TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
AI-assisted site reliability engineering tools are gaining attention as companies report more deployments and on-call pages. The current focus is faster incident investigation and alert triage; fully autonomous production fixes remain unproven in the report, and claims about effectiveness are difficult to compare across companies.
AI-assisted SRE tools are moving from conference demos into vendor products and some companies’ internal systems, as engineering teams contend with more code changes and rising on-call pages, according to a report by The Pragmatic Engineer. The tools are chiefly being used to investigate incidents and triage alerts; the report says systems that autonomously diagnose and fix production problems without human oversight are not yet a routine reality.
The report traces the latest interest to a May 2025 demonstration at Microsoft’s BUILD conference, where an Azure SRE Agent appeared to investigate an alert, identify a cause and ship a fix. The author questioned whether engineering teams would allow an AI system to make potentially risky production changes without a person reviewing them. The report says attention to the category has since increased, citing OpenAI’s Sevbot, an in-house system being developed at ING, and growing visibility for AI incident-management vendors.
For its reporting, The Pragmatic Engineer spoke with Rootly, incident.io and ResolveAI. It describes the practical focus as speeding up investigation, often within Slack or Microsoft Teams, rather than handing an agent unrestricted control over production. Multiple agents may examine an incident in parallel, while engineers guide them through chat. Some systems also use an internal “context layer”—such as a knowledge graph and stored operational information—to give investigations relevant company-specific detail.
Vendors and engineering teams are trying different ways to improve results, including having separate agents challenge an incident hypothesis, using more than one model, and testing systems against past incidents. But the report cautions that accuracy is difficult to measure consistently: a useful answer depends on each company’s infrastructure and each team’s incident history. Speed, cost and accuracy also involve trade-offs, with techniques such as model routing and shorter prompts used to manage them.
More Deploys, More On-Call Pressure
The appeal is tied to a basic reliability relationship outlined in the report: more changes can mean more opportunities for a change to cause a failure, even if the chance of failure per change stays the same. The author says many companies have seen pull-request frequency at least double over the past year as AI coding agents increase the pace of software development. That is a broad observation in the report, not a universal measurement for every company.
Rootly data cited by the report shows pages rising compared with a year earlier and outside-business-hours pages nearly doubling. The report does not provide enough detail here to treat those figures as an industry-wide rate. Still, more alerts can mean more interruptions for on-call engineers. If AI tools can identify routine issues or suppress unimportant pages, they could reduce that burden; the report says teams should expect the incidents that do reach engineers to remain difficult.
There is also a skills concern. If responders rely too heavily on automated investigation, they may get less practice diagnosing incidents themselves. The report raises incident-response simulations as one way to maintain those skills. It also notes an emerging category of outages involving AI systems, adding new failure modes even as AI tools are used to manage reliability.
Popular right now
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
iRobot Roomba Max 715 Pet Robot Vacuum with AutoEmpty Dock
As an affiliate, we earn on qualifying purchases.
From Stage Demo to Incident Triage
“AI SRE” has become a catch-all label for AI-augmented on-call and incident tools, though the report says the vendors interviewed do not particularly like the term. A site reliability engineer’s work extends well beyond responding to alerts, so the label can overstate what these products do. Alternatives such as automated incident investigation agents have been proposed, but have not caught on as widely.
The report describes two broad ways to reduce incidents: improve testing and verification before changes go live, or use lessons from incidents to make systems more resilient. AI incident tools are mainly aimed at the second area: helping teams learn what went wrong and respond faster. The report also notes that Meta was using AI for pager triage two years earlier, suggesting some current practices build on earlier internal systems rather than representing a wholly new approach.
The article’s author disclosed an investment in incident.io and said they had no other affiliation with the vendors discussed and were not paid to mention them. That disclosure is relevant because incident.io is among the three companies interviewed.
“The fully autonomous agents demoed on conference stages are still not a reality.”
— The Pragmatic Engineer report
How Reliable Are the Agents?
The report does not establish a common benchmark showing how often AI SRE tools correctly identify root causes, reduce response times or prevent outages. It says accuracy depends on the company and team, and that available evaluation approaches include backtesting and testing against previous incidents. The cited material does not provide comparable results across the vendors interviewed.
It is also unclear how widely these tools are deployed in production, how often humans must correct their findings, or what safeguards are used before an agent can change live systems. The report describes autonomous fixes as a conference-stage vision rather than routine practice. The scale and causes of the reported rise in pages are also not established as industry-wide findings by the material provided.
Testing Tools Against Real Incidents
The next test for AI SRE products is whether teams can show measurable gains in their own environments: faster investigations, fewer unnecessary pages, or fewer repeat incidents, without introducing unsafe changes. Companies will need evaluations grounded in their systems and incident records, alongside clear human review boundaries for actions that affect production.
As deployment activity and AI-generated code continue to grow, the report expects interest in on-call tooling to persist. It does not identify a shared industry standard for measuring these agents or a timetable for widespread autonomous remediation. For now, their clearest role is as assistants to human responders, not replacements for them.
Key Questions
What does “AI SRE” mean?
It is a loose label for AI tools that support site reliability work, especially incident investigation, alert triage and on-call response. The report says the term can be misleading because SRE work extends well beyond handling alerts.
Can AI SRE agents fix production incidents by themselves?
The report says fully autonomous production agents, as shown in conference demonstrations, are not yet routine. Current systems are more commonly used to investigate or triage incidents with engineers involved.
Why are companies paying more attention to these tools?
The report links the interest to faster code deployment and rising on-call pages. It cites Rootly data showing pages increased from a year earlier and outside-business-hours pages nearly doubled, while noting the figures are not presented as an industry-wide measure.
Do the tools work?
The report describes practical approaches such as parallel investigations, company-specific context, model checks and testing against past incidents. It does not provide a shared benchmark or comparable vendor results, so effectiveness remains dependent on the organization and task.
Could AI on-call tools create new risks?
Yes. The report raises concerns about unsafe production changes and the possibility that engineers could lose diagnostic practice if they depend too heavily on automation. It also notes that AI systems can themselves be involved in outages.
Source: newsletter.pragmaticengineer.com
