Within Microsoft 365, you can build and govern several types of agents, each aligned to different skill levels and use cases:

• SharePoint-based agents
  Use content stored in SharePoint sites and libraries as the foundation for agents. These are well suited for end users who want to support day-to-day tasks using the content they already work with.
• Copilot Studio Agent Builder (inside Microsoft 365 Copilot)
  Lets users create conversational templates directly from Copilot conversations. This is ideal for end users and makers who want reusable agents tailored to specific tasks or business needs, without deep development skills.
• Full Copilot Studio agents
  Use triggers, advanced logic, and connections to other Microsoft and third-party services. These are typically built by makers who need more autonomous agents that can handle complex scenarios.
• Pro developer tools (Teams Toolkit, Azure AI Foundry, etc.)
  Used by developers to build fully customized agents with the model and orchestration engine of their choice. These agents are usually managed and distributed via centralized IT catalogs.

The whitepaper identifies three main user audiences for these tools:

• End Users – Use intuitive tools like SharePoint and Agent Builder. They work within their existing permissions and don’t need deep technical expertise.
• Makers – Have a stronger understanding of technology and governance. They primarily use Copilot Studio to create more advanced and sometimes autonomous agents.
• Developers – Build sophisticated, highly governed solutions using pro developer tools. Their agents are typically surfaced through IT-managed catalogs.

Each audience requires a different level of governance, but all agents can be managed through a spectrum of controls (tool controls, content controls, and agent management) to keep your environment secure and compliant.

Agent governance in Microsoft 365 is organized around three main control areas:

1. Tool controls
   These govern what functionality is available in the tools used to create agents. They help ensure agents operate within parameters set by administrators.
   Key admin surfaces:
   • Microsoft 365 Admin Center (MAC) – Central portal to configure and manage Microsoft 365 services, including Microsoft 365 Copilot and many agent-related settings.
   • Power Platform Admin Center (PPAC) – Used for managing Power Platform-related capabilities that agents may rely on.
2. Content controls
   These govern the information agents can access, process, and generate, helping you comply with privacy and data protection requirements.
   Main tools include:
   • Microsoft 365 Admin Center (MAC)
   • Power Platform Admin Center (PPAC)
   • Microsoft Purview – For data governance and compliance.
   • SharePoint Advanced Management (SAM) – To control Copilot behavior at the content source level.

   For example, SAM’s Restricted Content Discovery policy can hide a site’s content from global search and Copilot indexing, which helps protect sensitive data. SAM also supports site sharing restrictions and block download policies to reduce oversharing and prevent downloading of confidential files.

3. Agent management
   Focuses on controlling and monitoring agent activities across their lifecycle. This includes:
• Staged rollouts for controlled deployment
• Usage reports with metrics on user engagement and per-agent activity
• Adoption tracking and identification of underused agents
• Lifecycle management, including decommissioning and deletion

Agent management is primarily handled through the Microsoft 365 Admin Center, including the Integrated Apps area for IT Catalog agents.

SharePoint agents rely on existing SharePoint Online permissions (e.g., Full Control, Edit, Contribute, Read). By design, a SharePoint agent can only access or update content that the current user already has access to, and because agents are stored as files, the same permissions also govern who can use or modify the agents themselves.

Together, these controls help you reimagine how agents operate in your environment while keeping data access appropriate, auditable, and compliant.

Microsoft 365 provides several capabilities to help IT teams manage agents at scale:

1. Centralized inventory and approvals via Integrated Apps

In the Microsoft 365 Admin Center (MAC), the Integrated Apps section acts as a centralized inventory for apps and agents:

• All apps and agents available in the tenant are listed, enabling admins to block agents or assign them to specific users.
• When an agent is submitted for approval, uploaded by an admin, or referenced from the public store, its metadata (capabilities, data sources, custom actions) is shown on the app detail tab.
• Admins can review an agent’s security and privacy posture and decide whether to allow or block it.
• The shared agents page and search make it easier to locate agents distributed across departments and take action if needed.

This approach helps maintain a curated catalog of allowed agents and supports quick intervention when an agent no longer meets organizational standards.

2. Trust signals: Publisher attested and Microsoft 365 certification

For agents and apps from the public store, two validation layers help build trust:

• Publisher attested – The publisher signs a legally binding attestation that their application meets defined security and development standards.
• Microsoft 365 certification – Microsoft performs a more in-depth evaluation of security, compliance, data handling, and performance within the Microsoft 365 ecosystem.

These signals help admins decide which external agents to allow into the environment.

3. Usage and activity reporting

The Microsoft 365 Copilot page in MAC (Reports > Usage) provides insights into:

• Which agents exist and how often they are used
• Usage frequency, duration, and distribution across platforms (desktop, web, mobile)
• Which agents are most adopted and which are underused

These metrics support data-driven decisions about which agents to promote into the IT Catalog, where to invest further development, and where to retire or consolidate agents.

4. Security monitoring with Microsoft Sentinel

Microsoft Sentinel, a cloud-native SIEM and SOAR platform, can be used to monitor agent-related activities as part of your broader security operations. It supports:

• Real-time monitoring and alerting on suspicious or non-compliant activities
• Customized alerts for critical events involving agents
• Historical analysis to identify patterns and improve security posture over time

By combining Integrated Apps, usage reports, trust certifications, and Sentinel, IT teams can reshape how they oversee agents—maintaining a secure, compliant, and well-governed catalog as adoption grows.