Post-Incident Review
What This Requires
Conduct post-incident review within 5 business days of all major AI incidents. Analyze root cause, identify contributing factors, define corrective actions, and share lessons learned. Track corrective action completion.
Why It Matters
Incidents are learning opportunities. Post-incident reviews prevent recurrence and build organizational resilience.
How To Implement
Trigger Criteria
Define major incident: SEV1/SEV2, data breach, bias complaint, regulatory inquiry, >4 hour outage, model failure requiring rollback.
Review Process
Schedule within 5 days. Attendees: incident responders, stakeholders, governance lead. Use template: (1) Timeline, (2) Root Cause (5 Whys), (3) Contributing Factors, (4) What Went Well, (5) Corrective Actions.
Corrective Actions
Define specific, actionable improvements: fix code bug, update runbook, add monitoring, conduct training. Assign owner and deadline. Track in Jira/ServiceNow.
Lessons Learned
Publish sanitized summary to internal wiki. Share in all-hands or engineering meeting. Add to training materials if applicable.
Evidence & Audit
- Post-incident review template
- Completed reviews for recent major incidents
- Root cause analysis documentation
- Corrective action tracking (Jira, ServiceNow)
- Lessons learned shared with team (wiki, meeting notes)