ROW EDITION · updated Aug 8, 2026

Microsoft Scout / Autopilot deep dive

Always-on agents change the governance question: identify the owner, boundary, enforcement, and receipt.

← All projector / phone kit pages

QR code for this page

Open this page on your phone

Program hub

Microsoft Scout / “Copilot moves to autopilot”

Autopilot — always-on agent, Frontier preview. The workshop's question is not whether autonomous agents are impressive. It is what changes when a tool acts without waiting for a prompt.

  • Microsoft described an OpenClaw-like enterprise agent in April. TechCrunch
  • At Build, Scout was introduced as an “autopilot”: active in the background and able to act without a prompt each time. Computerworld
  • Coverage describes connections to Teams, Outlook, OneDrive, SharePoint, browser, and external apps; it is gated through the Frontier program with policy / attestation requirements. TechCrunch
  • Execution containers and a large security scanning harness are described as defenses. Cybernews

Six-layer mapping

Workshop layerScout / Autopilot defense
1 · Policy & inventoryGoverned Entra identity: a named agent, not an anonymous service account.
2 · Development & testingSecurity scanning harness with 100+ specialized agents.
3 · Pipeline gatesFrontier preview, Intune policy, opt-in attestation before operation.
4 · Runtime enforcementExecution container: a bounded environment for action.
5 · Observability & intentPolicy conformance system checking behavior against guidelines.
6 · Evidence & attestationConformance checks generate audit trails — receipts by design.

Even the company shipping the autopilot will not let it fly without a named identity, a container, a conformance check, and a receipt. Neither should you.

Caution: an autonomous agent's inbox behavior can be surprising; use the HARMS worksheet to name the humans, defenses, owner, and receipt before enabling action.