Citation chains
A postmortem is a load-bearing document — it’s how a team agrees on what happened and what changed because of it. If an AI draft invents a participant or shifts a timestamp by ten minutes, it isn’t a postmortem any more. IncidentScribe’s answer is the citation chain: a verifiable path from any sentence in the draft back to the original log line that justifies it.
How it works
Each drafted section carries metadata about which timeline events it cites. Each timeline event carries metadata about which slice of the source — a Slack message, a log line, a PagerDuty payload — it was extracted from. Click any drafted claim and the Citations inspector opens with the cited events; click an event and the source pane opens to that slice with the actor’s name highlighted in the original text. Two clicks, no interpretation in between.
Stable IDs, not paraphrases
The links are stored as stable IDs rather than restated text. The claim that cites event 14, slice 7 today will still resolve to event 14, slice 7 in six months, when a new on-call engineer pulls up the postmortem trying to reconstruct what happened. Provenance that decays isn’t provenance.
Broken chains find you
When a section is drafted against too-sparse evidence — or no cited events at all — it gets a visible low-confidence chip next to the header. You see it in the draft UI before you sign off. You don’t have to hunt for the weak section; the broken chain surfaces itself so you know exactly where to fact-check harder.
Why it’s reliable
The chain is an architectural property, not a UI varnish. The drafter that writes the prose never sees the raw incident text — only the validated, reconciled timeline. It literally cannot make a claim that isn’t grounded in a cited event, because cited events are the only thing it’s given. The Timeline section itself is rendered deterministically from the structured events, never written as free prose.
That’s the difference between “trust the AI” and “verify the chain.” For a deeper treatment, read Citation chains: the trust UI for AI-drafted postmortems.