Turning agent sprawl into structured control

Artificial intelligence in the workplace has moved beyond chat interfaces and simple task automation. Organisations are now deploying AI agents that can read data, take decisions, trigger workflows and act on behalf of users and teams. As this accelerates, a new challenge has emerged. How do you manage, secure and govern AI agents at enterprise scale using the same standards you apply to people, identities and applications

Microsoft’s Agent 365 looks as if it may be the answer.

What is Agent 365

Agent 365 is Microsoft’s control plane for AI agents. It provides a central way to register, manage, secure and monitor AI agents across Microsoft 365, Azure and approved third party platforms. Microsoft introduced Agent 365 at Ignite in November 2025 as part of its shift towards what it calls the agentic enterprise.

At its core, Agent 365 recognises AI agents as standard, manageable components within Microsoft ecosystem. Each agent can be given its own identity in Microsoft Entra, its own permissions, and its own security and compliance policies. Agents are then surfaced and managed through familiar tools such as the Microsoft 365 admin centre, Entra, Defender and Purview.

This approach avoids the need to bolt on separate governance platforms or invent new security models for AI. Instead, organisations extend the controls they already use for users and applications to cover agents as well.

Why organisations need Agent 365

Most organisations will not deploy just one AI agent. They will deploy many.

Some will be built internally using Copilot Studio or developer frameworks. Others will come from software vendors such as ServiceNow or SAP. There’s a risk that over time agent sprawl will become apparent and with it, limited visibility and unclear accountability.

Without a control plane, common risks quickly appear.

Agents can gain access to data they should not see
Security teams cannot easily tell which agents exist or what they are doing
Compliance teams cannot audit how agents interact with sensitive information
IT teams struggle to control who can create or deploy new agents

Agent 365 addresses these issues by providing a single source of truth for all agents in the organisation. It maintains an inventory of agents, enforces access controls, applies security policies and provides monitoring and audit capabilities across the entire agent estate.

This becomes particularly important in regulated industries or hybrid environments where agents may interact with on premises systems, cloud services and third party platforms simultaneously.

In simple terms, Agent 365 makes AI agents easier to manage and govern.

How Agent 365 fits with Copilot and Microsoft 365

It is important to be clear about what Agent 365 is and what it is not.

Agent 365 does not replace Microsoft 365 Copilot. Copilot focuses on what AI can do for users inside applications such as Outlook, Word and Teams. Agent 365 focuses on controlling what AI agents are allowed to do, what they can access, and how they are monitored.

Similarly, Agent 365 is not included with Microsoft 365 E3 or E5 by default. Microsoft has been explicit that Agent 365 is a separate product because it serves a different purpose and targets agents rather than human users.

This separation allows organisations to scale their use of AI agents independently from their user licensing strategy.

The licensing and cost model

Agent 365 uses a dedicated licensing model that is separate from Microsoft 365 and Copilot.

Each AI agent that is brought under management in Agent 365 requires an Agent 365 licence assigned by an administrator. These licences are designed to cover the identity, governance, security and observability features provided by the platform.

Agent 365 is currently available through Microsoft’s Frontier programme, which provides early access to new AI capabilities. At the time of writing, Microsoft has confirmed that Agent 365 will not be bundled into existing Microsoft 365 or Copilot licences and will be sold as its own offering when it reaches general availability.

Microsoft has not yet published final public pricing. However, the direction is clear. Organisations should expect to license agents in a similar way to how they license users or service accounts today, with costs scaling based on the number of agents in use rather than the number of employees.

This model reflects a broader industry shift where AI agents are treated as digital workers with their own cost, risk and governance profile.

Why this matters now

AI agents are moving quickly from proof of concept to production. As they do, organisations that fail to put governance in place early will face security, compliance and operational challenges later.

Agent 365 provides a structured way to adopt agent based automation without losing control. It allows IT, security and compliance teams to enable innovation while maintaining the standards the business already relies on.

For organisations serious about scaling AI safely, Agent 365 is not an optional extra. It is foundational infrastructure.