This Week in IT Operations: October 26-November 1, 2026

Welcome back to This Week in IT Operations, our weekly roundup of the critical news that is shaping IT operations. Every week, we curate the stories that matter most to ITOps, AIOps, SRE, and incident management leaders, then tell you why each one matters.

A major theme that emerged this week is that as AI agents take on more of IT operations, the underlying infrastructure and integrations are becoming the real test. Gartner set the stage with its Top AI Trends in Infrastructure and IT Operations for 2027, predicting that by 2030, half of organizations will use AI agents with minimal human involvement in IT operations, and warning that uncapped token pricing will blow through cost expectations.

The rest of the week showed what scaling agents actually demands. GitHub is rebuilding its infrastructure as agents start more than 40% of pull requests, and the hyperscalers say cloud demand still outruns supply.

The week also showed how much of ITOps depends on third-party providers and integrations. Claude’s Google Calendar connector returned errors for more than two days before the first public status update, and a breach at IT provider Amadeus exposed customer data at Japan Airlines and several other carriers. In both cases, the lesson remains the same: detection and communication around third-party dependencies matter as much as they do for the systems inside your own walls.

Here’s what happened, and why it matters.

 

This week in IT Operations at a glance

• Gartner: By 2030, 50% of organizations will use AI agents with minimal human involvement in IT operations, and 25% of enterprises will merge their NOC and SOC.
• Anthropic: Claude’s Google Calendar connector had errors for more than two days before the first public status update.
• Hyperscalers: Azure grew 48%, Google Cloud grew 103%, and AWS grew 39%, but Microsoft trimmed and Alphabet held its capex forecast.
• GitHub: AI-agent load is forcing an infrastructure rebuild, with agents now starting more than 40% of pull requests.
• Japan Airlines: A breach at IT provider Amadeus exposed data for about 92,000 customers.

1. Gartner’s 2027 AI trends for I&O: agents on the loop, token costs, and NOC-SOC convergence

Gartner has named its top five AI trends for infrastructure and operations (I&O) leaders heading into 2027; the list was announced on October 26 and is based on research published in September. The urgency is real: in a Gartner survey of 272 I&O leaders, 58% said that scaling AI proofs of concept into I&O adoption will be a priority over the next 12 months. On autonomy, Gartner expects half of all organizations to run AI agents in IT operations with minimal human involvement by 2030, as teams move from human-in-the-loop to human-on-the-loop operations. On cost, the firm warns that through 2027, teams paying for agents through uncapped, token-based pricing will overshoot their initial cost expectations by more than 50% compared with those on predictable subscription or bundled plans, and it advises a strict “no telemetry, no deployment” rule to close blind spots in AI spending. On the organizational front, Gartner expects one in four enterprises to combine their NOC and SOC by 2030, up from 5% in 2026, saving more than 25% on combined tooling and staffing costs.

Sources: Gartner Newsroom, IT Voice

Key takeaways

• Plan for human-on-the-loop, not human-in-the-loop. Gartner describes a shift from people approving every step to people supervising agents that act on their own. That changes what your engineers do day to day, so start defining supervision roles, escalation paths, and override rules now.
• Pricing model is an operational decision. A 50% cost overrun is the difference between a program that scales and one that gets canceled. Model your expected agent usage before you choose a vendor, and favor pricing with caps or predictable tiers.

2. AI assistant connector outages highlight a gap in incident communication

Anthropic’s Google Calendar connector for Claude returned errors for some requests to list or search calendar events from 11:07 UTC on October 26 to 18:52 UTC on October 28, according to Anthropic’s status page, which says the company worked with an upstream provider to fix it.

A separate Gmail connector incident ran for about six and a half hours on October 27–28. According to status trackers, the first public status update on the Calendar issue was posted on October 28, about 53 hours after the errors began.

As enterprises bring AI assistants into daily workflows, detection and status communication for these integrations matter more than ever.

Sources: Claude status page, SessionWatcher

Key takeaways

• Monitor error rates on the integrations your workflows depend on, not just whether the service is up. A connector that fails some requests but not all rarely triggers basic availability checks.
• Treat provider status pages as a lagging record. A status page tells you what a provider has confirmed, not what your users are experiencing right now. Base your detection and failover decisions on your own telemetry.
• Communicate early, even with incomplete information. The gap between impact and acknowledgment is where trust erodes. The same rule applies to your own stakeholder updates, where a fast “we’re investigating” beats a perfect update hours later.

