What you build
Incident artifacts under pressure:
- A timeline of detection, mitigation, and recovery—with evidence links.
- Customer-facing or internal status language that is accurate and calm.
- A postmortem with root cause, contributing factors, and tracked action items—without blame theater.
Agents help organize and draft; humans decide severity, comms, and when to page.
Why CoWork OS is a strong fit
- Multi-channel inputs (Slack, email, ticketing) can feed one narrative if you wire them carefully.
- Approvals before sending external messages or closing incidents.
- Skills can encode your org’s postmortem template and severity definitions.
How to use
- Stabilize first—do not let drafting slow mitigation.
- Capture raw facts with timestamps as you go (pasted logs, messages).
- After calm returns, build the timeline from those facts only.
- Separate facts from hypotheses in the writeup.
- Review customer-facing text with someone who owns comms.
Prerequisites
- Severity rubric your team already uses (or agrees to adopt).
- Access to logs and tickets agents may summarize—within policy.
- A template for postmortems and status updates.
Steps
- Open a single working doc for the incident; avoid scattered threads as the source of truth.
- Record customer impact and scope as soon as known.
- Draft timeline bullets; attach links to evidence.
- Run a blameless review; focus on systems and process.
- File action items with owners and dates; track to completion.
Suggested prompts
- “From these messages, produce a chronological timeline with unknowns marked.”
- “Draft a status update for non-technical readers; no acronyms without expansion.”
- “List five contributing factors using the template’s categories.”
Never publish automated comms without a human who owns the customer relationship.
Launch readiness
- Timeline matches log evidence; no invented precision.
- Stakeholders sign off on external messaging.
- Action items have owners—not a vague “we should.”
Common pitfalls
- Premature root cause before data supports it.
- Leaking internals or customer data into public updates.
- Copy-paste from chat without verifying timestamps.
- Skipping the postmortem because the fire is “over.”