The Deploy Postmortem Agent: First Pass at the Writeup

A postmortem is a writing task and a forensics task. The agent that handles the forensics and produces a writeup that is 70% finished, leaving the analysis to humans.

Forensics: what happened

Pull the timeline: alerts that fired, services affected, deploys made, actions taken.

Pull the metrics: errors, latency, traffic. The agent extracts the time window of impact and characterises the drop.

Pull the deploy logs: what was deployed, by whom, with what changes. The agent links deploys to the impact window.

Draft the writeup

Section 1: summary (what happened, when, who was affected, total duration). The agent gets this 90% right because it is mechanical.

Section 2: timeline (the what-happened-when, with action attribution). Mechanical; agent is reliable.

Section 3: contributing factors (what allowed the incident to happen). The agent proposes; humans refine.

Section 4: action items (what to do to prevent recurrence). The agent proposes; humans decide ownership and priority.

70% finished

The agent's draft handles the rote sections (summary, timeline). The human owns the analytical sections (contributing factors, action items).

70% finished means the human starts from a draft, not a blank page. Time-to-postmortem drops from 4 hours to 90 minutes.

Resist drafting the analytical sections. The agent will produce plausible-looking text that misses the actual contributing factors. Save those for humans.

Human review checklist

Verify the timeline: are all major events represented? Did the agent miss something?

Verify the impact: does the duration and scope match what the team experienced?

Refine contributing factors: the agent suggested some; check whether they are right and what is missing.

Set action items: the agent's suggestions are starting points; humans pick the ones that matter and assign owners.

What the agent learns over time

Postmortem templates: each team has its preferences. The agent learns which sections to include and which to skip.

Common contributing factors: "deploy without canary" or "monitoring gap" appear repeatedly. The agent's prompt is updated to surface these.

Action item patterns: the team's culture about ownership, deadlines, follow-up. The agent's drafts adapt over months.