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.
What this session is about
Playbook
Editorial commentary · what to actually do about this on Monday
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.
---
Independent editorial perspective — not an official AWS or speaker statement. Designed for executives evaluating what to brief their teams on next.
Live updates related to this session LIVE
Sourced via Parallel AI Monitor — continuous web watch on 21 topical streams. Updated .
- businesswire.com high confidence Multi-agent collaboration patterns
ZoomInfo Introduces Agent Teams, Powered by the ...
ZoomInfo announced Agent Teams, a multi-agent orchestration product built on DoubleO.ai. Specialized agents have distinct roles, instructions, and tools; they share context from prior steps and use real-time evaluation of the current record state to choose the next action. The us
- arxiv.org Multi-agent collaboration patterns
FlowMAS: Learning Multi-Agent Workflow Topology via ...
FlowMAS is a framework for automatically learning multi-agent workflow topology using an information-guided Generative Flow Network. Its curiosity-driven component explores structurally novel workflows, while its information-guided component favors collaboration patterns that con
- arxiv.org Multi-agent collaboration patterns
RepoMAS: Solving Progressively Specified Tasks with ...
FlowMAS is a framework for automatically learning multi-agent workflow topology using an information-guided Generative Flow Network. Its curiosity-driven component explores structurally novel workflows, while its information-guided component favors collaboration patterns that con
- github.blog high confidence Autonomous coding agents
Agentic autofix now uses Copilot Memory - GitHub Changelog
GitHub’s agentic autofix now uses Copilot Memory for customers who enable it. The security-fix agent can use existing repository memories for context and save successful fix patterns for future alerts. These memories can also inform Copilot code review and Copilot cloud agent. Bo
- ovaledge.com Decision lineage & audit trails for agents
Governed context observability for AI agents explained
Google Cloud’s Gemini Enterprise Agent Platform observability documentation describes using OpenTelemetry telemetry to monitor agents and MCP servers through metrics, traces, and logs. It covers execution paths, multi-agent system visibility, and monitoring policy interceptions i
External links matched to this session via topic relevance. The KB does not endorse third-party content; verify before citing.