# PagerDuty introduces Paige to help investigate incidents

> The agent can bring incident evidence and suggested actions together; deeper investigation tools are experimental.

By BIGCHANGE.AI Editorial

Published: 2026-11-03T16:57:38.520Z
Updated: 2026-11-03T16:57:38.520Z
Canonical: https://bigchange.ai/blog/pagerduty-paige-sre-incident-agent

![Conceptual charcoal illustration of one person inspecting unlabelled trace-like strokes on a desktop monitor, with a short section accented in orange.](https://bigchange.ai/api/media/file/agent-run-diagnosis-inspection-hero-v1.png)
AI-generated conceptual illustration by BIG CHANGE; no actual agent run, customer trace, failure finding or repair result is depicted.

PagerDuty introduced Paige on November 3 as an AI agent intended to help operations teams investigate incidents.

## The big change

- Paige brings incident context and proposed responses into the responder’s workflow.
- Teams set what it can access and may approve suggested actions.
- Deep Investigation and SRE Workbench are experiments in PD Labs preview.

[PagerDuty says](https://www.pagerduty.com/blog/ai/meet-paige-sre-agent/) Paige combines alerts and telemetry with past incident context to suggest fixes. A team can review a proposed runbook action before approval. The company describes autonomous handling of defined incidents as coming soon, rather than an available general capability.

In the experimental [SRE Workbench](https://docs.pagerduty.com/ai-automation/pd-labs/workbench-user-guide), responders can ask Paige to investigate, add context and steer its work. The guide describes a shared timeline, findings and supporting evidence, plus action review and approval. As an incident develops, the team can add information and consider whether the explanation still fits.

The [Paige documentation](https://docs.pagerduty.com/ai-automation/advance/sre-agent) says console access requires both PagerDuty AI and AIOps; Slack or incident-page access requires PagerDuty AI. Requests in Slack or the Operations Console use four AI Actions. Plan and usage terms therefore matter, while Labs features may change.

A separate [Incident-Arena preprint](https://arxiv.org/abs/2610.00648) evaluated agents on 20 fault-injected tasks using deployed open-source applications in temporary Kubernetes clusters. Researchers reported diagnosis mistakes, incomplete repairs and unsafe regressions. The benchmark checks whether repairs work and avoid unsafe regressions. It did not evaluate Paige.

In a different integration, [AWS describes](https://aws.amazon.com/blogs/devops/scaling-autonomous-operations-with-aws-devops-agent-and-servicenow/) limiting its DevOps Agent’s ServiceNow tool access. [Earlier BIG CHANGE coverage](https://bigchange.ai/blog/servicenow-chatgpt-mcp-plugin-workflow-permissions) discussed the related ServiceNow developments. The AWS integration uses tool-level access controls.

## Sources

- [PagerDuty introduces Paige](https://www.pagerduty.com/blog/ai/meet-paige-sre-agent/) — Company announcement, capabilities and availability.
- [SRE Workbench user guide](https://docs.pagerduty.com/ai-automation/pd-labs/workbench-user-guide) — Preview workflow, evidence and action review.
- [Incident-Arena research](https://arxiv.org/abs/2610.00648) — Separate benchmark research on incident-repair agents.
- [AWS DevOps Agent and ServiceNow](https://aws.amazon.com/blogs/devops/scaling-autonomous-operations-with-aws-devops-agent-and-servicenow/) — Separate integration and access-control context.
- [Paige access and usage guide](https://docs.pagerduty.com/ai-automation/advance/sre-agent) — Access prerequisites and AI Action usage.