Operations management software company PagerDuty Inc. today introduced Paige, an artificial intelligence agent for site reliability engineering that is built to work as a member of operations and engineering teams.

Paige runs on the company’s flagship product, which as of today is called the PD Reliability Platform. Alongside both, PagerDuty opened PD Labs to put early features in front of customers while they are still being built.

The launch is aimed at a problem PagerDuty says is getting worse as AI coding tools write more production software, much of it shipped with little human review. Applications built on AI models also break in their own ways. Chief Executive John DiLullo pointed to model drift, token runaway and leaked personal data as examples. The only way to stay resilient, he said, is “to fight fire with fire, using AI itself as part of how we keep systems running.”

Customers set Paige’s permissions team by team. A team that keeps a person in the loop gets faster triage with remediation held for approval, and the agent acts on its own only in well-understood, low-risk scenarios the team has explicitly authorized. In March, PagerDuty said its SRE Agent would start joining escalation policies as a “virtual responder.” Paige, billed as the company’s next-generation SRE agent, can now be put on call that way. A lower-priority incident may never reach a human at all.

Paige’s recommendations arrive with evidence and a confidence level attached. By the time a responder reads one, many possible causes have been ruled out already. PagerDuty argues the bigger payoff comes later, once a resolved incident has been written into a shared memory that Paige consults whenever something similar fails. In a benchmark report released with the launch, the company says Paige did better than leading AI coding agents at diagnosing simulated information technology incidents.

The renamed platform, known until today as PagerDuty Operations Cloud, is where Paige gets its operational context. Incident commanders and team managers also get a new PD Reliability Platform Console, which puts incidents, on-call rotations and AI agents in one view. The console shows which agents are switched on and what each one is doing. Paige’s suggested fixes can be acted on from the same screen.

Responders can run an incident end to end in Slack or Microsoft Teams. A trigger opens a dedicated channel and a conference bridge without anyone having to set them up.

Steve Miller, director of business development at Slack, said having Paige and the platform inside Slack gives responders speed and visibility, which he called critical as “agentic and human teams operate side-by-side in the same channel.” Nobody has to write the post-incident review, which is generated as soon as the incident is resolved, and its follow-up actions come with owners and due dates already assigned.

Canadian telecommunications company Telus Corp. has wired the platform into its own centralized internal incident tool. Kulvir Gahunia, director of the Site Reliability Office at Telus, said PagerDuty’s Scribe Agent now runs through every incident, so responders “don’t have to stop and document what is happening.”

The earlier-stage work runs through PD Labs, which PagerDuty is building in the open with design partners who get early access and a direct line to the engineers writing the code.

Three experiments are in preview to start. Topology keeps a live map of an organization’s services, assembled from its code, infrastructure and runbooks. No configuration management database is needed, and the map scopes the blast radius of a problem automatically as conditions change.

Deep Investigation has Paige test several root-cause hypotheses at once while showing its reasoning on each as new evidence comes in. SRE Workbench gives people and agents one workspace to tackle a problem side by side, and whatever they find is saved to the same memory Paige learns from.

IT teams want measurable results from their spending on AI agents, according to Snow Tempest, research manager for IT service management at International Data Corp. Better uptime counts, as does time handed back to responders who would otherwise be “chasing logs and assembling the right people for an incident.” More organizations also want to prevent incidents, Tempest said.

Paige and the PD Reliability Platform are generally available today, and SRE Workbench is open to try through PD Labs. The platform is sold for a fee tied to operational scale that comes with a starting allotment of ingested events, automation actions and user seats. Customers who outgrow it can buy flexible credits without having to forecast usage ahead of time.

Image: PagerDuty/GPT Image 2.5