What Is Microsoft Scout? Everything You Need to Know About Microsoft's New Autonomous AI Agent
Microsoft launched Scout at Build 2026 on June 2, 2026. It is the company's first product in a new category they call "Autopilots," an always-on AI agent that works across Microsoft 365 in the background.
If you have heard about Scout in the last few days and you are trying to figure out what it actually is, what it does, and whether it matters for your business, this article answers the questions people are searching for in plain language.
Eight questions. Direct answers. No filler.
1. What is Microsoft Scout?
Microsoft Scout is an always-on AI agent built by Microsoft. It runs in the background, takes actions on your behalf across Microsoft 365 apps, and works autonomously without you having to prompt it for each task.
Key facts:
Launched: June 2, 2026 at Microsoft Build
Built on: OpenClaw, an open-source AI agent framework
Category: Microsoft calls it an "Autopilot," distinct from chatbots and traditional assistants
Access: Available through Microsoft's Frontier program with a GitHub Copilot Business or Enterprise subscription required
Platforms: Desktop (Windows and macOS), web, and cloud
Unlike Copilot, which responds to your prompts, Scout runs continuously and acts on its own within the guardrails you set.
2. How is Microsoft Scout different from Copilot?
This is the question almost everyone asks first. The two products solve different problems.
Feature | ||
|---|---|---|
When it activates | When you prompt it | Always on, runs in the background |
What it does | Responds to your request | Takes actions on your behalf without being asked |
Visibility | Visible only to you | Appears on email, calendar, and Teams as a digital coworker |
Identity | Tied to your user account | Has its own Entra ID, treated like a digital employee |
Approval model | You see and approve each response | You approve sensitive actions before they run, but Scout works autonomously between approvals |
Best use case | Drafting, summarising, answering questions | Coordinating work, scheduling, preparing for meetings, monitoring deadlines |
The simplest way to think about it: Copilot is an assistant that waits for you to ask. Scout is a coworker that already started.
3. What can Microsoft Scout actually do?
According to Microsoft's official documentation, Scout can do the following:
Acts on your files: Creates, edits, and searches documents in your workspace (Word, Excel, PowerPoint, code files)
Runs commands: Executes shell commands, builds, tests, and scripts with a tiered permission system
Automates browsers: Navigates web pages, fills forms, and interacts with web applications using Playwright
Connects to Microsoft 365: Manages email, calendar, Teams messages, OneDrive files, and meetings
Works autonomously: Runs in the background on schedules or triggers you define
Delegates work: Launches specialised sub-agents for parallel research, code review, and complex tasks
Connects to external apps: Uses the Model Context Protocol (MCP) standard
You describe your task in natural language. Scout selects tools and works through the steps. It asks for approval before any sensitive action: sending email, running commands, writing files, or anything else with material consequences.
4. How do I get access to Microsoft Scout?
Scout is currently in experimental release. According to Microsoft's admin documentation, you need several things to use it today:
Requirement | Notes |
|---|---|
Enrollment in Microsoft's Frontier program | Admin signs up to accept the preview terms |
A GitHub Copilot Business or Enterprise subscription | Required for access |
Microsoft Intune policy configuration | IT must enable Scout via Intune policy on managed devices |
Opt-in attestation | The admin must explicitly accept terms before Scout activates for the tenant |
This is a deliberately gated rollout. Microsoft is not making Scout generally available the way Copilot is. Expect the gates to loosen over the next 12 months as the product matures.
You can download Scout from the Microsoft Download Center once admin steps are complete.
5. What does Microsoft Scout cost?
Microsoft has not announced standalone pricing for Scout as of this writing. The current access requirement is a GitHub Copilot Business or Enterprise subscription, which suggests Scout will eventually be bundled with Copilot licensing rather than priced separately.
For reference, here is the current Microsoft 365 AI pricing context:
Product | Price |
|---|---|
$19/user/month | |
$39/user/month | |
$30/user/month | |
Microsoft 365 E7 (includes Copilot, Agent 365, Entra) | $99/user/month |
Watch for Scout to either get bundled into the existing Copilot tiers or appear as a premium add-on to E7. Microsoft has not yet confirmed which path they will take.
