PagerDuty launches Paige, an SRE agent to triage and fix incidents alongside humans — and learn from each one

PagerDuty has given its incident platform an AI teammate. The company announced Paige on November 3, 2026, a next-generation site reliability engineering (SRE) agent built to work as a member of operations, engineering and DevOps teams, detecting, investigating and remediating issues across increasingly complex systems. Paige orchestrates incident resolution on the PD Reliability Platform, PagerDuty’s flagship product, previously known as PagerDuty Operations Cloud.

The timing is the argument. Software is increasingly written, tested and released by AI with ever-diminishing human review before release, PagerDuty says, and production AI fails in new ways: model drift, token runaway, governance violations, PII exposure and hallucinations. Research the company cites puts numbers on the pressure: 95% of executives say the ability to reduce incidents is now a competitive advantage, with 77% planning to increase resilience investment over the next 12 months. Paige is PagerDuty’s answer to that pressure — not a dashboard, but an agent inside the response itself.

What Paige actually does

Paige coordinates signal intelligence, incident management and automation, so detection, investigation and remediation run as one workflow rather than three handoffs. It is designed for optionality, PagerDuty says: teams can run Paige with a human in the loop, accelerating triage while reserving remediation for approval, or let Paige act autonomously in well-understood scenarios, resolving lower-priority incidents automatically while keeping full visibility for its human operators. Independent coverage of the launch describes the same two postures — automated remediation for standard, well-documented disruptions, or triage-first with a human keeping the final call.

What PagerDuty says distinguishes Paige most is a capacity to learn. Every resolved incident becomes part of a shared memory, sharpening Paige’s future performance and making each subsequent incident easier to resolve. The new PD Reliability Platform Console gives incident commanders and team managers a single view of incidents, on-call rotations and AI agents — which agents are enabled and what they are doing — and lets teams act directly on Paige’s recommendations without leaving the screen. The console is available exclusively to PD Reliability Platform customers.

Where teams meet it: Slack and Teams

The platform is Slack- and Microsoft Teams-native, so responders can run incidents end to end without leaving the tools they already use. When an incident triggers, PagerDuty automatically spins up a dedicated channel and conference bridge. Responders acknowledge with one click, then collaborate with teammates and Paige to resolve it. Once resolved, a post-incident review is generated instantly, with follow-up actions pre-populated with owners and due dates, so longer-term remediation does not go unresolved.

That is also where PagerDuty’s scale claim sits. The company says it invests more than $100 million in research and development annually, manages approaching one billion incidents a year, is trusted by two-thirds of the Fortune 100 and tens of thousands of customers globally, and orchestrates workflows with context from more than 750 integrations. The announcement frames that history as the reason to trust an agent with production incidents — experience at a scale few incident platforms can match. Whether that experience transfers into Paige’s judgement is the claim the product now has to prove in customers’ hands.

PD Labs, a benchmark claim, and the price of entry

Alongside Paige, PagerDuty is launching PD Labs, a new AI innovation centre built in the open with design partners. It gives organisations hands-on access to what PagerDuty’s engineers are working on, the ability to try new capabilities against real problems, and a direct line to the people writing the code. The SRE Workbench, a collaborative investigation space for human responders and Paige working from the same data, is released concurrently on the PD Labs incubation platform at pagerduty.com/pd-labs.

PagerDuty also points to an SRE Agent Benchmark Report in which, it says, Paige outperformed leading AI coding agents at diagnosing simulated IT incidents, and says the report explores the methodology behind the results. The announcement gives no scores, and simulated incidents are not production incidents: treat the benchmark as a company-run test of diagnostic skill, not independent proof that Paige will diagnose a live outage better than the alternatives.

Pricing follows the platform model. The PD Reliability Platform carries a platform fee based on operational scale that includes a seeded amount of events ingested for signal intelligence, actions for AI and automation, and seats for responders, managers and stakeholders, with flexible credits to add events, actions and seats as teams go. Customers can allocate their investment between people and AI-driven actions rather than committing to seats alone. Customers who buy an Essential, Plus or Ultimate plan before January 15, 2027 will receive additional AI Actions capacity at no extra cost for the first year. No headline price was disclosed.

What to watch

Everything in the announcement is PagerDuty’s account, and the launch names no production customers using Paige itself. Autonomous action is scoped to well-understood scenarios, with human approval remaining the posture for the rest — the sensible scope for a first release, and the one to watch as PagerDuty widens it. Two questions will decide whether Paige matters beyond the launch: does shared memory demonstrably shorten repeat incidents, and do PD Labs prototypes graduate into the platform, or stay demos? The 95% and 77% figures are cited research about the market, not results from Paige. On the evidence published on November 3, Paige is a credible, carefully scoped bet: an agent that joins the incident queue rather than replacing the people in it, with its boldest claim — that it gets better with every incident it resolves — still to be shown.