Cristina DiasOctober 7, 2026 | 5 min read This blog post is part of PagerDuty’s ongoing series on how we’re helping customers navigate their journey towards autonomous operations. Read on to learn about how PagerDuty’s AI-Powered Incident Communications builds towards this vision.
The update is always late, and it is always the same update
An incident is live. The fix is not in yet, and the updates are already overdue. Internal teams want to know what is happening. Partners want to know whether they are affected. Customers want to know when it will be over. Those are the same story told three different ways, and each one pulls a responder out of the fix to rebuild the timeline from memory and write it again for a different reader. Twenty minutes later it happens again, and it keeps happening for the life of the incident.
Most incident management tooling treats this as a publishing problem. Give responders a faster editor, a better template, a button closer to the incident. That helps, and it still leaves the work exactly where it was. The context for a status update already exists in the bridge call and the incident channel. What has been missing is an agent that gathers it and turns it into the update.
AI-Powered Incident Communications is now in rolling GA. SRE Agent orchestrates the path from response to published update: it calls Scribe Agent to capture conference bridge audio and incident channel chat, keeps the incident channel updated automatically as the response unfolds, drafts internal and external status updates from that captured context, and surfaces those drafts in Slack for a responder to review and publish. Let’s look at each piece.
Capture the incident without assigning a scribe
On most bridge calls someone draws the short straw, stops investigating, and starts typing. When nobody draws it, the context is simply gone, and the update becomes somebody’s best recollection of a call that ended an hour ago.
Scribe Agent joins the incident bridge and captures both the audio and the incident channel chat, across Zoom, Microsoft Teams, and Google Meet. Live transcription appears in thread, so anyone who is not on the call can still follow along. SRE Agent invokes Scribe Agent as part of coordinating the incident, rather than waiting for a responder to remember.
That running record also drives the Periodic Incident Progress workflow action, which posts structured updates into the incident channel on a cadence your team sets. Internal stakeholders get a drumbeat without anyone being asked for one.
Draft internal and external updates from what was actually said
Updates written from memory go out in whatever shape the writer chose, so the version an executive reads bears little structural resemblance to the one on the public status page an hour later.
Standardized templates map to where you are in the incident, from scoping impact through recovery and resolution, so every update arrives in the same shape regardless of who wrote it. Generate with AI fills the draft from PagerDuty incident context plus what Scribe Agent captured, including the affected service, the customer-visible symptom, and what is not affected. Refine with AI tightens a draft a responder wrote themselves. Alongside the copy, the update carries the structured fields the status page needs:
- Status, from investigating through resolved
- Impacted business services
- Impact level
- Time until the next update
Internal and external status pages are both selectable from the same place, so the stakeholder update and the customer update come from one captured context rather than two separate writing sessions.
Review and publish in the channel you are already in
Approval is usually why updates sit. The person who should read the draft before customers do is in the incident, and the draft is in a doc or a DM that nobody in the response has open.
Running /pd update in the Slack incident channel opens the update directly where the team is working. Checking “post message to channel as draft” puts it in front of the channel first, where the incident commander can edit the message or refine its tone, either in Slack or on the incident details page, and then publish. The first update creates the status page post and sets the customer-facing title. Every update after it threads into that same post, so subscribers follow one story instead of hunting through separate entries.
Bringing it together
Separately these are three capabilities. Together they close the communication stage of the incident lifecycle, the stage that has stayed manual while detection, mobilization, and analysis were automated around it.
The human stays where the human matters. An agent gathers the context and proposes the words, and a responder decides what reaches customers. And none of it is disposable: the captured context and the published updates feed the post-incident review and SRE Agent’s memory, so the next incident of this shape starts further along than this one did.
Paving the path to autonomous operations
Your team shouldn’t be rewriting multiple updates for different audiences while the fix is still open.
Our approach to autonomous operations:
- Puts intelligent agents to work at scale, handling the noise, accelerating resolution, and keeping you in the loop where it matters
- Deepens the full incident management lifecycle by empowering teams to resolve incidents faster
- Broadens the platform and ecosystem with capabilities that help teams prevent incidents from happening
AI-Powered Incident Communications moves us closer to that vision. Every audience hears the same story, and nobody leaves the incident channel to tell it. Ready to see it in action?