Postmortems

A postmortem is a structured retrospective of an incident: what happened, who was affected, why it happened, and what should change. BlameTrail drafts a postmortem with AI as soon as an incident resolves, then lets you edit and publish it.

What's in a postmortem

Each postmortem has the following sections, all editable before publishing:

• Summary — a one-paragraph overview of the incident
• Severity — SEV-1 (critical) through SEV-4 (minor)
• Customer impact — description, duration in minutes, and affected services
• Timeline — chronological events from monitor checks, deploys, alerts, observability data, and human updates
• Root cause — the underlying technical or process failure
• Contributing factors — secondary issues that worsened or prolonged the incident
• What went well and What went poorly — blameless retrospective notes
• Action items — follow-up tasks with priority (high / medium / low) and a suggested owner
• Confidence — the AI's confidence in its analysis (high / medium / low)

Plan availability

Regenerating an existing postmortem within the same calendar month does not consume additional quota — only newly created postmortems count toward your limit.

Lifecycle

1. Auto-trigger — when an incident transitions to resolved (either automatically by the monitoring engine or manually by a user), a postmortem is queued for generation.
2. Generating — an AI draft is produced using the same context the incident summary uses (recent deploys, suspect commits, observability metrics and logs, human updates posted to the incident timeline). Generation typically completes in under a minute.
3. Draft — the postmortem is ready for human review on the incident detail page.
4. Edited — any teammate with edit permissions can update content; status changes to "Edited" while preserving the original AI draft for reference.
5. Published — an admin or owner publishes the postmortem, freezing it in its current state.

You can regenerate at any time before publishing — useful if you've added new information to the incident timeline since the original draft. Regeneration is rate-limited to once per minute.

Generating, editing, and publishing

1. Open the incident detail page for any resolved incident.
2. Scroll to the Postmortem card.
3. If no postmortem exists yet, click Generate postmortem.
4. Once the draft is ready, click Edit to revise any section. Save when you're happy with the result.
5. Click Regenerate if you'd like the AI to take another pass — for instance, after adding a teammate's update to the incident timeline.
6. Click Publish when the postmortem is ready to share. Published postmortems can still be edited, but the status remains "Published" rather than reverting to "Edited".

Permissions

Tips for better drafts

• Post updates to the incident timeline as the incident unfolds. Updates are included in the postmortem's user prompt, and they are the single biggest signal the AI uses to ground the timeline section.
• Connect an observability provider (Prometheus, Loki, Datadog, or CloudWatch). Metrics and logs around the incident window appear in the timeline and root cause sections when available.
• Resolve incidents promptly when remediation is complete. The auto-generated draft uses the resolution time as the end of the incident window.
• Regenerate after adding new information rather than editing every section by hand. The AI handles structural rewrites well.

Disabling AI features

Owners can disable AI features for the entire organization in Organization → Settings. With AI disabled, the Generate postmortem button is hidden and no postmortems are queued automatically.