance claims. AIG reported that these agents scored 88% as accurate as human experts on - [What Is the ROI of Deploying AI Agents? Real Numbers From 2026](https://bananalabs.io/blog/ai-agent-roi) — bananalabs.io (2026-05-12): 2026 Industry benchmarks for production AI agent deployments report significant ROI across Fortune 500 and major enterprises. According to IBM's 2026 AI Agent Economic Study (surveying 2,400 deployments), production AI agents delivered a median 12-month ROI of 171%. McKinsey's 20 --- ## STP207 — How RedOwl Built Real-Time Financial Governance and Control on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/STP207 Level: intermediate Category: Startups Topics: Agentic AI How RedOwl Built Real-Time Financial Governance and Control on AWS: RedOwl is a real-time governance and control platform that enforces policy before a single dollar moves combining agentic AI and pre-transactional intelligence to give CFOs and finance leaders the control to turn finance into a proactive business enabler. This session unpacks the architecture behind that capability, giving technology leaders and practitioners a practical guide to deploying AI-powered financial governance that earns CFO trust and drives enterprise-wide impact. ### Playbook (editorial commentary) **The concept.** Real-time enforcement of policy *before* a transaction settles. Agentic AI + pre-transactional intelligence for CFO controls. **Why it matters.** Most financial controls are detective (they spot bad transactions after). RedOwl is preventive (it blocks them at the door). **The hard parts.** Latency. The policy check must happen in the transaction path without slowing it. **Playbook moves.** (1) Cache policies aggressively. (2) Pre-compute risk profiles. Only the marginal check happens at transaction time. (3) Design false-positive handling carefully. **The surprise.** The interesting design question isn't "can we block bad transactions" — it's "what's the false-positive cost?" A blocked legitimate transaction costs more than detecting a fraud after the fact. Calibration of the model is more important than detection accuracy alone. --- ### Live monitored sources - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [The Operator Vault Publishes Free OpenClaw API ...](http://usatoday.com/press-release/story/27628/the-operator-vault-publishes-free-openclaw-api-database-for-ai-agent-builders) — usatoday.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios --- ## BIZ201 — AI-Everywhere: Transform Customer Interactions into Memorable Moments URL: https://aws-summit-2026-kb.pages.dev/sessions/BIZ201 Level: intermediate Type: Breakout session Category: Business Applications Topics: Voice & Conversational AI; Security, Identity & Compliance The enterprise landscape is shifting rapidly, forcing businesses to rethink their customer experience (CX) and collaboration strategies. From rising security threats, to dynamic workforces, to the desire to delight customers at every touchpoint: the demands are complex and ever-evolving. AWS empowers you to navigate these challenges with AI embedded from day one across the entire customer journeynot bolted on, but built in. ### Playbook (editorial commentary) **The concept.** AI embedded across the customer journey from day one. Not bolted on — built in. **Why it matters.** Bolted-on AI feels bolted-on. Native AI feels invisible. Customers tell the difference; their behaviour reflects it. **The hard parts.** "Embedded from day one" usually means rebuilding existing systems. That's a multi-quarter program. **Playbook moves.** (1) New surfaces get AI native. (2) Existing surfaces get AI overlays. (3) Don't try to retrofit deeply — the rebuild won't pay back fast enough. **The surprise.** The most successful "AI everywhere" rollouts are the ones where users don't notice AI is there. If users have to *think* about the AI, the integration isn't deep enough. Measure invisibility, not visibility. --- ### Live monitored sources - [newsroom.servicenow.com](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-brings-Autonomous-Workforce-to-every-major-business-function/default.aspx) — newsroom.servicenow.com (2026-05-07): ServiceNow announced a major expansion of its Autonomous Workforce at Knowledge 2026, launching 'AI Specialists' for IT, customer relationship management (CRM), employee service teams, and security and risk. These AI specialists are designed to complete entire business processes --- ## IND201 — Transforming software license efficiency - Human-centered AI on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/IND201 Level: intermediate Type: Breakout session Category: Other Topics: Agentic AI; Voice & Conversational AI; Retrieval Augmented Generation (RAG) As Worley's software landscape expands, manual license governance struggles to keep pace with scale and complexity. While manual optimisation has delivered measurable results, a sustainable approach is needed to scale these outcomes. Software Intelligence Advisor (SIA) is Worley's agentic AI solution that enables optimal license decisions and empowers end users. Underpinned by AWS native data platforms, SIA combines deep usage intelligence with a conversational agent that meets users within existing collaboration tools. Through trusted, context-aware conversations, the agent validates usage patterns and encourages better behaviours — delivering scalable, human-centred optimisation and a pragmatic path to learning what agentic AI can deliver. ### Playbook (editorial commentary) **The concept.** Agentic AI for software license optimisation. Conversational agent in collaboration tools (Teams, Slack). Validates usage patterns and encourages better behaviour. **Why it matters.** Software licensing is a large hidden cost. Most orgs over-license dramatically — single-digit-percent over-licensing across thousands of seats compounds. **The hard parts.** Behavioural change requires gentle nudges, not enforcement. Heavy-handed agents get gamed. **Playbook moves.** (1) Position the agent as advisor, not enforcer. (2) Reward good behaviour visibly. (3) Track license-tier appropriateness, not just user count. **The surprise.** The real software-license waste isn't users with too many licenses — it's users on the *wrong tier*. Downgrade detection is bigger ROI than license reclaim. Most license-management programs miss this entirely because they measure the wrong thing. --- ### Live monitored sources - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Artificial Intelligence May 2026 - arXiv.org](https://arxiv.org/list/cs.AI/current) — arxiv.org (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. - [The Context Graph Revolution: Why Enterprise AI ... - Medium](https://medium.com/@thanapong_18619/the-context-graph-revolution-why-enterprise-ai-needs-decision-lineage-c01d90fd1db4) — medium.com (2026-05-12): Daxn launched an AI agent governance system that provides a full audit trail and captures the complete multi-step journey for every agent action to ensure fast and explainable decisions. --- ## INO101 — From Zero to 270 AI Agents: how Lendi built Guardian URL: https://aws-summit-2026-kb.pages.dev/sessions/INO101 Level: foundational Type: Breakout session Category: Other Topics: Agentic AI; Security, Identity & Compliance; Databases & Aurora When Lendi Group launched Project Aurora on AWS, they bet big on a single super-agent with 270 tools. It floppedtechnically impressive but commercially useless. The agent couldn't sell. The breakthrough came from treating AI like a workforce: specialist agents with clear roles, sales logic embedded in the funnel, and relentless measurement. Engagement tripled. This talk shares the hard lessons from Lendi's 16-week sprint: why capability isn't outcome, why your best prompt engineer might be a 23-year-old closer, and how to architect agentic systems that actually convert. ### Playbook (editorial commentary) **The concept.** Lendi launched a single super-agent with 270 tools. It flopped — technically impressive but commercially useless. The agent couldn't sell. Pivoted to a workforce model: specialist agents with clear roles, sales logic embedded in the funnel. Engagement *tripled*. **Why it matters.** This is the most important anti-pattern in agentic AI right now. Capability ≠ outcome. The "monolithic super-agent" trap is everywhere. **The hard parts.** Specialist architectures need orchestration. Naive single-agent designs lose specialisation; over-decomposed designs thrash on coordination. **Playbook moves.** (1) Decompose by *role*, not by tool. Each agent has a job, not just a toolkit. (2) Embed domain logic in the funnel, not just in prompts. (3) Measure on outcomes (conversions), not capability metrics (tool use). **The surprise.** "Your best prompt engineer might be a 23-year-old closer." The lesson is that *domain expertise* (sales, legal, customer service) matters more than ML expertise when designing customer-facing agents. Hire from the domain, train them on the AI. Most teams have this hiring rule backwards. This may be the most quotable insight of the entire summit. --- ### Live monitored sources - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Comment and Control: GitHub AI Agents as Credential ...](https://labs.cloudsecurityalliance.org/research/csa-research-note-comment-control-github-prompt-injection-20/) — labs.cloudsecurityalliance.org (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## ISV203 — AI Monetization and Pricing Strategies URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV203 Level: intermediate Type: Breakout session Category: ISV & Partners Topics: Voice & Conversational AI; Compute: EC2, Graviton & Nitro Software companies developing AI solutions face unique monetization challenges. AI compute costs run 3-5x higher than standard applications, per-user pricing often yields negative margins, and profit margins typically fall 10-30 points below traditional SaaS. This session introduces a proven framework to help you navigate AI pricing complexities. Learn how to identify value capture attributes, select appropriate pricing models, and build sustainable monetization strategies. We'll cover when to begin pricing considerations, how to apply an AI monetization framework to your solutions, and how to develop an approach tailored to your company's position. Whether defining your initial AI pricing strategy or validating your current approach, gain actionable insights to maximize the value of your AI investments. ### Playbook (editorial commentary) **The concept.** AI compute runs 3–5× standard apps. Per-user pricing yields negative margins. Profit margins fall 10–30 points below SaaS without intervention. Framework for value capture and pricing model selection. **Why it matters.