Operations Manager
An incident scribe who builds the timeline while you fight the fire
Feed it the logs and it reconstructs the timeline, drafts the postmortem, and tracks the action items until they're actually done.
During an incident, nobody writes anything down, which is why postmortems get reconstructed a week later from fading memory and ten thousand lines of unread Slack scrollback. Give your agent the raw material instead, paste log excerpts into chat, drop files into its workspace, point it at status pages it can fetch, and it builds the timeline while the incident is still warm: first anomaly, detection, escalation, each intervention and its effect, resolution.
How it works
- 01
When the dust settles, the agent drafts the postmortem from that timeline: impact and duration, the contributing factors as evidenced by the logs rather than as remembered by the most senior person in the room, what worked in the response, and where the gaps were. The draft lands as a desk document for the team to correct and approve; the agent keeps it honest by citing which log line supports which claim.
- 02
Then comes the part every postmortem process fails at: the action items go onto a desk checklist with owners and dates, and the agent follows up on a schedule until each one is closed or explicitly dropped. Three incidents later, its memory starts paying out, "the connection-pool exhaustion in this one looks like the March incident, and that action item is still open" is the sentence that justifies the hire.