Turn your incident notes into a blameless postmortem first draft that marks everything it inferred, so you know exactly what to check.
Blameless by design: systems and gaps, not people.
A full worked example: a 92-minute database outage, written up end to end.
Action items typed as prevent, detect or mitigate.
6 SECTIONS / 1 PROMPT
What broke, for how long, and who felt it, with numbers wherever your notes have them. Anything it had to guess is flagged.
Your channel log and scratch notes, put in order. Anything it filled in rather than read is marked as inferred.
The underlying failure rather than the first symptom, with the open questions called out so you know what still needs proving.
The gaps that let it happen or slowed fix: missing alerts, missing runbooks, wrong first moves.
The calls that kept it from getting worse, written down so they become practice instead of luck. Credit goes to your decisions.
Each one typed as prevent, detect or mitigate, so the list covers more than the one fix everyone already agreed on.
On-call engineers who owe a postmortem after a long night
SRE and DevOps teams that run blameless reviews
Anyone moving into ops who wants a solid postmortem structure
Anyone hoping AI will find the root cause for them
Teams looking for someone to blame
Inferred: the notes do not give a request-failure count or an affected-customer figure. Both should be added before this is circulated.
FROM THE WORKED EXAMPLE
IMPACT SECTION · 92-MINUTE OUTAGE
One email to confirm, and the prompt opens straight away. Then five short emails on using AI in ops work, and one note a week after that.