Microsoft Scout.

Microsoft Scout was introduced in June 2026 as Microsoft’s first “Autopilot” agent, a personal work agent intended to keep tasks moving across Microsoft 365 and other tools. In September 2026, Microsoft described Autopilot as the new name for that personal-agent experience. The shift in naming can be confusing: Microsoft documentation still uses Microsoft Scout for a related desktop AI app distributed through its Frontier early-access program.

The distinction matters. Autopilot refers to the persistent, proactive agent experience. Microsoft Scout (Frontier) refers to the companion desktop app, which can work with local files, a browser, developer tools, and Microsoft 365 data. Product surfaces, requirements, and capabilities may differ, so confirm which one a product description or administrator is referring to.

The personal-agent concept is different from a conventional chatbot. Instead of waiting for a new prompt for each step, an agent can retain task context and continue work over time, within its configured access and controls. Microsoft positions Autopilot as a persistent agent that can work across Microsoft 365.

The desktop Scout app adds a local-computer surface. Its documentation describes work with a user-designated workspace, browser activity, shell commands, and Microsoft 365 content. These actions make the app more capable than a chat window, but also mean it requires more deliberate device and data governance.

A user gives an agent an objective and boundaries. The agent can then use available tools and work context to complete a sequence of tasks, such as preparing for a meeting, tracking follow-ups, or coordinating a recurring process. Some work can happen in the background, with the user reviewing progress or handling actions that require approval.

The agent’s output is not automatically correct or appropriate. It may misread context, misunderstand a request, or take an action that needs human judgment. Treat its work as delegated activity that needs defined limits and review, not as an independent decision-maker.

What the agent can see or do depends on the product surface, account, permissions, connected tools, and organizational configuration. For the desktop Scout app, Microsoft describes access to a selected workspace and connected Microsoft 365 content. The broader Autopilot experience is described as working across Microsoft 365 with its own identity and governance controls. Do not assume these access models are identical.

Organizations evaluating either experience should consider:

• Scope: Which files, mail, calendars, chats, sites, or other data are in reach?
• Actions: Can the agent draft, edit, send, share, schedule, or run commands?
• Approvals: Which actions require a person to review them before completion?
• Background activity: What recurring or triggered tasks can continue without a person actively prompting the agent?
• Data handling: Which AI model providers process prompts or content, and what policies govern that processing?

Approval prompts and access controls reduce risk but do not remove it. External content can contain misleading instructions, and mistakes in a background task may be harder to notice promptly. Start with limited access, test workflows using low-risk data, and review logs or activity indicators available in the specific experience.

A project coordinator could ask an agent to prepare for a weekly delivery review, identify unresolved follow-ups from authorized work content, draft a status summary, and suggest calendar time for owners to discuss blockers. The agent may reduce manual coordination, but a person should check the sources and approve messages or calendar changes before they affect others.

A careful pilot can proceed in stages:

1. Choose a bounded task. Start with a repeatable workflow that has a clear owner and low impact if delayed.
2. Limit access. Connect only the data and tools the task needs, and use a dedicated workspace where applicable.
3. Set action boundaries. Decide which activities may run automatically and which require explicit confirmation.
4. Test normal and unusual cases. Check how the agent handles missing information, conflicting instructions, and unexpected content.
5. Review outcomes before expansion. Confirm the work is accurate, authorized, and useful before allowing broader access or autonomy.

Microsoft Scout desktop app documentation describes a Frontier preview with prerequisites that include organizational setup and administrator controls. It also lists a Microsoft 365 work or school account, Microsoft 365 Copilot access, and a GitHub Copilot Business or Enterprise license. These are app-specific requirements and should not be assumed to describe every Autopilot experience.

The desktop app also uses a GitHub account for model access and token billing. A Microsoft 365 Copilot license alone should not be treated as proof that all Scout-related usage is included. Before a pilot, administrators should verify the applicable GitHub Copilot plan, model and usage policies, and how consumption is charged.

Administration can involve more than one team. Microsoft 365 administrators manage Frontier and Microsoft 365 tenant settings; endpoint administrators may configure the desktop app through Intune; GitHub administrators manage GitHub Copilot seats and access policies. If a user cannot sign in, check both Microsoft-side eligibility and GitHub-side licensing and policy. For a suspected product issue, use the organization’s established support route and verify the support terms for the specific preview and service involved.

• Autopilot: The current name Microsoft uses for the persistent personal-agent experience previously called Scout.
• Microsoft Scout (Frontier): The related desktop app name still used in Microsoft documentation, with local-device and Microsoft 365 integration.
• Copilot Chat: A conversational Copilot experience for questions, drafting, and other prompt-led tasks. Autopilot is oriented toward continuing delegated work. Access to work data and available capabilities depend on account, subscription, app, and tenant configuration.
• Microsoft 365 Copilot: A Microsoft 365 product experience with work-data grounding and app integration governed by user permissions and organizational controls. It is related to, but not interchangeable with, the Autopilot agent or the Scout desktop app.

Is Microsoft Scout still the current name?
Not consistently. Microsoft has described the personal-agent experience as Autopilot, while continuing to use Microsoft Scout for the Frontier desktop app.

Does Scout work only with Microsoft 365 data?
No such assumption is safe. The desktop app can also work with local files and browser or developer tools, subject to its configuration and permissions.

Is a Microsoft 365 Copilot license enough to use the desktop Scout app?
Not according to the documented prerequisites. The app also requires organizational access setup and a GitHub Copilot Business or Enterprise license.

Can Scout take actions without a person prompting each step?
Some agent workflows are designed to continue or run in the background. What can happen automatically depends on the product surface, configuration, and action controls.

Who should investigate a sign-in or access problem?
Start with the Microsoft 365 or endpoint administrator for Frontier and device configuration. A GitHub administrator may also need to check the user’s GitHub Copilot seat and organizational access policy.

The practical shift represented by Scout and Autopilot is from asking AI for an answer to delegating work that may continue across tools and over time. That can reduce coordination effort, but it also raises the importance of clear task boundaries, appropriate data access, and meaningful human review.

For organizations, the product name is only the starting point. Identify whether a deployment involves Autopilot, the Scout desktop app, or both, then evaluate the applicable permissions, billing, support, and governance controls for that specific experience.