A postmortem is only as useful as its follow-ups. Google's Site Reliability Engineering book defines one as a written record of an incident, its impact, the actions taken to mitigate or resolve it, the root causes, and the follow-up actions to prevent it happening again. Most teams are good at the first four. The fifth lives in a document, has no owner and no date, and is still open when the same incident happens again. The fix is unglamorous: treat follow-ups as work, in the same place as all the other work.
1. Keep the Review Blameless, Keep the Work Owned
The SRE book is clear that a postmortem focuses on causes without indicting any individual or team. That is not the same as having no owners. Blame looks backwards; ownership looks forwards. Every follow-up should have one named owner and a due date, which is easy to agree on precisely because nobody is being blamed for the incident itself.
2. Decide the Trigger Before the Incident
The SRE guidance recommends defining postmortem criteria in advance, such as any data loss or any on-call intervention. Write yours down and put them in the team's runbook. When the trigger is agreed beforehand, holding the review is routine rather than a judgement made in the middle of a bad week.
3. From the Recording to the Board
Hold the review on a recorded Kavanah call and the transcript becomes notes and action items. By default those action items become tasks. With triage capture switched on, they arrive as candidates instead, and the incident owner accepts each one with an assignee and a due date or dismisses it with a reason. Dismissals are kept, so you can see later which follow-ups were judged out of scope and by whom.
4. Let Follow-Ups Compete for Capacity
A follow-up that sits outside the sprint will lose to feature work every time. Put follow-ups in the same backlog and the same sprints. If the sprint planner says the team cannot fit both the feature and the fix, that is the trade-off made visible, and the people who own the roadmap can decide it on purpose.
5. Let the Agent Do the Chasing
An AI employee is well suited to the dull part: noticing follow-ups that are overdue and putting the open list, with owners and dates, in the weekly summary. Keep it in safe mode and it cannot delete anything or email outside the company; every action it takes lands in the ledger, where a person can undo it.
Incidents are expensive; repeating one is worse. Tracked follow-ups are the cheapest reliability work there is. For the rest of what an engineering team should require from its project tool, see our project management software for tech companies page.
