Windows 365 for agents: a new era of automation in the cloud

Imagine delegating your most tedious computer tasks to an AI that works like a colleague — logging into systems, filling in forms, running reports, bridging gaps between applications that have no API. That's what Windows 365 for Agents makes possible. Announced at Ignite 2025, it's a platform that lets AI agents operate within secure Cloud PCs, governed by the same tooling you already use for your people.

The Scale of What's Coming

IDC projects up to 1.3 billion AI agents in operation by 2028. That's not a number organisations can ignore. The question isn't whether AI agents will become part of the enterprise IT landscape — it's whether you'll have the infrastructure to deploy them securely and at scale. Windows 365 for Agents is Microsoft's answer to that question.

What Windows 365 Agents Actually Are

Think of them as cloud-hosted workers. They combine Cloud PC technology with AI capabilities, operating on a check-in/check-out model — an agent reserves a Cloud PC, completes its task, and releases the resource. During the public preview, pricing runs at approximately $0.40 per hour of Cloud PC usage, which makes it accessible for experimentation without committing to dedicated always-on infrastructure.

The key capability is computer vision. These agents can view screens, click interfaces, and type — mimicking human interaction rather than requiring APIs. If an employee can do it on-screen, an AI agent can be taught to do the same. That opens up automation for legacy systems, third-party tools, and anything else where a modern API integration isn't an option.

How Security and Compliance Work

Each agent receives a unique Entra Agent ID, creating a complete audit trail of activity. Agents operate within your organisation's security perimeter — Conditional Access policies, role-based access controls, and data loss prevention controls apply to agents the same way they apply to employees. This is a fundamental difference from traditional RPA approaches, where bots often operate outside normal governance frameworks.

From an IT operations perspective, agent Cloud PCs appear in Intune alongside employee devices. The same workflows you use for software deployment, policy updates, and compliance reporting apply. No new toolset to learn, no shadow IT sprawl.

Windows 365 Agents vs. Traditional RPA

Legacy robotic process automation relies on brittle scripts tied to specific screen coordinates and UI states. When an interface changes — a button moves, a field is renamed — the automation breaks. Modern computer-use agents adapt through computer vision, understanding what they're looking at rather than following a fixed sequence of instructions. The maintenance burden is significantly lower.

What This Means for EUC Teams

End user computing professionals have an opportunity here that's worth taking seriously. The expertise you've built in policy management, security compliance, and device governance translates directly to managing AI agents. You already know how to govern access, enforce controls, and monitor activity through Intune. Those skills apply to agent management with minimal translation required.

Microsoft's own Copilot team uses Windows 365 for Agents operationally — this isn't a research project. Early adoption by solution providers like Manus AI and Genspark is already underway. The practical question for most organisations is which processes to automate first, not whether the infrastructure is ready.

Where to Start

The platform is currently in public preview. The most useful starting point is identifying workflows that are repetitive, well-defined, and currently manual — inventory checks, report generation, data entry across systems, compliance evidence gathering. Start with a monitored pilot, treat the agent like a new hire with limited initial access, and expand scope as confidence grows.

The governance framework already exists in your Intune and Entra ID environment. The hard part isn't the infrastructure — it's building institutional confidence in a new kind of workload.