Copilot was the first wave of AI in most tenants. The second wave is agents: Copilot Studio builds, third-party agents, and custom apps that read data, call tools, and take action. Agents are built like apps, but they behave more like users. They have access, they touch data, and they act on someone’s behalf.
That creates a governance gap. We have Entra for people and Intune for devices, but until recently nothing equivalent for agents. That is the gap Agent 365 targets.
So what is Agent 365?
Think of it as a control plane, not another agent builder. Microsoft describes it as a way to observe, govern, and secure the growing number of agents within organizations. It doesn’t replace the tools you already run. It extends them to cover agents.
You start in the Microsoft 365 admin center, where the agent registry gives you a single inventory of agents. Three existing products then do the heavy lifting:
- Microsoft Entra gives agents a real identity through Entra Agent ID, so organizations can manage them using many of the same identity principles applied to employees and applications. Conditional Access and risk-based controls apply to agents as well.
- Microsoft Purview handles the data side (more below).
- Microsoft Defender adds continuous threat detection and real-time protection that can block unsafe agent behaviour at runtime.
A simple analogy for your readers: Entra is the ID badge, Purview is the data handling policy, and Defender is the security team watching the cameras.
Where Purview fits
This is the part that matters for data governance. Agent 365 extends Microsoft Purview capabilities to help organizations understand and manage how agents interact with your data. In practice, that means the same controls you already know, applied to agent interactions:
- Information protection (sensitivity labels)
- Data Loss Prevention
- Auditing of agent activity
- Retention and compliance controls
- Insider risk signals
Microsoft’s framing is that Defender assesses risk of agent compromise, Entra evaluates identity risk, and Purview evaluates insider risk. Security and compliance teams see the same agent inventory in the workflows they already use, so you don’t need a parallel governance model.
How it relates to Copilot
This is the question I’d expect most readers to ask, so answer it directly. Copilot governance is about how people use AI. Agent 365 governs the agents themselves, including ones that aren’t Copilot. Microsoft says it supports adopting agents across Microsoft and third-party platforms without creating a separate security and governance model for each one.
Licensing and availability
As of May 1, 2026, Agent 365 is generally available for the Commercial segment on a per-user basis, and it works best with Microsoft E5 as a prerequisite. It’s also bundled into Microsoft 365 E7, which combines Microsoft 365 E5, Microsoft 365 Copilot, Microsoft Entra Suite, and Agent 365. Agents themselves don’t need their own Agent 365 licence.
Where to start
- Open the Microsoft 365 admin center and look at the agent registry. You can’t govern what you can’t see.
- Review your Purview baseline (labels, DLP, audit) before agents start touching data.
- Decide who owns agent lifecycle: IT, security, or the business.
- Pilot with a small set of agents before rolling out broadly.
Closing thought
Agent 365 isn’t a new product category so much as your existing identity, data, and threat tooling pointed at a new kind of actor. If your Purview foundations are solid, you’re already most of the way there.
