PagerDuty launches Paige to investigate outages, with autonomous fixes still to come

An on-call engineering team can now put an AI responder into its escalation chain to gather evidence, recommend fixes and execute approved workflows. PagerDuty launched Paige on November 3, 2026, alongside its renamed PD Reliability Platform. The promise of handling familiar incidents without waking a person, however, remains a coming-soon feature.

Paige is a site reliability engineering agent: software designed to help the people responsible for keeping services running. PagerDuty says it draws on more than 16 years of operational data and works across more than 750 integrations, including Slack and Microsoft Teams. Its role spans sorting alerts, finding owners, coordinating responders and preparing the record after an incident.

From an alert to an approved fix

As a virtual responder, Paige can be added to an escalation policy—the rules that determine who gets called when something breaks. It combines current monitoring data with past incidents, uses documentation to identify the right owners and posts summaries before responders join. Supporting agents handle distinct jobs: Scribe keeps records, Shift manages scheduling and Insights supplies proactive recommendations.

For remediation, PagerDuty describes Paige as adapting fixes that worked in similar incidents. It can run available runbooks, the predefined instructions for tasks such as restarting a service or repairing a Kubernetes deployment. It can also apply an approved resolution and report whether it worked. Teams configure what it can see and do for each user group, and can override recommendations.

That division is important to the launch’s scope. Paige and the platform are generally available, according to SiliconANGLE, but PagerDuty’s launch post explicitly labels autonomous resolution of team-defined, well-understood incidents as coming soon. For unfamiliar incidents, the company says human judgment remains necessary. Teams also retain responsibility for communications with executives and customers, even when Paige drafts the messages.

Three experiments behind the launch

PagerDuty also opened PD Labs, where design partners get early access to work still in development and direct contact with its engineers. Three initial previews extend the incident workflow, rather than representing finished capabilities available on the same terms as the core agent:

• Topology builds a live map of services from code, infrastructure and runbooks. It is designed to show how widely a problem could spread without requiring a configuration management database.
• Deep Investigation tests several possible root causes at once, ranks them by confidence and explains which possibilities it ruled out. PagerDuty says responders can inspect its reasoning as evidence arrives.
• SRE Workbench gives humans and Paige a shared investigation space using the same data. Responders choose whether to investigate further, apply a suggested fix or take another path.

Outside those previews, the renamed platform gets a new console that brings incidents, on-call rotations and AI agents into one view. It shows which agents are enabled and what they are doing, with suggested fixes actionable from that screen. Responders can also manage incidents in Slack or Teams, where a trigger creates a dedicated channel and conference bridge.

The memory needs access to the work

PagerDuty’s longer-term pitch rests on shared memory. It says resolutions and human corrections feed back into organizational knowledge, while Paige reads Jira tickets, monitoring data and Slack conversations to identify patterns. Through MCP, a connection mechanism for AI tools, it can connect to Claude or Cursor and warn about proposed code changes resembling those behind earlier incidents.

That design requires more than switching on an agent. PagerDuty says Paige needs access to the resources engineers use, including ownership records, escalation rules and runbooks. It warns against both broad system access without operational context and such narrow access that people still do the work. Its recommended setup pairs full context with permissions matched to the team’s comfort level.

The PD Reliability Platform replaces the PagerDuty Operations Cloud name. Its pricing is tied to operational scale, with an initial allotment of ingested events, automation actions and user seats. Customers can purchase flexible credits for additional usage.