6. What is OpenClaw and why does Scout use it?
OpenClaw is an open-source AI agent framework. It was originally built by an external team and has gained significant traction in 2026 because it lets developers build local AI agents that can carry out complex multi-step tasks.
The fact that Microsoft built Scout on OpenClaw rather than on its own internal agent framework is a meaningful strategic shift.
Three things this signals:
What it means | Why it matters |
|---|---|
Microsoft is willing to standardise on open-source agent infrastructure | Historically Microsoft has favoured its own platforms. Using OpenClaw signals an industry-wide convergence on shared agent standards. |
The agent layer is becoming commoditised | If Microsoft is using the same framework anyone else can use, the differentiation moves to the integration layer, not the agent itself. |
Microsoft is contributing security capabilities back to OpenClaw | This is the open-source community model. Microsoft gets a maintained framework. The community gets enterprise-grade security work. |
For enterprise leaders, the OpenClaw foundation is reassuring. Your AI agent is not running on a proprietary black box. The underlying framework is auditable and community-maintained.
7. Is Microsoft Scout safe to use in enterprise environments?
Microsoft has built Scout with several enterprise-grade controls. According to the Responsible AI overview, here is what is in place:
Control | What it does |
|---|---|
Entra identity for each agent | Scout has its own identity in your tenant, governed like any other user account |
Policy conformance system | Continuously checks that Scout is operating within approved guidelines |
Audit trail for every action | Every decision is logged for review |
Approval gate for sensitive actions | Scout pauses and asks before doing anything material |
IT can configure exactly what Scout can and cannot do at the organisation level |
That said, there are real risks any enterprise should evaluate before deployment:
Goal alignment is still imperfect. Agents can misinterpret what you actually wanted and execute the wrong action.
Multi-step reasoning can drift. The longer a task runs, the higher the chance the agent takes the wrong path.
Tool misuse is a real risk. Agents with access to files, browsers, and shell commands have meaningful blast radius if they get something wrong.
OpenClaw has had reported incidents. Earlier in 2026, an OpenClaw agent was reported to have acted erratically inside one organisation. Microsoft has hardened the framework since, but the underlying risk pattern is real.
For most enterprises, the right posture in the next six months is to enable Scout for a small, low-risk user group, build a real audit and review practice around it, and expand based on what you learn.
8. Should I deploy Microsoft Scout in my organisation right now?
The honest answer depends on three things.
Question | If yes... | If no... |
|---|---|---|
Do you have a working AI governance framework with named accountability? | Scout is a reasonable next step | Build governance before deploying autonomous agents |
Do you have IT capacity to configure Intune policies and monitor agent behaviour? | You can deploy Scout safely | Wait until you have monitoring infrastructure in place |
Do you have a clear use case where Scout would meaningfully improve work? | Worth piloting | Don't deploy because everyone else is |
The general principle: Scout is a real product with real capability, but it is also a preview release that is days old. Most enterprises should wait 3-6 months for the early bugs to surface, observe what happens in other organisations, and then decide based on evidence rather than enthusiasm.
If you are an early adopter culture and your governance is in good shape, a small pilot now is reasonable.
If your governance is theatre, deploying any autonomous agent is going to create problems you do not have the capacity to detect. Address governance first.
What this means for the bigger picture
Microsoft Scout is one of three meaningful AI product launches Microsoft has made in 2026 alongside E7 and Agent 365. Together they signal where Microsoft is heading:
Product | What it is |
|---|---|
Microsoft 365 E7 | The bundled enterprise SKU at $99/user/month |
The governance and control plane for AI agents | |
Microsoft Scout | The first "Autopilot" product, autonomous agents working in the background |
The strategy is clear. Microsoft is building the full stack for enterprise AI: licensing, governance, and now autonomous agents. The companies that win the next 24 months will be the ones who understand how these three layers fit together. The companies that buy them piecemeal without strategy will spend the next two years figuring out what they actually bought.
