The updates covered in this post show Microsoft Agent 365 developing into a central governance layer for that landscape. They also provide useful context for Microsoft Scout, an always-on agent that can act across Microsoft 365 on behalf of a user. Before organizations enable agents with that level of autonomy, they need to know what agents exist, who uses them, what they do, and how their activity is monitored.
Microsoft Scout introduces Microsoft's first Autopilot agent
Anyone following Microsoft's Copilot announcements over the past few months has probably come across Microsoft Scout and the broader concept of Autopilots. While much of the attention has focused on what Scout can do, the more interesting discussion for administrators may be what it represents: a shift from prompt-driven assistants to agents that operate continuously on behalf of users.
Microsoft Scout is the first agent in Microsoft's new Autopilot category and is currently available through the Microsoft Frontier program. Unlike traditional Copilot interactions that begin with a user prompt, Scout is designed to remain connected to user priorities and act proactively across Microsoft 365 services such as Teams, Outlook, OneDrive, and SharePoint. The agent runs through a dedicated desktop application for Windows and macOS and operates using the user's identity rather than a separate service account.
From an administrative perspective, Microsoft Scout introduces several prerequisites that should be reviewed before any pilot deployment. The feature is disabled by default and requires administrator enablement before users can sign in. Organizations must configure the required Intune device and access policies, complete Microsoft's opt-in and attestation process, and verify that eligible users have the necessary GitHub Copilot licensing. Microsoft notes that the attestation requirement exists because some data flows occur outside Microsoft 365.
The governance implications may be just as significant as the functionality itself. Scout is designed to operate autonomously on behalf of users, making identity, access controls, device compliance, and monitoring particularly important. Organizations considering a Frontier deployment should review the feature with security and compliance stakeholders before enabling it and ensure support teams understand that user sign-in requires all administrative prerequisites to be completed successfully.
Microsoft Scout is currently available to Frontier-enrolled organizations on Windows and macOS. Microsoft has not yet announced a broader General Availability timeline. Additional details are available in Message Center post MC1332811.
More information:
- Introducing Microsoft Scout: Your always-on personal agent
- Microsoft Scout (Frontier) documentation
- Introduction to billing and licensing
- Set up Microsoft Intune
Monitor third-party agents through Agent 365
Microsoft is extending Agent 365 observability capabilities to supported third-party agent platforms. Agent 365 can now collect and display activity data from synchronized agents running on Amazon Bedrock, Google Vertex AI, and Anthropic Claude Managed Agents. Runs, sessions, tool calls, and exceptions are surfaced through the agent's Activity tab, providing administrators with a consolidated view of activity across supported platforms.
The feature is optional for existing connections and must be enabled by an administrator. Organizations planning to use the feature should validate that the necessary permissions and telemetry collection are configured in the connected platform before enabling observability collection. New connections will have observability collection enabled by default. Microsoft notes that Agent 365 only reads activity data using the existing connection configuration and does not execute third-party agents. Organizations should also review potential tracing, logging, storage, or query costs charged by the connected provider.
Public Preview is expected to begin in late October 2026 and reach General Availability in late November 2026. These dates were updated from earlier timelines published by Microsoft. As always, Microsoft timelines may change. Additional details are available in Message Center post MC1480873.
Agent Registry consolidates duplicate agent entries
Microsoft is changing how agents are represented in the Agent Registry by consolidating multiple entries for the same agent into a single record. Previously, agents could appear multiple times when published to different scopes or when several published versions existed.
After the change, each agent will appear only once in the Registry. Administrators can access published and draft versions through a version selector in the agent details pane, while governance actions such as blocking an agent continue to apply across all versions.
The update simplifies agent inventory management, but it may also affect reporting. Agent counts in the Registry are expected to decrease because duplicate entries are being consolidated rather than removed. Organizations that rely on Registry counts, exports, dashboards, scripts, or other reporting processes should review those dependencies and validate any downstream reporting after rollout.
Rollout began in mid-September 2026 and is expected to complete by mid-October 2026, replacing an earlier rollout target of late September. Organizations should review Message Center post MC1477184 for the latest status and verify that reporting processes continue to produce the expected results after the update.
Export active user reports for Agent 365
Microsoft Agent 365 now includes a reporting capability that allows AI Admins and Global Admins to export a list of users who have actively used agents registered in the tenant's Agent 365 Registry during the previous 30 days. The export includes user principal names, the number of agents used, total sessions, and the most recent activity date.
The feature is available through the Microsoft 365 admin center under Agents > All Agents > Export and is enabled by default for eligible administrators. While the report does not change permissions, billing, or agent behavior, it provides a centralized view of agent adoption that can support licensing reviews and usage analysis.
The feature reached General Availability in late August 2026. Organizations using Agent 365 should review who has access to this reporting data, as exports contain user identifiers and usage information that may fall under existing governance or privacy policies. Message Center post MC1462914 contains additional details.
Taken together, these updates clarify the intended role of Agent 365. It is not simply an inventory of Microsoft-built agents, but a governance and operational layer covering usage, versions, activity, and supported agents from other platforms.
Microsoft Scout makes the need for that layer more tangible. An always-on agent acting through a user's identity across Microsoft 365 requires more preparation than enabling another Copilot feature. Identity, device controls, data flows, licensing, monitoring, and support processes should be understood before a pilot begins. Agent 365 does not remove those responsibilities, but it can provide some of the visibility needed to manage them.
Thank you for reading. If you'd like to stay up to date with Microsoft 365, Copilot, Teams, security, compliance, and AI-related announcements, you can follow me on LinkedIn.
Comments