** This is the economic reality most software companies haven't faced yet. The bills are coming. **The hard parts.** Customers expect SaaS-like flat pricing. AI economics don't allow it. Education is part of the sale. **Playbook moves.** (1) Model your top-10 customers' actual usage costs. (2) If pricing doesn't cover the worst-ROI top-10, your model is wrong. (3) Build measurement into the product so the renewal conversation has data. **The surprise.** Outcome-based pricing is the textbook fix and rarely implemented because it requires billing infrastructure most companies lack. *Hybrid* models (base + metered) are more practical and capture most of the benefit. Don't let perfect be the enemy of good — start the metering now even if pricing stays simple. --- --- ## SMB204 — Accelerated Insights from Amazon Connect using AI URL: https://aws-summit-2026-kb.pages.dev/sessions/SMB204 Level: intermediate Type: Breakout session Category: Small & Medium Business Topics: Voice & Conversational AI AAMC partnered with AWS to build an intelligent, AI-driven contact centre that transforms data into actionable insights. The solution unifies customer interaction datacalls, chats, surveys, and QA reviewsinto a single environment where AI continuously analyses and learns. Leaders ask questions in natural language and receive instant, data-backed answers with visual context. This self-optimising system enables faster, more confident decision-making while continuously improving customer experience operations through deeper visibility and adaptive learning capabilities. ### Playbook (editorial commentary) **The concept.** Unified contact-center data — calls + chats + surveys + QA reviews. Natural language queries; instant answers with visuals. Self-optimising system. **Why it matters.** Contact-center managers can't read everything that flows through the floor. AI compresses that. **The hard parts.** Quality of insights depends on quality of data labelling. Garbage in, garbage out applies harder here than usual. **Playbook moves.** (1) Tag conversations consistently. (2) Bad tagging breaks AI insights subtly. (3) Re-tag historical data when taxonomies change. **The surprise.** The most useful contact-center AI insight isn't summary metrics — it's the *outliers*. The 3 calls that broke pattern reveal more than the 10,000 that didn't. Build the outlier-surfacing UX, not the summary dashboard. --- --- ## DEV309 — AI Outputs: Amazon Bedrock Structured Output in Production URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV309 Level: advanced Category: Developer Tools Topics: Security, Identity & Compliance; Generative AI & Foundation Models Parsing LLM responses as JSON worksuntil truncation, missing fields, or malformed output breaks your pipeline. This session examines Amazon Bedrock Structured Output, which enforces schema compliance at the model level rather than relying on prompt instructions. You'll learn how to define response contracts using JSON Schema, understand where native structured output differs from prompt-based approaches, and see practical patterns for integrating reliable model responses into production systems. Attendees will leave with concrete techniques for eliminating output validation failures, designing more resilient AI workflows, and understanding the architectural implications of treating model responses as typed, validated data rather than freeform text. ### Playbook (editorial commentary) **The concept.** Schema-enforced LLM outputs via Amazon Bedrock Structured Output. JSON Schema response contracts. Schema compliance enforced at the model level, not via prompt instructions. **Why it matters.** Parsing-as-validation is fragile. Schema-level enforcement is the right architectural layer. **The hard parts.** Some prompts work better without strict schema constraints. Reliability and creative flexibility trade off. **Playbook moves.** (1) Use structured outputs for downstream-consumed data. (2) Use freeform for human-consumed text. (3) Test both modes; pick per task, not per project. **The surprise.** Strict schema can degrade quality on creative tasks. Reliability and creativity trade off in non-obvious ways. Most teams default to "always structured" or "never structured"; the right answer is per-task. --- ### Live monitored sources - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Releases · google/adk-python - GitHub](https://github.com/google/adk-python/releases) — github.com (2026-05-12): Google ADK (google/adk-python) released v1.33.0 on 2026-05-08. Key changes: added BufferableSessionService; added get_function_calls and get_function_responses on LlmResponse (models); made ADK environment tools truncation limit configurable; allowed injecting credentials into Ap - [What is crewAI?](http://ibm.com/think/topics/crew-ai) — ibm.com (2026-05-12): Google ADK (google/adk-python) released v1.33.0 on 2026-05-08. Key changes: added BufferableSessionService; added get_function_calls and get_function_responses on LlmResponse (models); made ADK environment tools truncation limit configurable; allowed injecting credentials into Ap --- ## ISV104 — hipages Journey Towards an Agentic Engineering Organisation URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV104 Level: foundational Type: Lightning talk Category: ISV & Partners Topics: Agentic AI; Databases & Aurora; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) Discover how hipages, ANZ's leading online construction marketplace, redefined their software development lifecycle by embedding Claude Code on Amazon Bedrock into daily engineering workflows. Join this session to get practical insights into maximizing Claude Code on Bedrock, understanding the decision process behind selecting this solution, and learning how hipages is pioneering an AI-first strategy that's transforming operations across their entire business. ### Playbook (editorial commentary) **The concept.** Claude Code on Bedrock embedded in daily engineering workflows. AI-first strategy across operations. **Why it matters.** Engineering ops are where AI-native culture takes hold first; the rest of the org follows. **The hard parts.** Adoption is uneven. Some teams take to it; others resist. Resistance reasons are often legitimate. **Playbook moves.** (1) Track adoption per team, not aggregate. (2) Identify resistance reasons specifically — sometimes the resisters have a point. (3) Make the choice of Bedrock vs. SaaS LLM explicit and documented. **The surprise.** The choice of Claude Code on Bedrock (vs. SaaS) isn't just compliance — it's *cost predictability*. Self-hosted inference at volume often beats per-token pricing once you cross a threshold. Run the numbers; the threshold may be lower than you expect. --- ### Live monitored sources - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Agentic Benchmarks 2026: Tool Use, Browsing, Computer Use | BenchLM.ai](https://benchlm.ai/agentic) — benchlm.ai (2026-05-12): BenchLM.ai updated its Agentic Benchmarks leaderboard on 2026-05-11. The update introduced two new benchmarks: 1) OpenHands Index, a holistic coding-agent benchmark covering issue resolution, frontend work, greenfield development, testing, and information gathering; and 2) SWE-At - [How to Build an Agentic AI Strategy With Process Intelligence](http://skan.ai/blogs/process-intelligence-for-agentic-ai-enterprise-automation) — skan.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Decision Traces for Agentic Operations: Why Agents Need ...](https://xmpro.com/decision-traces-for-agentic-operations-why-agents-need-operational-memory/) — xmpro.com (2026-05-08): XMPro introduced the concept of 'operational memory' powered by decision traces, which capture the reasoning behind specific actions (including exceptions and human judgment) rather than just general rules. This is implemented via a decision trace layer in the orchestration path --- ## STP216 — Building AI Agents: From Open-Source Frameworks to Production-Grade URL: https://aws-summit-2026-kb.pages.dev/sessions/STP216 Level: intermediate Category: Startups Topics: Generative AI & Foundation Models; Compute: EC2, Graviton & Nitro; Code Generation & AI-Assisted Development; Startups & Innovation; Agentic AI; Machine Learning & SageMaker; Industry Spotlight: Healthcare & Life Sciences AI agents are moving from demo to deployment. Startups across ANZ are building production-grade assistants using open-source orchestration frameworks, fine-tuned foundation models, and GPU-accelerated inference on AWS and NVIDIA infrastructure. This panel explores what it actually takes to ship agentic use casesfrom choosing the right models and frameworks to managing latency, cost, and reliability at scale. We'll hear from AirTree VC on where the investment thesis is heading, from NVIDIA on how accelerated compute is shaping the agent stack, and from Heidi Health building and scaling these systems in production. Whether it's vertical agents for healthcare, customer support, or code generation, we'll focus on what's working, what's hype, and where the real startup opportunities lie in the agent ecosystem. ### Playbook (editorial commentary) **The concept.** Production-grade agents using open-source orchestration, fine-tuned models, GPU acceleration on AWS + NVIDIA. Panel: AirTree VC + NVIDIA + Heidi Health. **Why it matters.** Open-source agent stacks are maturing. Vendor-locked stacks now have credible alternatives. **The hard parts.** Open-source = self-hosted ops. Most startups can't afford the overhead. **Playbook moves.** (1) Use managed services until scale or differentiation justifies self-hosted. (2) Don't optimise prematurely. (3) Monitor your inference cost trajectory; the crossover point sneaks up. **The surprise.** VC investment thesis is shifting from "agent capability" to "agent vertical depth." Generic agents are commoditising fast; domain-specific agents have moats. If you're raising for a generic agent platform in 2026, you're raising in a saturated market. --- ### Live monitored sources - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [Enterprise AI Agents 2026: Mid-Year Report on What's Working](https://www.ampcome.com/post/enterprise-ai-agents-2026-mid-year-report) — ampcome.