PagerDuty launches Paige, an AI agent that answers the page

Overview

Updated Yesterday

For 17 years, PagerDuty's product did one thing well: when software broke, it woke someone up. On November 3, 2026, the company started selling an agent that responds to the page itself.

Paige investigates incidents, proposes fixes, and can run them with or without a human approving each step. The launch also renames PagerDuty's operations cloud the PD Reliability Platform and opens PD Labs, a program where customers test unfinished AI features.

Why it matters

If Paige works as advertised, on-call engineers spend less time diagnosing and more time approving an agent's proposed fixes.

Questions about this story

Free account needed to ask — your question is kept and asked for you right after sign-up. Answers are public.

No questions yet — be the first to ask.

Key Indicators

95%

Executives who call incident reduction a competitive advantage

Figure PagerDuty cites from recent research to justify resilience spending.

77%

Executives planning to raise resilience investment

Share who say they will increase spending over the next 12 months.

Steps in Paige's incident workflow

Connect, detect, mobilize, diagnose, remediate, learn.

Voices

Curated perspectives — historical figures and your fellow readers.

People Involved

Jennifer Tejada

Chief executive of PagerDuty

Leads PagerDuty through the Paige launch

Organizations Involved

PagerDuty

Public company

Launched Paige and renamed its core product the PD Reliability Platform

San Francisco incident-management company whose software alerts engineers when production systems fail.

PD Labs

AI innovation center

Newly launched with design partners

A PagerDuty program that gives customers early access to AI features under development.

Timeline

2009 November 2026

3 events Latest: Yesterday

November 2026
Paige, PD Reliability Platform, and PD Labs launch
Latest Product launch

PagerDuty releases its SRE agent, renames its operations cloud, and opens an AI innovation center.
Yesterday
April 2019
PagerDuty goes public
Company

Shares begin trading on the New York Stock Exchange under ticker PD.
April 11th, 2019
2009
PagerDuty founded
Company

The company starts in San Francisco to automate on-call alerting.
2009

Scenarios

Paige resolves most incidents without a human approving each fix

Possible Resolves by Nov 3, 2027

Discussed by: PagerDuty product materials and analysts covering AI operations

PagerDuty designed Paige to act autonomously in well-understood cases, and says every resolved incident feeds a shared memory. If customers grant that permission widely, incident response becomes a review queue rather than a live firefight.

Customers keep Paige on a short leash

Likely Resolves by Nov 3, 2027

Discussed by: Reliability engineers and industry surveys on AI adoption

Remediation is the riskiest step. PagerDuty defaults to human approval and says agent permissions are still planned for early access, so many teams may run Paige as an investigator that never touches production.

A rival ships a competing SRE agent

Likely Resolves by May 3, 2027

Discussed by: Enterprise software analysts

Observability and service-management vendors already sell into the same budgets. A generally available agent from a large competitor would turn Paige from a category creator into one option among several.

A Paige remediation causes a public outage

Unlikely Resolves by Nov 3, 2028

Discussed by: Reliability engineers and post-incident reviewers

Automated fixes can make an outage worse. PagerDuty's guardrails, including human approval and tested releases, are meant to prevent that. A public failure would test whether customers still trust autonomous remediation.

Historical Context

3 moments from history that rhyme with this story — and how they unfolded.

August 2012

Knight Capital trading failure (2012)

Knight Capital deployed new trading software but left old code running on one of eight servers. When markets opened on August 1, the firm sent millions of erroneous orders. It lost about $440 million in roughly 45 minutes.

Then

Knight's stock collapsed, and the firm was acquired by Getco the next year.

Now

The Securities and Exchange Commission fined Knight $12 million and regulators tightened rules for automated trading systems.

Why this matters now

Paige automates production changes rather than trades, but the failure mode is the same: software acting at machine speed needs limits before it acts.

May 2010

The Flash Crash (2010)

On May 6, 2010, the Dow Jones Industrial Average fell about 9% in minutes before recovering. Regulators later pointed to a large sell algorithm interacting with high-frequency traders.

Then

Exchanges introduced circuit breakers to pause trading during sudden moves.

Now

Automated market controls became standard, built on the premise that fast machines need fast brakes.

Why this matters now

The debate over how much autonomy to give Paige echoes the question regulators faced in 2010: how fast should automation be allowed to act without a person in the loop?

2011

Netflix Chaos Monkey (2011)

Netflix released software that randomly terminated production servers so engineers would build systems that survive failure. The practice became known as chaos engineering.

Then

Netflix teams designed services to tolerate constant, expected failure.

Now

Controlled failure testing spread across the industry as a way to earn trust in automation.

Why this matters now

PagerDuty's pitch is similar: let an agent act, but only after testing and guardrails show it can be trusted.