Microsoft has announced Windows 365 for Agents in public preview, a cloud PC platform designed specifically to run AI agents in secure, isolated cloud environments. The product addresses one of the most significant practical concerns about enterprise AI agent deployment: the risk of giving autonomous agents access to sensitive endpoints, files, and systems alongside human workers.
How It Works
Windows 365 for Agents provisions dedicated cloud PCs for agent workloads. Organisations direct agents with natural language to interact with applications, browsers, files, and enterprise systems — all within a contained environment governed by the same Defender and Intune policies that protect human endpoints. Agents can access legacy applications, UI-based systems, and tools without APIs, resolving the long-standing problem of automating processes that were never designed for programmatic access.
The Security Architecture
The key architectural decision is separation. Rather than running agents on the same infrastructure as human workers, Windows 365 for Agents creates an isolated fabric. If an agent behaves unexpectedly — or is compromised — the blast radius is contained to the agent environment. For UK organisations in regulated industries, this isolation model is likely to be a prerequisite for deploying any autonomous agent in production. The FCA and ICO have both signalled interest in how financial services and data processors govern AI agents.
Pricing and Availability
The product is in public preview with consumption-based pricing. A dedicated agent compute SKU reaching general availability is expected in Q4 2026. UK organisations interested in the preview can apply through the Microsoft 365 admin centre.