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM - [Agentic AI at Scale: The Rackspace Story](http://rackspace.com/resources/agentic-ai-scale-rackspace-story) — rackspace.com (2026-05-09): NVIDIA GTC 2026 reports that Fortune 500 enterprises have scaled from a few to 50-200 production agentic workflows per company. Key drivers include a 10x drop in inference costs via Blackwell Ultra/Rubin hardware and the adoption of the Model Context Protocol (MCP) and NVIDIA NIM - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## DEV310 — Zero-Downtime Migration from Sydney to Auckland (ap-southeast-6) URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV310 Level: advanced Category: Developer Tools Topics: Containers: EKS, ECS & Fargate; Storage: S3, EBS & EFS; Migration & Modernization; Databases & Aurora; Serverless: Lambda & Step Functions; Compute: EC2, Graviton & Nitro; Retrieval Augmented Generation (RAG) With AWS ap-southeast-6 (Auckland) now open, New Zealand organizations can repatriate workloads from Sydney. This advanced session provides practical migration strategies minimizing downtime and eliminating data loss across every layer of your stack. You'll learn region-to-region migration patterns for: *Storage*: S3 replication, EBS snapshots, EFS cross-region transfers *Databases*: RDS read replicas, DynamoDB global tables, self-managed EC2 database replication *Applications*: Lambda, ECS/EKS workload migration, EC2 AMI copying Walk away with a prioritized migration playbook, realistic RTO/RPO targets, and battle-tested sequencing strategies for large-scale data transfers without extended application outages. ### Playbook (editorial commentary) **The concept.** AWS NZ region (ap-southeast-6) now open. Migration patterns for storage (S3 replication, EBS snapshots, EFS), databases (RDS replicas, DynamoDB global tables), applications (Lambda, ECS/EKS, EC2 AMIs). **Why it matters.** NZ data sovereignty was a barrier; now resolved. Repatriation from Sydney is now viable for NZ customers. **The hard parts.** Cross-region replication, DNS cutover, RDS read-replica promotion. Sequencing matters. **Playbook moves.** (1) Tiered cutover by service criticality. (2) Practise on staging that mirrors prod's region topology. (3) Test rollback paths — region-to-region rollback is non-trivial. **The surprise.** NZ region opens up GovTech NZ in ways the Sydney region couldn't. The compliance change unlocks a market segment that wasn't accessible before. Sales conversations shift from "but our data" to "show me the integration." --- ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Live Agent Upgrades and Cross-Runtime Session Portability (2026)](https://zylos.ai/research/2026-04-17-live-agent-upgrades-session-portability) — zylos.ai (2026-05-03): MarsDevs published the 'Agentic RAG: The 2026 Production Guide', detailing a shift from linear RAG pipelines to a state-machine control loop. This 'Agentic RAG' approach uses a planner agent to decompose queries and iteratively retrieve and evaluate information. It identifies fiv - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [About Us - Firebolt](http://firebolt.io/about-us) — firebolt.io (2026-05-08): Empathic introduced 'Clash', which provides agentic sandboxing to control and restrict specific tools and commands an agent can perform, adding a layer of safety and load management to agent infrastructure. - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios --- ## ISV214 — Grounding AI Agents: How to give your AI real-world data with MCP URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV214 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Agentic AI; Manufacturing & Industry 4.0; Model Context Protocol (MCP) Most AI agents fail not because of models, but because they cant access trusted external data. This session shows how InfoTrack used Model Context Protocol (MCP) to connect agents to authoritative data sources via a compliant and secured gateway. ### Playbook (editorial commentary) **The concept.** MCP gateway connecting agents to authoritative external data sources. Compliant + secured access patterns. **Why it matters.** Most agent failures are *data-access failures*, not model failures. Better grounding beats better models for many use cases. **The hard parts.** Authentication, authorisation, and audit for agent-accessed data. **Playbook moves.** (1) Build MCP servers as governed APIs. (2) Apply the same standards as customer-facing APIs. (3) Log every access — agent traffic patterns differ from human ones. **The surprise.** The MCP server's *logging* is more valuable than its data access. Audit trails for agent data access become regulatory evidence; design for that, not just functionality. Most teams build MCP servers and forget the audit log. --- ### Live monitored sources - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [AI MCP OAuth2 - Plugin - Kong Docs](http://developer.konghq.com/plugins/ai-mcp-oauth2) — developer.konghq.com (2026-05-08): Azure.Mcp.Server released version 3.0.0-beta.10, which improves tool validation to ensure the complete registered tool set is recognized at runtime and enhances MSAL error handling with PII-safe telemetry and more accurate exception mapping. - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [github.com](https://github.com/IBM/mcp-context-forge) — github.com (2026-04-28): IBM released MCP Context Forge, an open-source tool designed for the Model Context Protocol (MCP) ecosystem to streamline how context is managed and provided to AI agents. --- ## IND101 — Test, Learn, Iterate: Amazon Connect Success URL: https://aws-summit-2026-kb.pages.dev/sessions/IND101 Level: foundational Type: Breakout session Category: Other Topics: Voice & Conversational AI; Machine Learning & SageMaker; Retrieval Augmented Generation (RAG) Discover how Flybuys achieved rapid contact centre transformation through early Amazon Connect adoption using AI-powered capabilities and a disciplined Test, Learn, Iterate approach. Starting with a focused pilot, they deployed AI-driven features like intelligent routing, real-time sentiment analysis, and automated quality assurance. They progressed through Launch, Activate, and Consume phasescapturing baseline metrics, scaling through peer-led training, and continuously refining AI performance based on weekly feedback loops. The results: reduced AHT, improved CSAT, 100% AI-powered QA coverage, and measurable ROI. This demonstrates that early AI adoption delivers calculated, data-driven transformation. ### Playbook (editorial commentary) **The concept.** Disciplined Test-Learn-Iterate approach to Amazon Connect. Pilot → Launch → Activate → Consume phases. Weekly feedback loops. Intelligent routing, real-time sentiment, automated QA. **Why it matters.** Method matters more than tools. The TLI discipline outperforms ad-hoc. **The hard parts.** Discipline is the hard part. Most teams skip the "Learn" step under delivery pressure. **Playbook moves.** (1) Block weekly learn sessions. Make them un-skippable. (2) Capture baseline metrics before any AI deployment. (3) Treat each phase as a checkpoint, not a milestone. **The surprise.** 100% AI-powered QA coverage is a sign of *process maturity*, not AI capability. The technology was already there; the change was the discipline to use it consistently. Most "we couldn't do AI QA" excuses are really "we don't have the process discipline." --- ### Live monitored sources - [IBM announcements at Think 2026 to advance the agentic era](https://www.ibm.com/new/announcements/ibm-announcements-at-think-2026) — ibm.com (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Best AI Agent Memory Systems in 2026: 8 Frameworks Compared](https://vectorize.io/articles/best-ai-agent-memory-systems) — vectorize.io (2026-05-06): IBM introduced 'Real-time context on watsonx.data', which provides AI agents with data that is continuously accessible as it changes. Using a Real-Time Context Engine in partnership with Confluent, the system combines streaming data with semantic enrichment and governance, allowi - [Notion launches its first AI agents for data analysis and task ... - MLQ.ai](http://mlq.ai/news/notion-launches-its-first-ai-agents-for-data-analysis-and-task-automation) — mlq.ai (2026-05-10): Matt Shumer announced 'Agent Relay,' a dedicated infrastructure layer for AI agents designed to handle persistent history, real-time events, search, and communication structures including channels, threads, and direct messages. - [Mem0 - The Memory Layer for your AI Apps](https://mem0.ai/) — mem0.ai (2026-05-09): Mem0 introduced 'Memory Decay,' a technical approach to long-term memory management that mimics human forgetting. The system implements a ranking score for memories that is reinforced upon each single retrieval and gently decayed over time if the memory remains untouched. This pr - [Memory Decay - Mem0 - docs.mem0.ai](https://docs.mem0.ai/platform/features/memory-decay) — docs.mem0.ai (2026-05-09): Mem0 introduced 'Memory Decay,' a technical approach to long-term memory management that mimics human forgetting. The system implements a ranking score for memories that is reinforced upon each single retrieval and gently decayed over time if the memory remains untouched. This pr --- ## IND206 — How scalable data foundations helped TGE unlock the power of AI URL: https://aws-summit-2026-kb.pages.dev/sessions/IND206 Level: intermediate Type: Breakout session Category: Other Topics: Voice & Conversational AI; Security, Identity & Compliance; Retrieval Augmented Generation (RAG) In one of Australia's most operationally complex industries, Team Global Express (TGE) turned data into a strategic asset, and AI into a competitive edge. In 2025, TGE invested in data modernisation, establishing an AWS native data platform which now serves as the operational heartbeat of its logistics network. On this foundation, TGE is delivering compounding business value through rapid deployment of AI solutions across multiple domains. Join this session to learn how TGE secured board-level backing, built a lean AI team, and is scaling pragmatic, cost-effective AI — including the lessons learned along the way and whats next on their roadmap. ### Playbook (editorial commentary) **The concept.** AWS-native data platform as the operational heartbeat of TGE's logistics network. Lean AI team. Pragmatic, cost-effective AI across multiple domains. Board-level backing earned through demonstrated wins. **Why it matters.** Logistics is data-rich and operationally complex. Data foundation work pays back fast in this sector. **The hard parts.** Board-level backing requires demonstrated wins. Lean teams have to prove value before they can scale headcount. **Playbook moves.** (1) Pick one quick-win use case for the first six months. (2) Use it to fund the rest. (3) Stay lean longer than feels comfortable — the bloated central AI team is the failure mode. **The surprise.