Agentic SRE Advanced By Samson Tanimawo, PhD Published Apr 5, 2026 5 min read

Multi-Agent Workflows for Postmortem Generation

One agent gathers data. One writes. One reviews. One files. The workflow, with the inter-agent messages typed and bounded.

The pipeline

Gatherer: pulls timeline, metrics, deploy logs, action history. Read-only; produces a structured incident object.

Writer: drafts the postmortem text from the incident object. Produces sections 1-4 of the standard template.

Reviewer: reads the draft, identifies gaps, asks clarifying questions. Edits where appropriate; flags where humans must intervene.

Filer: posts the final draft to the postmortem store; notifies the team; opens action-item tickets.

Gatherer → Writer: the incident object. Schema: timeline, metrics summary, services affected, actions taken, duration.

Writer → Reviewer: the draft. Schema: text with section markers, confidence per section, gaps identified.

Reviewer → Filer: the reviewed draft. Schema: text, action items, owner suggestions, follow-up dates.

Each message is typed and validated. Schema mismatches halt the pipeline.

Each message has a size cap. The gatherer's incident object is at most 10k tokens; the writer's draft is at most 5k tokens.

Caps prevent runaway context growth. Without caps, the pipeline accumulates state and slows down.

When a message would exceed the cap, the producer summarises. The summary is part of the API; the agent has to be good at producing it.

Gaps the writer flagged. Sections it could not draft confidently get filled by humans.

Action items the reviewer suggested. Humans pick which to actually file and assign owners.

Final approval before filing. The filer does not auto-publish; a human approves.

Solo human: 4-6 hours from incident close to filed postmortem.

Pipeline: 90 minutes (50 minutes of agent work + 40 minutes of human review).

The 4x speedup is real and survives quality review. Drafts are not perfect but they are starting points; humans finish.