Microsoft rolls out Agent 365

Microsoft is trying to solve a very specific enterprise problem: companies are filling up with AI agents faster than their security and IT teams can track them. On May 1, Microsoft moved Agent 365 into general availability, which means this is no longer a preview experiment — it is a product Microsoft says is ready for production use. The pitch is simple. Give companies one control plane to see, govern, and secure AI agents across Microsoft tools, partner platforms, cloud services, and even local machines. ### What is Agent 365, exactly? Agent 365 is basically an admin layer for AI agents. Not the agents themselves — the management system around them. Microsoft puts it directly in the Microsoft 365 admin center and ties it into Defender, Entra, Intune, and Purview, so IT teams can manage agents with the same security and compliance stack they already use for employees, devices, and apps. ### Why does Microsoft think this is urgent? Because agents are showing up everywhere. Some are built inside Microsoft 365. Some come from SaaS vendors. Some run locally on endpoints. Some use delegated human permissions, while others operate with their own credentials. That creates a new attack surface fast — data oversharing, over-permissioned workflows, tool misuse, and agents interacting in ways that manifest as “agent sprawl” and “shadow AI.” ### What changed on May 1? The big change is status. Agent 365 is now generally available for commercial customers, and Microsoft also used the launch to preview extra capabilities around discovering and managing shadow AI agents. At the same time, Microsoft made its Microsoft 365 E7 bundle generally available across its enterprise product line. ### What can admins actually do with it? The useful part is the registry. Agent 365 keeps a centralized system of record for agents across the organization, with metadata like publisher, owner, platform, deployment status, permissions, data and tool access, compliance details, and usage activity. The owners are reviewing pending requests. Basically, it is trying to make agents look less like random bots and more like governed enterprise assets. ### How broad is the coverage? Broader than just Microsoft-built agents. Microsoft says Agent 365 is meant to cover agents from Microsoft AI platforms, ecosystem partners, SaaS environments, cloud runtimes, and local agents. That cross-platform angle is the point — if companies end up with one set of agents in Copilot, another in third-party SaaS, and another on employee devices, Microsoft wants its control plane sitting above all of them. ### What about governance and policy? This is where Agent 365 connects to work Microsoft has already been doing in Microsoft 365. Admins already have agent access, publishing, inventory, and lifecycle controls in the Copilot Control System inside the Microsoft 365 admin center. Agent 365 extends that idea outward — from managing Microsoft 365 agents to governing a broader fleet with audit, retention, access, and security controls attached. ### What does Microsoft charge for it? Microsoft lists Agent 365 at $15 per user per month with an annual commitment. It is also bundled into Microsoft 365 E7, which Microsoft lists at $99 per user per month, or $90.45 without Teams. That pricing tells you who this is for — large organizations that expect agents to become normal infrastructure, not side experiments. ### So what is the real significance? The real story is that Microsoft thinks AI agents are crossing the line from software feature to managed enterprise identity. Once an agent can access data, call tools, and act on its own credentials, it starts looking less like a chatbot and more like a new class of worker — one that needs inventory, policy, monitoring, and security controls from day one. This layer will own a big part of enterprise AI rollout too.