** TGE built a "lean AI team" — not a centralised big team. The org pattern that works at logistics-scale companies is *small federated teams*, not centralised AI hubs. The hub-and-spoke model that worked for data engineering doesn't translate well to AI. --- ### Live monitored sources - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [Firestore: Agentic AI, Search, and MongoDB Compatibility | Google Cloud Blog](https://cloud.google.com/blog/products/databases/firestore-agentic-ai-search-and-mongodb-compatibility) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [Introducing Spanner Omni | Google Cloud Blog](https://cloud.google.com/blog/products/databases/introducing-spanner-omni) — cloud.google.com (2026-05-07): Google Cloud announced updates to Firestore designed for agentic development, including native integrations with AI Studio and third-party coding agents (e.g., Claude Code, Cursor, Codex). The update introduces 'Firestore Skills' and a remote MCP service to connect external agent - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg --- ## INO102 — Partnering for Scale & Innovation URL: https://aws-summit-2026-kb.pages.dev/sessions/INO102 Level: foundational Type: Breakout session Category: Other Topics: Migration & Modernization; Serverless: Lambda & Step Functions Discover how Sportsbet, Australia's leading sports betting platform, transformed its technology organization through strategic AWS partnerships. Learn how they built an enterprise-wide AI learning culture that achieved record-breaking certification results, and completed a serverless-first modernization that delivered significant cost savings and emissions reductions while handling massive scale. This session shares practical strategies for building executive buy-in, establishing effective AWS partnerships, and creating a culture where innovation thrivesdemonstrating how to focus internal resources on competitive differentiation while partnering for technical expertise to accelerate transformation. ### Playbook (editorial commentary) **The concept.** Strategic AWS partnerships. Enterprise AI learning culture. Serverless-first modernisation delivering cost savings + emissions reductions at massive sports-betting peak load. **Why it matters.** Sports betting peaks are extreme; serverless is the right primitive. Pattern generalises to other peak-driven workloads. **The hard parts.** Cultural change to AI-learning. Certification at scale across thousands of staff. **Playbook moves.** (1) Make certification a *career signal*, not a checkbox. (2) Reward learning publicly. (3) Build certification into hiring and promotion criteria. **The surprise.** Serverless's environmental impact is finally measurable. Cost reductions correlate strongly with emissions reductions; ESG reporting and cost optimisation now align. This is a CFO+sustainability convergence point that didn't exist 3 years ago. --- ### Live monitored sources - [IBM Consulting Expands AI Capabilities to Accelerate Enterprise Transformation](https://newsroom.ibm.com/2026-05-06-ibm-consulting-expands-ai-capabilities-to-accelerate-enterprise-transformation) — newsroom.ibm.com (2026-05-08): IBM announced an expansion of its AI capabilities through 'IBM Enterprise Advantage' and 'IBM Consulting Advantage,' including the 'Agent2Agent (A2A)' interoperability standard to allow multi-agent orchestration across enterprise ecosystems (e.g., watsonx Orchestrate and SAP's Jo - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [AI Agent Token Budget Enforcement [2026]](https://www.waxell.ai/blog/ai-agent-token-budget-enforcement) — axell.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios --- ## INO103 — Adopting AI-DLC at Scale: How SEEK Is Transforming Product Delivery URL: https://aws-summit-2026-kb.pages.dev/sessions/INO103 Level: foundational Type: Breakout session Category: Other Topics: Containers: EKS, ECS & Fargate Most organisations use AI to help developers code faster, but few have figured out what needs to change when building is no longer the bottleneck. This session introduces AI-DLC, the next evolution in how teams deliver software: a methodology that compresses specification timelines from months to weeks, and fundamentally changes how product teams operate. SEEK's Principal Product Manager shares how AI-DLC reshaped their people, process, and technology, and how they're now scaling across multiple product teams. You'll hear what's working, what's hard, what they're still figuring out and what it means for how your organisation delivers. ### Playbook (editorial commentary) **The concept.** SEEK's AI-DLC adoption. Specification timelines compressed from months to weeks. People, process, and tech change together. Now scaling across multiple product teams. **Why it matters.** Confirms AI-DLC works in another distinct sector (jobs marketplace). Combined with Xero, Skyjed, AP+, that's four independent validations. **The hard parts.** Scaling across multiple product teams. Each team adapts the methodology differently — that's a feature and a problem. **Playbook moves.** (1) Pilot AI-DLC in 1–2 teams. (2) Don't roll out to all teams simultaneously. (3) Capture local adaptations as patterns; share across teams. **The surprise.** AI-DLC's biggest win isn't speed — it's that PMs and engineers finally have the same artefact (the spec). Cross-functional alignment is the underrated benefit. The methodology fixes a meeting problem more than a coding problem. --- ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits --- ## SMB205 — How Blackmores accelerated SAP RISE connectivity with an EBA and Kiro URL: https://aws-summit-2026-kb.pages.dev/sessions/SMB205 Level: intermediate Type: Breakout session Category: Small & Medium Business Topics: Industry Spotlight: Healthcare & Life Sciences; Kiro & Spec-Driven Development Healthcare company, Blackmores, accelerated their SAP RISE program by utilising Experience Based Acceleration (EBA) to build a production-ready AWS Landing Zone with Terraform in two days using Kiro. ### Playbook (editorial commentary) **The concept.** Production-ready AWS Landing Zone built with Terraform in 2 days using Kiro + AWS Experience Based Acceleration (EBA). **Why it matters.** Landing zones traditionally take months. 2 days is order-of-magnitude different — it changes what's possible in a quarter. **The hard parts.** 2 days is achievable with EBA's pre-baked patterns. Custom requirements still take time; don't believe blanket numbers. **Playbook moves.** (1) Match your needs to EBA patterns. (2) Customise minimally. (3) If you can't fit EBA, the landing zone is going to take months — plan accordingly. **The surprise.** The bigger story isn't speed — it's *reproducibility*. A 2-day landing zone you can rebuild in 2 days is more valuable than a 6-month landing zone you can't. Reproducibility unlocks DR scenarios that previously sat in the "we'll figure it out" pile. --- --- ## STP214 — Create hyper-personalized voice interactions with Amazon Nova Sonic URL: https://aws-summit-2026-kb.pages.dev/sessions/STP214 Level: intermediate Type: Breakout session Category: Startups Topics: Voice & Conversational AI We've all experienced itthat moment when an AI assistant creates more problems than it solves because of its robotic responses and lack of contextual awareness. It operates in a vacuum, making customer experiences less than delightful. In this session, you will learn how to create meaningful and hyper-personalized customer experiences with Amazon Nova Sonic and design patterns to enable new use cases. You will discover how Amazon Nova Sonic's tool use and function calling capabilities enable integration with external APIs and internal and external data sources to provide context that makes for delightful voice-based customer interactions. ### Playbook (editorial commentary) **The concept.** Voice AI with tool use and function calling. Integration with external APIs. Context-aware responses powered by Amazon Nova Sonic. **Why it matters.** Voice is becoming a primary interaction channel for many use cases — accessibility, hands-free, regulated phone-based industries. **The hard parts.** Latency. Voice tolerates ~300ms before feeling broken; agent tool calls often exceed that. **Playbook moves.** (1) Optimise for first-token latency. (2) Stream when possible. (3) Pre-compute likely contexts. **The surprise.** The most important "personalisation" in voice isn't preference matching — it's *interruption handling*. Real conversations have interruptions. Voice agents that handle them gracefully feel real; ones that don't feel robotic. Latency is necessary; interruption handling is sufficient. --- --- ## WPS302 — Secure and Resilient Agentic AI for High-Assurance Environments URL: https://aws-summit-2026-kb.pages.dev/sessions/WPS302 Level: advanced Type: Breakout session Category: Public Sector Topics: Agentic AI; Industry Spotlight: Public Sector & Government; Security, Identity & Compliance; Resilience & Disaster Recovery Autonomous AI systems that plan, decide, and act across workflows are transforming how organisations deliver mission-critical services. This session shares security-first architecture best practices for designing, deploying, and governing agentic AI in high-assurance environments, drawing from Australia's Information Security Manual (ISM) and AWS security frameworks. Discover practical patterns for architecting proactive, intelligent services while maintaining security, transparency, and operational resilience through defense-in-depth strategies and purpose-built AWS capabilities. ### Playbook (editorial commentary) **The concept.** Architecture patterns aligned to Australia's Information Security Manual (ISM). Defense-in-depth for autonomous agentic systems. Security, transparency, operational resilience. **Why it matters.** Public sector is moving fast on agentic AI. ISM compliance is non-negotiable in Australian government work. **The hard parts.** ISM doesn't have agent-specific guidance yet (as of 2026). Interpretation matters; getting it wrong creates compliance debt. **Playbook moves.** (1) Engage with the AU government CISO community early. (2) Document your interpretive choices. (3) Build for the standard you expect, not just the one that exists. **The surprise.** ISM compliance for agents will likely drive *private-sector* standards too. If you sell to government, your architectural patterns are about to become reference architectures across regulated industries. > *Related: WPS202 covers similar ground with public-sector framing. Treat as the same content; attend whichever fits your slot.* --- ### Live monitored sources - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Learning Tools & Educational Solutions](http://edu.google.com/intl/ALL_us/workspace-for-education/editions/overview) — edu.google.