Week 22

Enterprise AI Weekly: May 26–30, 2026

By Attila · May 30, 2026

A quieter week than Google I/O, but three stories that matter significantly for how enterprises are actually deploying and governing AI right now — plus one workforce story that signals something bigger about where this is heading.

1. Microsoft Copilot Studio Gets Computer-Using Agents — Now Generally Available

What happened: Microsoft announced that computer-using agents in Copilot Studio are now generally available across all commercial Power Platform geographies. Computer-using agents are exactly what they sound like: AI agents that interact with applications through the UI — clicking, reading screens, filling forms, navigating web pages — rather than through APIs. The same May update also brought agent-to-agent (A2A) communication to GA, remote MCP server support into preview, and a redesigned workflows canvas that combines process orchestration with AI reasoning steps on a single surface. Real-time voice agents through Dynamics 365 Contact Center also hit GA for North America.

Why it matters to sysadmins and IT decision-makers: Computer-using agents close the automation gap that has existed for years in enterprise IT. There is a long tail of business-critical applications — legacy portals, vendor web interfaces, proprietary internal tools — that have no API, no PowerShell module, no way to automate programmatically. Until now, the options were either a brittle RPA script that breaks whenever the UI changes, or a human clicking through the process manually. Computer-using agents navigate live UIs using vision and reasoning, adapting when layouts change rather than failing.

The governance angle is what IT teams need to focus on. GA means these agents can now be deployed to production under your tenant’s data residency and compliance boundaries. The Power Platform admin center controls which agents are active, what credentials they can access, and what they can interact with. Before your Power Platform team starts building agents that click through HR portals or vendor billing systems, make sure those governance controls are configured — credential management, access scoping, audit logging, and approval workflows for high-risk actions.

The A2A and MCP additions compound this: agents can now hand off tasks to other agents and connect to external MCP-compliant tools. The practical result is that multi-step cross-system automation is becoming significantly easier to build in the Microsoft stack. The administrative overhead of governing it is growing at the same rate.

2. Intuit Cuts 3,000 Jobs — And the Pattern Is Worth Paying Attention To

What happened: Intuit announced on May 20 that it is eliminating approximately 3,000 positions — 17% of its global workforce of 18,200 — to reduce organizational complexity and reallocate resources toward AI product development across TurboTax, QuickBooks, Credit Karma, and Mailchimp. CEO Sasan Goodarzi framed the decision as removing layers of management to ship AI products faster, not replacing specific roles with AI directly. Impacted US employees remain until July 31 and receive severance. The layoffs came on the same day as the Intuit Q3 earnings call, which showed strong revenue — consistent with the broader 2026 pattern of profitable tech companies cutting headcount while citing AI investment.

Industry-wide context: more than 114,000 tech workers have been laid off in 2026 so far, per Layoffs.fyi. Companies explicitly citing AI as a factor in restructuring decisions include Amazon, Block, Cisco, Cloudflare, Meta, Microsoft, and Oracle.

Why it matters to sysadmins and IT decision-makers: Two things here, one immediate and one longer-term.

Immediately: if your organization uses QuickBooks, TurboTax for Business, or Mailchimp, monitor support quality over the next 6–12 months. A 17% workforce reduction at a SaaS company typically affects customer support, implementation, and professional services teams before it affects engineering — those are the people your IT team calls when something breaks.

Longer-term: Intuit is one of many enterprise software companies going through the same calculus this year — strong revenue, profitable operations, and a board that sees AI as a reason to run leaner. The pattern matters because these are the vendors whose products are in your environment. Understand the AI roadmap of your key vendors, not just the features, but what it means for support staffing, product deprecation timelines, and the humans on the other end of your support tickets.

3. GPT-5.5 Is Now the Default Enterprise ChatGPT Model

What happened: OpenAI’s GPT-5.5 and GPT-5.5 Instant are now fully rolled out to all ChatGPT tiers — Plus, Pro, Business, and Enterprise — as well as the API. GPT-5.5 Instant replaced GPT-5.3 Instant as the default model for all ChatGPT users on May 5. The model delivers meaningful improvements in coding, computer use, reduced hallucinations in regulated domains (law, medicine, finance), and agentic performance — specifically its ability to handle ambiguous multi-step tasks without needing step-by-step guidance. OpenAI president Greg Brockman described the core advancement as: “What is really special about this model is how much more it can do with less guidance. It can look at an unclear problem and figure out just what needs to happen next.”

For Business and Enterprise users, GPT-5.3 Instant remains available for three months via model configuration settings before being retired.

