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 SRE tools are moving from conference demos into real engineering workflows, as companies report more deployments and on-call pages. The tools mainly help investigate incidents and triage alerts; fully autonomous production fixes remain unproven in routine use, and accuracy varies by company.
AI SRE tools are gaining traction among software companies and incident-management vendors, but their current role is mostly to speed up investigations and triage alerts—not to independently make production changes. The shift comes as engineering teams face more frequent deployments and on-call pages, according to The Pragmatic Engineer’s report, which spoke with Rootly, incident.io and ResolveAI and described in-house efforts at OpenAI and ING.
The report describes AI SRE as a loose label for tools that use AI to support site reliability and incident response. In practice, the products commonly investigate alerts, gather context from company systems, and report findings in collaboration platforms such as Slack or Teams. Engineers can steer investigations, while systems may use multiple agents and a company-specific knowledge layer to improve the information available to them.
The report says OpenAI built an incident-response bot called Sevbot, and an engineer at ING said the bank is developing an in-house AI SRE. It also identifies Rootly, incident.io and ResolveAI as vendors in the growing market. The article’s author disclosed being an investor in incident.io and said they were not paid by the companies mentioned.
Some tools also triage pages or suppress alerts judged unimportant, but the report distinguishes that work from the more ambitious conference-stage demonstrations of an agent identifying a root cause and shipping a production fix without waking a human responder. According to the report, fully autonomous incident handling is not yet a routine reality; whether agents should be permitted to change production code remains debated.
More Pages, Different On-Call Work
The appeal is practical: if teams deploy more often and receive more alerts, responders need ways to find relevant evidence faster and avoid spending time on low-value pages. The report cites Rootly data showing that pages have increased from a year earlier and that outside-business-hours pages have nearly doubled. It does not provide the underlying counts or a detailed measurement method in the supplied material, so the figures should be read as vendor-reported data rather than a universal industry measure.
AI support could reduce repetitive investigation work and help teams learn from incidents, but it does not remove the reliability tradeoff. More changes can mean more opportunities for failures, and a fast but mistaken automated response could worsen an outage. Human oversight, clear permissions and ways to verify an agent’s reasoning remain central to how teams deploy these systems.
The report also flags a potential deskilling risk: if engineers rely on automated triage for too long, they may get less practice diagnosing incidents themselves. It suggests incident-response simulations as one way to maintain those skills, while noting that engineers may eventually receive fewer pages but face more difficult incidents when paged.
Products Worth Considering
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Jackery HomePower 3600 Plus Portable Power Station, 3584Wh, 3600W
As an affiliate, we earn on qualifying purchases.
From BUILD Demo to Incident Tools
At Microsoft’s BUILD conference in May 2025, the company demonstrated an Azure SRE Agent responding to an alert, identifying a possible root cause and shipping a fix without waking a human on-call engineer. The report’s author described the demonstration as difficult to imagine teams allowing in real production, given the risk of an automated change making an outage worse. The source does not establish that this demo represented a generally available or routinely used autonomous system.
The renewed interest described in the report is tied to the rise in AI-assisted software development and deployment. Its reliability framing is that incident rates depend both on how often changes are made and on the chance that an individual change causes a failure. The report says many companies have seen pull-request frequency at least double over a year, but that observation is not independently verified in the supplied source. More deployments can increase operational pressure unless testing and incident learning reduce the risk per change.
The vendors interviewed generally prefer more descriptive labels, such as automated incident investigation or remediation agents, but the term AI SRE has caught on. The report notes that the label is imprecise: SRE work includes much more than handling alerts.
“The fully autonomous agents demoed on conference stages are still not a reality.”
— The Pragmatic Engineer report
Accuracy and Autonomy Remain Open
The report does not provide a shared benchmark showing how accurately AI SRE tools diagnose incidents across companies. It says accuracy is difficult to measure and depends on each organization’s systems and even individual engineering teams. Suggested approaches include testing agents against past incidents, running evaluations, using different models, and having one agent challenge another’s hypotheses, but the source does not give comparative results for these methods.
It is also unclear how broadly the tools are deployed in production, how often their recommendations are accepted, and whether they measurably reduce incident duration or engineer workload. Vendor-reported page trends do not by themselves establish that AI tools caused a change in reliability. The scope of autonomous permissions—and the safeguards teams use before an agent changes production—also remains unsettled.
Testing the Tools in Production
The next test is whether teams can show that AI-assisted investigation improves response without introducing new operational risk. That requires company-specific evaluations, comparisons with past incidents, and reporting on outcomes such as time to diagnose, alert quality and the rate at which engineers accept or reject recommendations. The source report does not identify a common industry standard or a scheduled milestone for such evidence.
For now, the described direction is toward agents that gather evidence, triage pages and work alongside human responders, with production actions kept under human control or limited by company policy. As adoption grows, teams will also need to monitor incidents involving AI-generated code and preserve engineers’ incident-response skills through practice.
Key Questions
What does AI SRE mean?
AI SRE is a broad label for AI-assisted reliability and incident-response tools. The systems described in the report mainly investigate alerts, assemble context and help triage on-call pages.
Can an AI SRE fix an outage without a human?
Some conference demonstrations have shown agents proposing or making fixes, including Microsoft’s May 2025 BUILD demo. The report says fully autonomous incident handling is not routine, and whether an agent should change production code remains debated.
Why are companies looking at these tools?
The report links interest to higher deployment activity and rising on-call pages. AI tools may help responders investigate incidents faster and filter less important alerts, though the source does not establish that they have reduced outages overall.
How well do AI SRE tools work?
There is no single accuracy measure in the report. Results depend on each company’s systems and data, and the source says teams are using evaluations, historical incident tests and other checks to assess performance.
Could AI SRE make engineers less prepared?
The report identifies deskilling as a risk if engineers get less practice diagnosing incidents. It points to incident-response simulations as one way organizations can keep those skills active.
Source: newsletter.pragmaticengineer.com