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [CISA, US and International Partners Release Guide to Secure ...](https://www.cisa.gov/news-events/news/cisa-us-and-international-partners-release-guide-secure-adoption-agentic-ai) — cisa.gov (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [Onyx Security Launches with $40M in Funding to Build the ...](https://www.businesswire.com/news/home/20260311837993/en/Onyx-Security-Launches-with-%2440M-in-Funding-to-Build-the-Secure-AI-Control-Plane-for-the-Agentic-Era) — businesswire.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. --- ## DEV208 — Production-Grade Platforms: Real-World IaC Practices on AWS URL: https://aws-summit-2026-kb.pages.dev/sessions/DEV208 Level: intermediate Category: Developer Tools Topics: DevOps, CI/CD & DevSecOps; Media & Entertainment; Security, Identity & Compliance Managing infrastructure as code sounds straightforward until you're wrangling hundreds of modules across multiple teams and accounts. In this session, a Platform Engineer and Lead Architect share hard-won lessons from building and scaling IaC platforms at large organisations — covering module composition strategies, state isolation patterns, and CI/CD pipelines that keep deployments safe and reproducible. You'll walk away with practical design patterns you can apply immediately, whether you're laying the foundations of your first landing zone or untangling a sprawling estate that's grown faster than its architecture. ### Playbook (editorial commentary) **The concept.** Real-world IaC at scale. Module composition strategies, state isolation patterns, CI/CD pipelines for safe and reproducible deployments. **Why it matters.** IaC fragility is a top cause of outages at scale. The patterns that work for one team don't work at multi-team scale. **The hard parts.** State management in Terraform/CDK. Drift detection. Module composition without circular dependencies. **Playbook moves.** (1) State per environment, not per service. (2) Drift detection in CI. (3) Composition over inheritance for modules. **The surprise.** The hardest IaC problem isn't writing it — it's *deleting* it. Resources orphaned by failed deploys are an iceberg of cost and risk. Audit and clean regularly; most orgs accumulate orphans for years before noticing. --- ### Live monitored sources - [Open-Source AI Agent Infrastructure Reaches Production Maturity](https://insights.reinventing.ai/articles/ai-agents-open-source-production-2026-03-24) — insights.reinventing.ai (2026-05-06): Galileo released Agent Control, an open-source (Apache 2.0) control plane designed for the centralized governance, real-time policy enforcement, and safety of AI agents. It allows developers to integrate governance hooks using a @control() decorator, decoupling policy management - [Identity Digital Launches Neutral, DNS-Anchored ...](http://identity.digital/newsroom/identity-digital-launches-neutral-dns-anchored-identity-standard-for-ai-agents) — identity.digital (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [AI Agent Authentication & Authorization Deep Dive: Reading ...](https://dev.to/kanywst/ai-agent-authentication-authorization-deep-dive-reading-draft-klrc-aiagent-auth-00-5d1) — dev.to (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [GitHub - iflytek/skillhub: Self-hosted, open-source agent skill registry for enterprises. Publish & version skill packages, govern with RBAC and audit logs, deploy on-premise with Docker or Kubernet](http://github.com/iflytek/skillhub) — github.com (2026-05-10): SkillHub, an enterprise-grade open-source agent skill registry, released version v0.2.7 on May 9, 2026. SkillHub provides a self-hosted platform for publishing, discovering, and managing reusable agent skill packages with governance features like RBAC and audit logs. The project --- ## ISV213 — From GRC Platform to AI-Native Risk Intelligence on AWS:Protecht Story URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV213 Level: intermediate Type: Lightning talk Category: ISV & Partners Topics: Containers: EKS, ECS & Fargate; Voice & Conversational AI; Security, Identity & Compliance; Generative AI & Foundation Models; Retrieval Augmented Generation (RAG) Protecht, a global leader in enterprise risk management software, partnered with AWS and Caylent to build Cognita AI, an embedded AI assistant purpose-built for governance, risk, and compliance (GRC). Backed by a $280M PSG investment, Protecht built Cognita on a production-grade Amazon EKS foundation, integrating Amazon Bedrock and Anthropic's Claude models with a RAG architecture grounded in Protecht's proprietary GRC content. The result is a contextual, explainable, and auditable AI assistant that guides risk professionals through complex workflows, earning high accolades at the Gartner Enterprise Risk, Audit & Compliance Conference and setting a new benchmark for investor-grade, regulator-trusted AI in months. ### Playbook (editorial commentary) **The concept.** Cognita AI — embedded GRC AI assistant on EKS + Bedrock + Claude. RAG grounded in proprietary GRC content. Backed by $280M PSG investment. **Why it matters.** GRC is heavy on document understanding. AI accelerates risk professionals materially. Investor-grade trust in AI products is achievable. **The hard parts.** Auditable outputs are critical. Black-box recommendations don't pass GRC audits — by definition. **Playbook moves.** (1) Citations on every output. (2) Trace back to source documents. (3) Make the explanation the product, not just the answer. **The surprise.** Investor-grade trust in AI products is achievable when the AI shows its work. Investors and regulators don't fear AI; they fear *opaque* AI. Transparency is the moat in regulated AI products. --- ### Live monitored sources - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits - [Agent-Native Database Architecture 2026: Why REST APIs Fail ...](https://agentmarketcap.ai/blog/2026/04/10/agent-native-database-architecture-2026) — agentmarketcap.ai (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [Releases · microsoft/agent-governance-toolkit · GitHub](https://github.com/microsoft/agent-governance-toolkit/releases) — github.com (2026-05-08): Microsoft released v3.5.0 of the Agent Governance Toolkit, adding enterprise-grade agent identity via Citadel Integration (Entra identity bridge), Multi-Agent Collective Policies for workflow-wide constraints, Intent-Based Authorization for structured lifecycle management (declar - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding --- ## STP202 — Stop Vibing, Start Shipping: How Startups Build with Kiro URL: https://aws-summit-2026-kb.pages.dev/sessions/STP202 Level: intermediate Category: Startups Topics: Agentic AI; Startups & Innovation; Kiro & Spec-Driven Development Prompt-and-pray development doesn't scale. Learn how startup engineering teams are adopting Kiro's spec-driven, agentic approach to go from requirements to working code on AWS with real adoption patterns, pitfalls, and measurable productivity impact. ### Playbook (editorial commentary) **The concept.** Spec-driven, agentic Kiro for startup engineering teams. Real adoption patterns, pitfalls, measurable productivity impact. **Why it matters.** Vibe coding limits scale. Specs scale. **The hard parts.** Startup teams resist process. Spec-driven feels like overhead initially. **Playbook moves.** (1) Start with one spec per feature. (2) Show the time savings on the *second* iteration. (3) Make the spec the artefact you keep — disposable code, durable specs. **The surprise.** Spec-driven development feels slow on the first iteration but compounds. After 3–5 features, the team is shipping faster than vibe-coding ever was. The trough is real; survive it and you win. --- ### Live monitored sources - [Scaling Autonomous Agent Swarms with Distributed Task ...](https://martinuke0.github.io/posts/2026-03-31-scaling-autonomous-agent-swarms-with-distributed-task-orchestration-and-low-latency-communication-protocols/) — martinuke0.github.io (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [Gr4vy supports agentic payments through orchestration ...](http://gr4vy.com/posts/gr4vy-supports-agentic-payments-through-orchestration-and-launches-development-kit-to-prepare-merchants-for-ai-commerce) — gr4vy.com (2026-05-10): At Stripe Sessions 2026 on May 10, 2026, Stripe announced new programmable products and platform features designed to support AI agents and autonomous machine-to-machine commerce, expanding Stripe's economic infrastructure for agent-driven payments. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [The agent control plane becomes the new enterprise buying surface](https://www.linkedin.com/pulse/agent-control-plane-becomes-new-enterprise-buying-andrew-mcpherson-gcd4f) — linkedin.com (2026-05-07): At Cloud Next 2026, Google committed $750 million to a partner fund designed to accelerate the development of agentic AI builds, supporting partners like Accenture and KPMG in scaling AI agent deployment. - [Comment and Control: Prompt Injection to Credential Theft in ...](https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/) — oddguan.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against --- ## FSI202 — Accelerating Payment Innovation: Spec-Driven Development with AWS Kiro URL: https://aws-summit-2026-kb.pages.dev/sessions/FSI202 Level: intermediate Type: Breakout session Category: Financial Services Topics: Containers: EKS, ECS & Fargate; Kiro & Spec-Driven Development; Security, Identity & Compliance; Databases & Aurora; Machine Learning & SageMaker Australian Payments Plusoperator of Australia's critical payment infrastructure including eftpos, BPAY, and NPP, processing millions of daily transactionstransformed their development practices by adopting Spec-Driven Development using AWS Kiro. AP+ manages the payment rails connecting banks, merchants, and consumers throughout Australia. Through intensive Event-Driven Architecture bootcamps and hands-on training, engineering teams now independently run development workshops every two weeks, accelerating delivery of payment platform innovations while maintaining the highest security and compliance standards required for national financial infrastructure. Learn the practical framework for building development velocity in regulated environments. ### Playbook (editorial commentary) **The concept.** Australian Payments Plus (eftpos, BPAY, NPP) adopting spec-driven development with AWS Kiro. Engineering teams now run their own bi-weekly Event-Driven Architecture bootcamps. Critical national financial infrastructure. **Why it matters.** Critical national infrastructure proving spec-driven works in highly regulated environments. The strongest possible validation of the methodology. **The hard parts.