PagerDuty launches SRE agent Paige and rebrands its reliability platform

PagerDuty has launched Paige, an AI agent for incident response, and is presenting its reliability platform under a new name, PD Reliability Platform, according to a 3 November StreetInsider report. Framing the product launch and the rebrand together fits a broader repositioning of the company around an AI-first reliability stack rather than a single feature update.

On its product page, PagerDuty calls Paige an SRE Agent and an “agentic on-call teammate”. The claim is that the agent is already on call when an incident fires, rather than being summoned after the fact, and that it pulls the right people and other agents into the response.

The workflow is split into six stages: connect, detect, mobilize, diagnose, remediate and learn. Paige plugs into development and observability tools, gathers change events, documentation, runbooks, logs, metrics and topology, then tests several hypotheses in parallel while showing its reasoning at each step.

Remediation is the part that deserves scrutiny. Paige recommends the fix that best fits the incident, with the reasoning and evidence attached, and every action waits for human approval unless a team explicitly lets it act on its own. PagerDuty says the agent is governed by the same trust layer and permissions as the rest of the platform, and that its AI capabilities are reviewed by security, legal, data science and product leaders and tested against prompt injection.

Paige also coordinates a small team of specialised agents: Shift Agent handles schedules and escalation paths, Scribe Agent captures key moments and communications, and Insights Agent surfaces patterns from operational history. Paige acts as the orchestrator across a single incident.

The agent meets responders in Slack and Microsoft Teams, and reaches AI tools such as Claude and Cursor through MCP. PagerDuty says Paige ships with the PD Reliability Platform or PagerDuty Advance, and that setup takes three steps: connect the tools, add Paige to an escalation policy or incident workflow, and set approval rules. The vendor puts the platform at more than 750 integrations and 900 million incidents managed at 99.99% average reliability.

The pitch is that Paige does not stop at diagnosis: it proposes a fix, shows the evidence behind it, and can execute it once a human approves. Incident response is one of the few agent use cases with high frequency, a clear success metric and real cost pressure, which is why AI SRE has become a crowded entry point. The next things to watch are pricing and packaging, whether customers actually grant autonomous execution, and verifiable reductions in mean time to repair rather than another polished demo.