AI tools are helping some organisations investigate and respond to infrastructure incidents faster, according to figures reported by PYMNTS. The useful distinction: these examples concern AI helping teams manage disruption, not AI causing the recent outages.
Microsoft AI Watch analysis
What happened
PYMNTS reports that T-Mobile’s AutoPilot made real-time network adjustments in about half the time during testing. At financial data provider MSCI, Resolve AI said incidents fell by 77% on one infrastructure team and 76% on another after its agents began initial investigations.
Microsoft said its Azure team cut mitigation times for high-severity incidents by more than 50%. Its agents assemble incident histories, identify likely causes and prepare root-cause reports, the company said.
The reported figures come from different organisations and sources, and are not a single comparative trial. PYMNTS says the recent Microsoft, FAA and JPMorganChase outages were unrelated.
Why it matters
These examples show where AI can be practically useful in operations: sorting alerts, connecting signals and helping engineers find what needs attention. At MSCI, PYMNTS says one investigation that had taken engineers three to four hours was traced to an outdated server address in 14 minutes.
Faster triage is not the same as preventing every outage, and these results should not be treated as proof that every team will see similar gains. But they make a more concrete case for AI in incident response than the usual promise that software will somehow become more reliable by being given a chatbot.
Our read
The strongest point here is not that AI can stop outages. It is that teams are using it to narrow the search when something has already gone wrong. Operators weighing these tools should look for evidence from their own systems, including whether the AI’s explanations help engineers act faster.
What to watch
- Whether the reported response-time gains are sustained as deployments expand.
- How teams check AI-generated diagnoses before acting on them.
- Whether future evidence compares performance across different incident types.
Discussion spark: Should infrastructure teams let AI agents take first action on low-risk incidents, or keep them strictly in an advisory role?
Sources and evidence
- Chase, Microsoft and FAA Meltdowns Hand AI Its Big Moment – PYMNTS.com (18 November 2026, 20:43 UTC)
- Microsoft Introduces New Models to Cut the Cost of AI Coding and Cybersecurity – PYMNTS.com (18 November 2026, 20:43 UTC)
not affiliated with or endorsed by Microsoft