3. Hyperscaler earnings: cloud growth is strong, but AI capex forecasts stall

Microsoft, Alphabet, and Amazon all reported earnings this week. Azure grew 48%, Google Cloud grew 103% to $30.7B, and AWS grew 39% to about $46B, its fastest growth in nearly five years. But CNBC reports that the AI buildout hit a snag: Microsoft and Alphabet both reported capital expenditures below analysts’ estimates.

Microsoft trimmed its full-year capex forecast to $170B from $175B and disclosed delays in some third-party capacity deliveries, while Alphabet held its guidance at $195B to $205B rather than raising it. Alphabet’s finance chief said outside computing power came online only at the very end of the quarter. Alphabet shares closed down 12.6% on October 29 after the company missed on earnings and reported negative free cash flow.

Cloud capacity and cost remain the main constraints on enterprise AI plans.

Sources: CNBC, 24/7 Wall St.

Key takeaways

• Demand still outruns supply. Both Microsoft and Amazon say demand for increased cloud capacity exceeds what they can deliver. If your AI roadmap assumes capacity will be there when you need it, build in buffers for regions, instance types, and timelines.
• Your AI timeline depends on your provider’s supply chain. Delays at third-party capacity suppliers are now showing up in hyperscaler earnings. That risk cascades to the enterprises waiting on that capacity for new workloads.
• More growth means more concentration. As more critical workloads run on a handful of providers, a single provider incident hits more of your business, as last week’s Google Cloud London outage showed. Resilience planning has to keep pace with cloud adoption.

4. GitHub Universe: AI-agent load forces an infrastructure rebuild

At Universe ’26 (October 28–29), GitHub addressed platform stability head-on after a year of reliability problems. According to The New Stack, GitHub has added 4 million CPU cores, 160 petabytes of high-speed storage, and 50 Tb/s of network capacity this year, and is re-architecting its Git backend to separate storage from compute. It’s also moving to a multi-region deployment on Azure, which it expects to finish early next year. The demand driver is clear in GitHub’s Octoverse 2026 data: the platform saw 7.4B commits in September alone, and by September more than 40% of pull requests started with a draft from an AI agent, up from 17% in March.

It’s a clear example of agentic workloads breaking long-held assumptions about operational scale.

Sources: The New Stack, The Next Web

Key takeaways

• Agent traffic doesn’t follow human patterns. More than doubling the share of agent-started pull requests in six months is the kind of growth no human-based capacity plan anticipates. Revisit your own scaling assumptions wherever agents are starting to generate work.
• Capacity buys time, but architecture fixes the problem. Millions of new cores helped GitHub absorb the surge, but the lasting fix is a backend redesigned for agent-scale write volume. If your systems are struggling under new load, ask whether you’re facing a capacity problem or a design problem.
• Agent workloads need their own guardrails. As agents generate more activity, teams need limits on how much a single agent or automation can consume. Budgets, rate limits, and visibility into agent-driven load are becoming core reliability controls.

5. Japan Airlines hit by a breach at IT provider Amadeus

On October 28, Japan Airlines (JAL) disclosed that Amadeus, which runs its reservation, ticketing, and check-in systems, had suffered a breach. According to JAL, the attacker accessed reservation codes, names, flight dates, and routes for about 90,000 customers, and a further 2,000 also had email addresses, phone numbers, and passport numbers exposed. Flight operations weren’t affected. Amadeus notified JAL on October 23 that unauthorized access had occurred between August 2023 and September 2026. JAL isn’t alone: ANA, Korean Air, and Asiana have also reported impact from the same breach.

Sources: The Record (Recorded Future News), Travel Radar, TechNadu

Key takeaways

• Shared vendors mean shared incidents. One provider’s breach reached several major airlines at once. Map which critical providers your business shares with the rest of your industry, because their incidents can become yours overnight.
• Dwell time is the real story. Unauthorized access reportedly ran for more than three years before notification. Ask your critical vendors how they detect intrusions, how quickly they’ll notify you, and what evidence they’ll share when they do.
• Plan for the follow-on wave. A data breach rarely ends at disclosure. Customer communications, fraud and phishing that reuse exposed details, and regulatory reporting all belong in the same incident plan as the technical response.

On the calendar: upcoming ITOps events

Meet BigPanda at Gartner IOCS 2026 in Las Vegas

As AI agents take on more of the work, this week made it clear that cost control, trusted integrations, and visibility into third-party dependencies will decide which agentic programs scale and which stall. We’d love to talk about what that means for your team.

Join BigPanda at the Gartner IT Infrastructure, Operations & Cloud Strategies Conference, December 8–10, 2026, at The Venetian in Las Vegas. Stop by booth 301 to see how BigPanda helps enterprises detect, triage, and resolve incidents with agentic AI. See you there, and check back next week for the next edition of This Week in IT Operations!