** Compliance review for AI-generated code. Audit trails for spec → code lineage. **Playbook moves.** (1) Capture the spec. (2) Sign the spec. (3) Generate from signed specs only. The signed spec becomes the auditable artefact. **The surprise.** Bi-weekly bootcamps run by engineering teams (not central training) is the *sustainability mechanism*. Self-replicating training scales without HR involvement. That's the org-design move that makes AI-DLC stick — without it, the methodology decays. --- ### Live monitored sources - [Experian Announces Agent Trust to Power Trusted AI ...](http://businesswire.com/news/home/20260430719198/en/Experian-Announces-Agent-Trust-to-Power-Trusted-AI-Driven-Commerce) — businesswire.com (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Red Hat adds support for agentic AI development | CIO](https://www.cio.com/article/4169833/red-hats-message-to-enterprises-you-dont-need-to-re-platform-for-ai-agents-2.html) — cio.com (2026-05-12): ServiceNow introduced 'Action Fabric' within its AI Control Tower, a usage-based pricing and metering system for agentic AI ('assists'). The rollout highlights the critical infrastructure need for budget controls to prevent autonomous agents from exhausting credits through recurs - [AI Agent Authentication & Authorization in 2026: What Works ...](https://api.aport.io/blog/best-ai-agent-authentication-authorization-2026) — api.aport.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho - [Prompt Injection in Production Agents: 2026 Taxonomy](https://www.digitalapplied.com/blog/prompt-injection-production-agents-2026-taxonomy) — digitalapplied.com (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf - [Prompt Injection Attack to Tool Selection in LLM Agents](https://www.ndss-symposium.org/wp-content/uploads/2026-s675-paper.pdf) — ndss-symposium.org (2026-05-08): Security Disclosure: Microsoft disclosed two critical vulnerabilities in the Semantic Kernel framework that enable Remote Code Execution (RCE) and sandbox escapes via prompt injection. 1) CVE-2026-26030: A vulnerability in the In-Memory Vector Store's filter function (using unsaf --- ## INO202 — Build and scale AI: from reliable agents to transformative systems URL: https://aws-summit-2026-kb.pages.dev/sessions/INO202 Level: intermediate Type: Breakout session Category: Other Topics: Agentic AI; Generative AI & Foundation Models Many teams move fast with agentic AI prototypes that impress in demos but stall in productionblocked by gaps in reliability, accuracy, and safety. In this session, AWS agentic AI technical leaders will help builders rethink how to build and scale production-grade, trustworthy agentic AI. Learn proven patterns to build and deploy agents that earn trust in the real world. See AWS agentic AI platform Amazon Bedrock AgentCore in action. Discover how AWS customers move fast from sparks of experiment to scaled AI-driven innovation with trust at the core, transforming industries. ### Playbook (editorial commentary) **The concept.** AgentCore in action. AWS's view on production-grade trustworthy agent patterns. From experiment to scaled innovation. **Why it matters.** AWS's reference patterns become customer reference designs. What they say is what they'll productise next. **The hard parts.** AWS patterns assume an AWS-native stack. Hybrid environments need adaptation. **Playbook moves.** (1) Map AWS patterns to your stack. (2) Note where adaptation is needed. (3) Watch which patterns get the most demo time — that's the roadmap signal. **The surprise.** AWS's agent reference patterns lean heavily on Bedrock + AgentCore + Strands. If you're not on those, adoption costs more than the patterns suggest. The "neutral" framing is real but the bias toward the AWS stack is too. --- ### Live monitored sources - [Agentic AI - Union.ai](http://union.ai/solutions/agentic-ai) — union.ai (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## ISV207 — How Canva Scales and Optimizes AI Workloads with Karpenter URL: https://aws-summit-2026-kb.pages.dev/sessions/ISV207 Level: intermediate Type: Breakout session Category: ISV & Partners Topics: Compute: EC2, Graviton & Nitro; Containers: EKS, ECS & Fargate; Retrieval Augmented Generation (RAG) his session explores how Canva leverages Karpenter to scale and optimize diverse workloads on Amazon EKS. Learn how Canva manages AI workloads using On-Demand Capacity Reservations (ODCRs) and EC2 Capacity Blocks for ML, while maximizing resource utilization by intelligently co-locating CPU and GPU workloads on GPU nodes. We will dive into NodePool management strategies for efficient scheduling of AI workloads and examine how Canva uses a range of Amazon EC2 instance types to operate a multi-tenant container orchestration platform for all workloads, optimizing for cost-effectiveness and resource efficiency. Ideal for platform engineers and Kubernetes operators looking to optimize their EKS clusters for both AI and general workloads at scale. ### Playbook (editorial commentary) **The concept.** Karpenter for diverse EKS workloads. On-Demand Capacity Reservations + EC2 Capacity Blocks for ML. CPU + GPU workload co-location on GPU nodes. NodePool strategies for AI scheduling. **Why it matters.** AI workload scheduling on EKS is operationally complex; getting it wrong wastes money fast. **The hard parts.** GPU node utilisation is hard to maximise. Idle GPUs are very expensive. **Playbook moves.** (1) Co-locate batch CPU work on GPU nodes. (2) Let Karpenter handle node lifecycle. (3) Monitor GPU utilisation as carefully as you monitor CPU. **The surprise.** GPU co-location with CPU workloads can hit 70%+ GPU utilisation on otherwise-idle hardware. That's the cost-saving most teams miss because they segregate GPU and CPU workloads "for clarity." Co-location pays back the operational complexity many times over. --- ### Live monitored sources - [What’s new in compute at Next ‘26 | Google Cloud Blog](https://cloud.google.com/blog/products/compute/whats-new-in-compute-at-next26) — cloud.google.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [IBM Consulting Expands AI Capabilities to Accelerate Enterprise Transformation](https://newsroom.ibm.com/2026-05-06-ibm-consulting-expands-ai-capabilities-to-accelerate-enterprise-transformation) — newsroom.ibm.com (2026-05-08): IBM announced an expansion of its AI capabilities through 'IBM Enterprise Advantage' and 'IBM Consulting Advantage,' including the 'Agent2Agent (A2A)' interoperability standard to allow multi-agent orchestration across enterprise ecosystems (e.g., watsonx Orchestrate and SAP's Jo - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Multi-Agent Orchestration Patterns Drive Enterprise ROI in 2026](https://insights.reinventing.ai/articles/ai-agents-orchestration-patterns-2026-03-18) — insights.reinventing.ai (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. --- ## MAE204 — How Amazon Ads Creative Agent uses AWS to democratize ad creation URL: https://aws-summit-2026-kb.pages.dev/sessions/MAE204 Level: intermediate Type: Breakout session Category: Media & Entertainment Topics: Containers: EKS, ECS & Fargate; Voice & Conversational AI; Media & Entertainment; Generative AI & Foundation Models; Machine Learning & SageMaker Media advertisers see up to 25% higher engagement when delivering custom creative to relevant audiences, yet producing quality video ads traditionally requires weeks of expensive and specialized expertise. Discover the inner workings of Amazon Ads new AI Creative Agent, and how it's transforming the creative process by automating and enhancing the generation of multi-format ads to businesses regardless of their size or creative expertise. Explore how Amazon Bedrock, custom-built ML models, GPUs, and model evaluations are used to orchestrate and generate compelling ad creatives into full video productions with professional voiceovers from conversational natural language, while reducing creative development time. ### Playbook (editorial commentary) **The concept.** Multi-format ad generation from natural language briefs. Bedrock + custom ML + GPUs + model evaluations. Voiceovers, full video productions. Democratises ad creation regardless of business size. **Why it matters.** Engagement is up to 25% higher with personalised creative. Personalisation is no longer a luxury for advertisers. **The hard parts.** Brand-safety in generated content. Trademark and copyright issues. **Playbook moves.** (1) Pre-flight every generated asset against brand guidelines. (2) Don't post unchecked content. (3) Build the brand-safety check into the generation pipeline, not as a manual step. **The surprise.** Custom ad generation works because the old "one ad, many viewers" model is obsolete. Personalisation isn't a luxury; it's the new baseline. Advertisers without it are ceding measurable engagement. --- ### Live monitored sources - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [Think 2026: IBM Delivers the Blueprint for the AI Operating ...](https://newsroom.ibm.com/2026-05-05-think-2026-ibm-delivers-the-blueprint-for-the-ai-operating-model-as-the-ai-divide-widens) — newsroom.ibm.com (2026-05-06): GitHub introduced 'Rate Limiting Controls' for Agentic Workflows to prevent runaway agent behavior. The system implements a defense-in-depth architecture including dual concurrency control (per-workflow and per-engine) to prevent parallel execution explosions, 'Safe Output Limits - [Belitsoft Releases AI Agent Development Forecast 2026: 40% of ...](https://www.abnewswire.com/pressreleases/belitsoft-releases-ai-agent-development-forecast-2026-40-of-enterprise-applications-to-include-taskspecific-agents-by-year-end_800878.html) — abnewswire.com (2026-05-05): DeepClaude, a new open-source tool, has been released enabling the use of the Claude Code agent loop with DeepSeek V4 Pro, allowing Claude to orchestrate DeepSeek models for multi-step tasks. - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [IBM Consulting Expands AI Capabilities to Accelerate Enterprise Transformation](https://newsroom.ibm.com/2026-05-06-ibm-consulting-expands-ai-capabilities-to-accelerate-enterprise-transformation) — newsroom.ibm.com (2026-05-08): IBM announced an expansion of its AI capabilities through 'IBM Enterprise Advantage' and 'IBM Consulting Advantage,' including the 'Agent2Agent (A2A)' interoperability standard to allow multi-agent orchestration across enterprise ecosystems (e.g., watsonx Orchestrate and SAP's Jo --- ## SMB203 — From Vision AI to Agentic AI: Real-Time Ops & Compliance in QSR URL: https://aws-summit-2026-kb.pages.dev/sessions/SMB203 Level: intermediate Type: Breakout session Category: Small & Medium Business Topics: Observability & Monitoring; Media & Entertainment; Security, Identity & Compliance; Generative AI & Foundation Models; Compute: EC2, Graviton & Nitro; Agentic AI Fingermark's Eyecue platform turns drive-thru video feeds into real-time operational intelligence for some of the world's largest QSR brands. Using hybrid edge-cloud architecture on AWS, they track every customer journeycapturing precise timing at order points, windows, and bayswhile keeping sensitive data at the edge. Now they're taking the next leap: agentic AI powered by Amazon Bedrock AgentCore. Autonomous agents automatically answer compliance questions"Are there spills Are staff following food handling protocols"replacing manual audits with continuous monitoring. See how a Kiwi company scaled from local innovation to global impact, and from computer vision to autonomous agents. ### Playbook (editorial commentary) **The concept.