Microsoft Scout & the Autopilot Era
A New Agent Category: Autopilots
At Microsoft Build 2026 (June 2–3), Microsoft introduced Autopilots — a new category of agents that are always-on, act with their own identity, and work autonomously in the background without needing to be re-prompted for every step. This is a distinct pattern from conversational Copilot Studio agents, which respond turn-by-turn inside a chat.
Microsoft Scout: The First Autopilot
Microsoft Scout is Microsoft's first Autopilot agent — an always-on personal agent for work. It's a desktop application (Windows 11+ and macOS 12+) that stays integrated with your Microsoft 365 flow of work: Teams, Outlook, OneDrive, and SharePoint, plus the chats, email, calendar, and contacts that power your day. From the desktop app it also reaches your browser, local files and shell, and Model Context Protocol servers.
| Capability | What it does |
|---|---|
| Acts on files | Creates, edits, and searches Word/Excel/PowerPoint and code files in your workspace |
| Runs commands | Executes shell commands, builds, tests, and scripts under a tiered permission system |
| Automates browsers | Navigates pages and fills forms using Playwright |
| Heartbeat mode | Periodic background check-ins (every 15–120 minutes) that run prompts while you're away |
| Automations | Scheduled or condition-triggered tasks that execute independently |
| Delegates work | Launches specialized sub-agents for parallel research, code review, or complex tasks |
Powered by Work IQ
Microsoft Scout builds context over time using Work IQ — learning how you work, what you care about, and what needs to happen next, so it becomes more useful and aligned to your priorities the longer it runs. Microsoft also shipped standalone Work IQ APIs (GA, Build 2026) so any first- or third-party agent — not just Scout — can plug into the same real-time work-context signal.
Built on OpenClaw — With Enterprise Guardrails
Microsoft Scout is powered by OpenClaw, an open-source agent technology, and Microsoft is contributing policy-conformance work back upstream so organizations running OpenClaw themselves can get a verifiable, audit-ready compliance answer. On top of that open-source core, Microsoft layers the enterprise controls that make it safe to run at scale:
- Every agent runs under its own governed Microsoft Entra identity — never a shared, anonymous service account — so every action is attributable to a known actor.
- Credentials are scoped to the task at hand and redacted from logs and diagnostics.
- Sensitive actions (sending email, running commands, writing files) require your explicit approval before they execute.
- Microsoft Purview sensitivity labels and data-loss-prevention policies are enforced in the moment, before anything is sent or written — Scout operates within these controls, it doesn't bypass them.
Scout vs. Copilot Studio Agents vs. Copilot Cowork
| Product | Pattern | Best for |
|---|---|---|
| Copilot Studio agent | Conversational, built/governed by makers, published to Teams/web/M365 Copilot | Purpose-built business agents (HR, IT helpdesk, support) |
| Microsoft Scout | Always-on Autopilot with its own identity, runs on your desktop | Personal work coordination — scheduling, prep, follow-through |
| Copilot Cowork (GA June 2026) | Multi-agent teamwork surface inside Microsoft 365 Copilot | Delegating whole work packages to a team of agents |