Why it matters to sysadmins and IT decision-makers: If your organization has deployed ChatGPT Enterprise or ChatGPT Business, your users are now running on GPT-5.5 by default. This is worth a brief check-in with your AI governance team for two reasons.

First, model behavior changes between versions — particularly in how the model interprets ambiguous instructions and how much autonomous action it takes in agentic workflows. If you have any Copilot or ChatGPT integrations running automated tasks against business systems, validate that behavior hasn’t shifted in a way that produces different outputs than expected.

Second, the hallucination improvements in law, medicine, and finance are relevant if your organization has been cautious about deploying AI in those domains. It may be worth revisiting use cases that were previously deprioritized due to accuracy concerns — the evaluation bar has moved.

4. Microsoft Build Is Next Week — Here’s What to Watch

What happened: Microsoft Build 2026 runs June 2–3 at Fort Mason in San Francisco, with keynotes livestreamed globally. Pre-event reporting and confirmed session topics point to a focus on: Azure AI Foundry updates, autonomous GitHub Copilot agent capabilities, Entra Agent ID (identity governance for AI agents), multi-agent orchestration maturity, and on-device Windows AI APIs. This is Microsoft’s developer-focused counterpart to Google I/O — the venue where the infrastructure and tooling behind Copilot gets detailed for the engineering audience.

Why it matters to sysadmins and IT decision-makers: Two things in the confirmed agenda are directly relevant to IT teams rather than just developers.

Entra Agent ID is the one to watch closely. As AI agents proliferate in Microsoft 365 and Azure environments, they need identities — and those identities need to be managed, audited, and scoped like any other service principal. Entra Agent ID is Microsoft’s answer to the non-human identity problem. If your org manages Entra ID and is starting to deploy Copilot agents, this session will define how you govern them going forward.

GitHub Copilot autonomous agents — a year after the initial announcement, real-world deployment data will come out at Build. If your organization has developers using GitHub Copilot, the autonomous agent capabilities (write tests, fix bugs, open PRs without prompting) are expanding. Understanding what governance controls exist in GitHub Enterprise for these agents is worth reviewing before they land in your repositories.

Build sessions are livestreamed free at build.microsoft.com. Come back next week for the Week 23 post covering what actually shipped.

5. The Bigger Pattern: 114,000 Tech Layoffs, Record Revenue, and What It Means for IT Budgets

What happened: Layoffs.fyi data shows more than 114,000 tech sector job cuts in 2026 through May — on pace to exceed both 2024 and 2025. The consistent pattern: companies reporting strong or record revenue are simultaneously cutting 5–20% of their workforce, with AI cited as both the reason for optimism (future growth) and the justification for cuts (current efficiency gains). Intuit this week, Oracle (30,000 earlier this year), Microsoft (8,750 in a “Rule of 70” retirement program), Cisco, Cloudflare, Meta, Amazon. All profitable. All cutting.

Why it matters to sysadmins and IT decision-makers: This matters for two reasons that are easy to miss in the individual announcements.

First, it changes the vendor landscape you operate in. When a SaaS vendor cuts 15–20% of staff, the first impact is usually felt in professional services, customer success, and tier-2 support — the teams that help you with complex implementation issues, migration projects, and edge-case troubleshooting. The AI chatbot on their support portal gets better; the human expert who understood your environment gets a severance package.

Second, it signals where AI investment is actually coming from. These aren’t companies borrowing to fund AI — they are reallocating existing labor spend. That math eventually reaches your organization’s own IT budget conversations. If leadership is watching Microsoft, Oracle, and Intuit cut 17% of headcount while citing AI productivity, the question of how AI is changing your team’s productivity is not far behind. Getting ahead of that conversation — with data about what your team actually does that AI cannot do — is a better position than waiting to be asked.

The Week in Summary

Three operational signals from this week:

If you run a Microsoft 365 or Power Platform environment, the Copilot Studio GA announcements mean computer-using agents and multi-agent orchestration are now production-ready features, not previews. Your Power Platform governance policies need to cover them.

If your organization uses any of the major enterprise SaaS vendors that have cut staff this year — and the list is long — build a short-term vendor health check into your IT risk review. Support quality degrades 3–6 months after a major layoff wave, not immediately.

Microsoft Build next week (June 2–3) will answer what Entra Agent ID actually looks like in practice. That is the identity governance story for the agentic era, and it belongs in IT planning conversations the same way Entra ID does today.

Next edition publishes June 6 — covering Microsoft Build 2026 and anything else that lands during the week.

More Enterprise AI Weekly coverage:

Enterprise AI Weekly: May 12–18, 2026 — Week 20
Enterprise AI Weekly: May 19–25, 2026 — Week 21