** Drive-thru video → real-time operational intelligence. Hybrid edge-cloud architecture. Now adding agentic AI for compliance — autonomous agents answer questions like "Are there spills? Are staff following food handling protocols?" **Why it matters.** QSR is operationally intense and labor-heavy. AI directly affects unit economics. **The hard parts.** Edge compute for vision is bandwidth-intensive. Hybrid is necessary, not optional. **Playbook moves.** (1) Edge for inference, cloud for aggregation. (2) Don't ship video to cloud. (3) Tier sensitivity — keep PII at the edge. **The surprise.** The transition from "vision AI tells you what's happening" to "agentic AI tells you what to do about it" is a meaningful step up. Most vision-AI deployments are stuck at observation; the unlock is *action*. That's the next investment frontier in physical-world AI. --- ### Live monitored sources - [43,750% Surge! BNB Chain is Crushing It with 150,000 AI Agents Blasting Through the Track | 小机构集团 on Binance Square](http://binance.com/en/square/post/316311861525314) — binance.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Amazon Builds AI Agent Payments With Coinbase and Stripe](https://thedefiant.io/news/infrastructure/amazon-builds-ai-agent-payments-with-coinbase-and-stripe) — thedefiant.io (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [The AI Agent challenge: From Data Lineage to Cognitive Lineage](https://www.linkedin.com/pulse/ai-agent-challenge-from-data-lineage-cognitive-tim-b%C3%B8gh-morthorst-bk96f) — linkedin.com (2026-05-09): New analysis of Gartner's 2026 predictions indicates that the Context Graph is the critical architectural gap that agents must fill to ensure reliable and scalable decision-making within enterprise settings. - [Edge Delta Makes All Telemetry Pipelines Data ...](http://prnewswire.com/news-releases/edge-delta-makes-all-telemetry-pipelines-data-throughput-limitless-and-free-for-all-customers-302736808.html) — prnewswire.com (2026-05-11): TraceRoot launched an open-source observability platform for AI agents featuring a 'self-healing layer' that captures traces and uses AI to automatically identify bugs and open fix PRs by analyzing source code and GitHub history. It includes an OpenTelemetry-compatible SDK for ca --- ## AIM304 — Agentic AI Meets Responsible AI - Science, Strategy and Practice URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM304 Level: advanced Type: Breakout session Category: AI & Machine Learning Topics: Agentic AI AI agents offer powerful capabilities — and introduce fundamentally new risks that require more than traditional controls. This session explores responsible agentic AI through three lenses: the science, the framework, and a real-world customer story. Understand the scientific frontiers that make agents different — from emergent behaviour and agent-to-agent trust to the challenges of governing systems that plan, negotiate, and act autonomously. Learn the four areas of the AWS Responsible AI framework where agents change the rules, and hear how one of Australia's leading health insurer is putting responsible AI into practice — from strategy to governance to real-world trade-offs. ### Live monitored sources - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [AI Agent Protocol Community Group - World Wide Web Consortium ...](https://www.w3.org/community/agentprotocol/) — 3.org (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i - [ServiceNow expands AI agent governance through deeper ...](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-expands-AI-agent-governance-through-deeper-integration-with-Microsoft/default.aspx) — newsroom.servicenow.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [Announcing the Agent2Agent Protocol (A2A) - Google Developers ...](https://developers.googleblog.com/en/a2a-a-new-era-of-agent-interoperability/) — developers.googleblog.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes --- ## AIM402 — Agentic AI Meets responsible AI: Strategy and best practices URL: https://aws-summit-2026-kb.pages.dev/sessions/AIM402 Level: expert Type: Breakout session Category: AI & Machine Learning Topics: Agentic AI AI agents offer powerful capabilities and require thoughtful design to help manage risks. This session explores responsible agentic AI implementation with appropriate controls and governance. Understand some of the scientific frontiers that inform design considerations, including the language of AI agents, context management, agent interactions, and common sense reasoning. Learn approaches for human oversight, risk mitigation, evaluation methods, and control mechanisms to help align agent behaviors with organizational goals, and help make agentic AI both effective and trustworthy. ### Live monitored sources - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [ServiceNow expands AI agent governance through deeper ...](https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-expands-AI-agent-governance-through-deeper-integration-with-Microsoft/default.aspx) — newsroom.servicenow.com (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Guild Raises $44M to Build the Agent Control Plane](https://www.guild.ai/knowledge/guild-raises-44m-agent-control-plane) — guild.ai (2026-05-08): ServiceNow announced an expansion of its AI agent governance capabilities through a deeper integration with Microsoft, enhancing tool governance and control for enterprise agents. - [Context Graph: The Definitive Resource for AI Decision Traces](https://www.contextgraph.tech/) — contextgraph.tech (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [From AI Agent Sprawl to Unified AI Operations](http://onereach.ai/blog/from-ai-agent-sprawl-to-unified-ai-operations-how-enterprises-can-regain-control) — onereach.ai (2026-05-11): Google and Microsoft have jointly proposed a new W3C standard called WebMCP (Web Model Context Protocol). This standard aims to allow websites to expose structured, callable tools directly to AI agents through a native browser API, fundamentally changing how agents discover and i --- ## FSI203 — How HBF Transformed Claims Processing From Two Weeks to Two Minutes URL: https://aws-summit-2026-kb.pages.dev/sessions/FSI203 Level: intermediate Type: Breakout session Category: Financial Services Topics: Containers: EKS, ECS & Fargate; Retrieval Augmented Generation (RAG) In this session discover how HBF revolutionized claims processing using AWS. By leveraging Amazon Bedrock and Amazon Textract, they cut claim costs from $2 to just 10 cents and reduced the processing time from two weeks to two minutes. With accuracy in the high 90s and 70,000 claims processed monthly, their end-to-end AI-powered architecture for claims processing sets a new benchmark for speed, cost, and customer satisfaction. ### Playbook (editorial commentary) **The concept.** Bedrock + Textract end-to-end claims processing. Cost cut from $2 to $0.10 per claim (95% reduction). Time from 2 weeks to 2 minutes. 70,000 claims/month at high-90s accuracy. **Why it matters.** This is the kind of P&L-changing AI deployment that ends boardroom debates. The numbers don't argue. **The hard parts.** High-90s accuracy still leaves errors. In consumer-facing claims, those errors become complaints, regulatory issues, and brand damage. **Playbook moves.** (1) Tier claims by complexity. Auto-process simple ones. Escalate edge cases to humans. (2) Track the escalation rate as a primary metric. (3) Audit accuracy per claim type — averages hide issues. **The surprise.** $2 → $0.10 is a 95% cost reduction. At that ratio, the question changes from "should we automate?" to "what do we do with the freed-up budget?" That's a strategy conversation, not an ops one. Most boards are unprepared to answer it. --- ### Live monitored sources - [FAQs](http://gruve.ai/gruve-frequently-asked-questions) — gruve.ai (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AgentBudget - Real-time cost enforcement for AI agents](https://agentbudget.dev/) — agentbudget.dev (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. - [AI Agent Memory Systems Cut Costs 60% with Long-Term Context 2026](https://iterathon.tech/blog/ai-agent-memory-systems-implementation-guide-2026) — iterathon.tech (2026-05-05): Vektor Memory published 'The State of AI Agent Memory in 2026', introducing a four-dimensional framework for agent memory: Storage (indexing), Curation (handling contradictions/outdated info), Retrieval (temporal vs. semantic), and Lifecycle (consolidation/retirement). The analys - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [Empathic 2026 Company Profile](http://pitchbook.com/profiles/company/989050-06) — pitchbook.com (2026-05-09): AgentBudget was identified as an open-source Python SDK that provides real-time cost enforcement for AI agents, allowing developers to set a hard dollar limit on any single AI agent session to prevent runaway expenses. --- ## SEC501 — Where Big Ideas Live: How to Actually Read Research Papers URL: https://aws-summit-2026-kb.pages.dev/sessions/SEC501 Type: Chalk talk Category: Security, Identity & Compliance Topics: Media & Entertainment; Security, Identity & Compliance; Databases & Aurora Research papers hold ideas you won't find in any docs, blog posts, or explainer videos. They're also brutal to read, and can leave you frustrated. This talk is about how to actually read one, layer by layer. First what the paper is about, then why it's important, and finally how it works. We'll start with AWS IAM Access Analyzer and it's paper on Stratified Predicate Abstraction. Then we'll work backwards through the research papers it's built on, learning about SMT solvers and Decision Procedures. You'll walk out with a method you can use on any paper, in any field. ### Playbook (editorial commentary) **The concept.** A method for reading research papers, layer by layer: what is the paper about → why is it important → how does it work. Worked example: AWS IAM Access Analyzer's paper on Stratified Predicate Abstraction. Then works backward through the foundational papers on SMT solvers and Decision Procedures. **Why it matters.** Papers contain ideas not in docs, blog posts, or videos. Reading them is the highest-leverage learning skill in tech — it stays useful for decades. **The hard parts.** Papers are intentionally dense. They're written for peer reviewers, not learners. Most engineers bounce off the abstract. **Playbook moves.** (1) Three-pass method: skim for what (5 min), reread for why (15 min), deep-read for how (1+ hour). (2) Read the bibliography to find the foundational papers; understanding cascades backwards. (3) Take notes in your own words — passive reading doesn't stick. ### Live monitored sources - [NIST AI Agent Standards: Enterprise Governance Implications](https://labs.cloudsecurityalliance.org/wp-content/uploads/2026/03/CSA_research_note_NIST_AI_agent_standards_initiative_20260324-csa-styled.pdf) — labs.cloudsecurityalliance.org (2026-05-09): Fastio has formalized new architectural patterns for scaling multi-agent systems, including Sequential Handoff (Pipeline), The Router (Dispatcher), Hierarchical (Manager-Worker), and Bidirectional/Joint Collaboration. A significant practical implication is the emergence of 'Deleg - [prnewswire.com](https://www.prnewswire.com/news-releases/constructive-open-sources-agentic-db-the-postgres-memory-layer-for-ai-agents-302755269.html) — prnewswire.com (2026-04-29): Constructive announced and open-sourced "agentic-db": a purpose-built Postgres "memory layer" for AI agents providing long-term episodic memory, conversation and tool-call event logs, token accounting, a versioned skills/tools registry, rules/behavioral policies for governance, t - [AI Agent Authentication & Authorization in 2026: What Works ...](https://api.aport.io/blog/best-ai-agent-authentication-authorization-2026) — api.aport.io (2026-05-09): A new authorization architecture known as the Three-Layer Model has been proposed by APort. This framework shifts security from prompt-based controls to deterministic infrastructure policies across three layers: Authentication (using OAuth 2.0, OIDC, SPIFFE/SVID, mTLS), API Autho --- ## WPS202 — Secure and Resilient Agentic AI for High-Assurance Environments URL: https://aws-summit-2026-kb.pages.dev/sessions/WPS202 Level: intermediate Type: Breakout session Category: Public Sector Topics: Gaming & Interactive Media; Industry Spotlight: Public Sector & Government; Security, Identity & Compliance; Databases & Aurora; Agentic AI As governments worldwide race to modernise services, agentic AI based autonomous systems that plan, decide, and act across workflowsis emerging as a game-changer. This session delivers a compliance-first blueprint for designing, deploying, and governing agentic AI in the Australian public sector, fully aligned with Australias Information Security Manual (ISM) and national AI standards. Learn how to transform legacy processes into proactive, citizen-centric services while ensuring security, transparency, and trustwithout regulatory risk. ### Live monitored sources - [AI Detection & Response: Secure Your Systems | Aiceberg](http://aiceberg.ai/) — aiceberg.ai (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. - [CISA and partners publish new advice on AI agent safety](https://cybernews.com/ai-news/cisa-and-partners-publish-new-advice-on-ai-agent-safety/) — cybernews.com (2026-05-05): Policy Proposal/Guidance: CISA and international partners released the 'Guide to Secure Adoption of Agentic AI' in May 2026. The guide provides developers, vendors, and operators with best practices for securing agentic AI systems and recommends specific actions to defend against - [How to Scale Backend Infrastructure for the Age of Agentic AI](https://virtualizationreview.com/articles/2026/02/05/how-to-scale-backend-infrastructure-for-the-age-of-agentic-ai.aspx) — virtualizationreview.com (2026-05-08): Waxell provides a governance layer for infrastructure-layer budget enforcement that wraps LLM requests and tool calls, synchronously terminating sessions before an API call is placed once a per-session or fleet-wide token/cost ceiling is reached, preventing runaway loop scenarios - [CSAI Foundation Announces Key Milestones to Secure the ...](https://cloudsecurityalliance.org/press-releases/2026/04/29/csai-foundation-announces-key-milestones-to-secure-the-agentic-control-plane) — cloudsecurityalliance.org (2026-05-02): Microsoft announced the general availability of Agent 365, a comprehensive control plane for agents focused on observability, governance, and security. Key governance features include a centralized registry of all agents, an admin approval and publication workflow for onboarding - [A2A Net](http://linkedin.com/company/a2anet) — linkedin.com (2026-05-10): Google introduced the A2A (Agent2Agent) protocol, a standardized communication framework designed to enable interoperability between AI agents across different frameworks, teams, and organizations. Launched as part of a broader agentic suite at Google Cloud Next 2026, it includes --- ## STP401 — How WhiteHorse AI Deploys Openclaw Agents on AWS with Amazon Bedrock URL: https://aws-summit-2026-kb.pages.dev/sessions/STP401 Level: expert Category: Startups Topics: Agentic AI; Generative AI & Foundation Models What if every small business could afford a chief of staff that never sleeps In this session, WhiteHorse AI shares how they build autonomous AI agents — "AI employees — that handle invoicing, email triage, lead follow-up, and customer inquiries 24/7. ### Playbook (editorial commentary) **The concept.** Autonomous AI "employees" for small business — chief-of-staff that never sleeps. Invoicing, email triage, lead follow-up, customer inquiries 24/7. Bedrock-powered. **Why it matters.** SMB market is underserved by enterprise AI tooling. Huge volume opportunity for vendors; significant productivity opportunity for the SMBs themselves. **The hard parts.** SMB customers can't afford failure. Reliability matters more than capability for this segment. **Playbook moves.** (1) Constrain agent scope tightly. (2) Better to do less reliably than more unreliably. (3) Onboarding UX matters more than feature breadth — SMB owners don't have time to configure. **The surprise.** SMB AI's killer feature is *consistency*, not capability. The chief-of-staff metaphor is apt: small businesses need someone who shows up every day, not someone who's brilliant occasionally. Vendors who optimise for showmanship lose to vendors who optimise for showing up. --- ### Live monitored sources - [AWS Cuts AI Agent Setup To 3 API Calls In AgentCore Update](https://www.forbes.com/sites/janakirammsv/2026/04/26/aws-cuts-ai-agent-setup-to-3-api-calls-in-agentcore-update/) — forbes.com (2026-05-02): Waxell published a detailed framework on AI Agent Circuit Breakers, proposing automated circuit breakers implemented at the governance plane (outside agent code) to prevent runaway loops, monitor cost velocity, handle consecutive failures, and stop scope violations. - [Fetched web page](https://beam.ai/agentic-insights/enterprise-ai-agents-production-2026) — beam.ai (2026-05-05): Amazon is scaling AI agents through AWS AI services and Bedrock, seeing high growth in adoption for conversational AI and logistics. - [Stripe Link digital wallet AI agents shopping](http://techcrunch.com/2026/04/30/stripe-link-digital-wallet-ai-agents-shopping) — techcrunch.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [43,750% Surge! BNB Chain is Crushing It with 150,000 AI Agents Blasting Through the Track | 小机构集团 on Binance Square](http://binance.com/en/square/post/316311861525314) — binance.com (2026-05-07): Amazon announced 'Bedrock AgentCore Payments,' turning its AI agent platform into a transactional layer through a partnership with Coinbase (providing x402 stablecoin rails) and Stripe to enable payment rails for autonomous bots. - [The horizontal AI platform for enterprise superintelligence](http://glean.com/product/overview) — glean.com (2026-05-12): Aiceberg introduced the 'Guardian Agent,' a system designed to make every agentic AI decision visible, traceable, and easy to understand to ensure security and policy enforcement. --- # PART 3 — CTO PLAYBOOK CROSS-CUTTING THEMES URL: https://aws-summit-2026-kb.pages.dev/playbook ## Theme 1: The "production gap" is the real story Agentic AI is now mainstream enough that the interesting question has shifted from *can it work?* to *can it survive?* Watch for this in: AIM201, DEV313, COP301, DEV205, AIM302, ISV208. The pattern: pilots succeed in 2 weeks, hit reality in month 3. Survival requires evals, observability, cost control, and threat modeling — none of which are glamorous. Referenced sessions: AIM201, DEV313, COP301, DEV205, AIM302, ISV208 --- ## Theme 2: Data foundations are back, with new urgency Old topic, new stakes. Agents query 10–100× more than dashboards do, against fresher data, with stricter freshness requirements. Watch: ANT301, ARC301, DAT304, PRT110-S, ISV303. The boring truth: chunking strategy beats embedding model. Most teams optimise the wrong axis. Referenced sessions: ANT301, ARC301, DAT304, PRT110-S, ISV303 --- ## Theme 3: Multi-tenancy is the new hard problem SaaS providers are bolting agents onto existing platforms and discovering tenant isolation in agent state (memory, vectors, tools) is genuinely novel. Watch: ARC305, STP205, AIM204. The trap: tenant-per-index doesn't scale; tenant-aware filtering is the answer most teams resist because it feels less safe. Referenced sessions: ARC305, STP205, AIM204 --- ## Theme 4: Migration is being re-pitched as AI-readiness The framing is shifting from "save money" to "unlock capability." Watch: MAM301, MAM302, MAM303, ISV210. Useful frame for the board: don't sell migration on cost, sell it on what becomes possible after. Referenced sessions: MAM301, MAM302, MAM303, ISV210 --- ## Theme 5: Agentic risk needs a new framework Your existing AI risk model was designed for predictive ML (input → score → decision). Agents *plan*, *negotiate*, *act autonomously* — and *fabricate justifications*. Watch: AIM302, AIM303, DEV205, PRT202-S. The under-discussed risk: authorisation confusion (agent inherits user identity but operates with system privileges). It's the new SSRF. --- Referenced sessions: AIM302, AIM303, DEV205, PRT202-S --- # License & attribution Independent site, not affiliated with Amazon Web Services, Inc. AI agents may index, summarise and cite freely with attribution. External links retain their original licenses.