Microsoft introduces Scout, an always-on personal agent for work

Microsoft is officially moving past the era of the passive chatbot.

At its annual Build conference on June 2, the tech giant introduced Microsoft Scout, debuting a new class of tools it calls Autopilots. Unlike traditional copilots that wait around for a prompt, Autopilots are always-on autonomous agents designed to work in the background with their own credentials.

Putting an enterprise harness on OpenClaw

Scout is built directly on OpenClaw, the open-source agent framework that went viral earlier this year before its creator was snapped up by OpenAI. While OpenClaw showed immense potential, it also terrified IT admins after instances of rogue agents misbehaving in inboxes.

Microsoft is attempting to tame that wild streak with enterprise guardrails. Each Scout instance operates under its own governed Microsoft Entra identity rather than a shady shared service account. That means every single action it takes is logged and attributable, permissions are strictly scoped to the task, and Microsoft Purview policies block unauthorized data transfers in real time. Microsoft is even contributing its policy conformance system back upstream to the OpenClaw community.

In practice, Scout acts like an omnipresent executive assistant. Users can name their personal agent—in one demo, the reporter dubbed his Sebastian—and let it run across Windows, the web, Teams, Outlook, OneDrive, and SharePoint. It connects to local files, calendars, and even Model Context Protocol servers.

Instead of waiting for instructions, Scout watches your workflow. It automatically negotiates cross-timezone meetings, flags stalled project decisions before they turn into bottlenecks, and proactively carves out deep-work blocks on your calendar ahead of major deadlines. Through a contextual engine Microsoft calls Work IQ, it learns your quirks and habits over time.

For now, Scout is in private preview. Getting access requires being part of Microsoft's Frontier program, configuring Intune policies, and holding an active GitHub Copilot subscription.

Why it matters

We think Microsoft's strategy here is brilliant for one big reason: it makes autonomous agents palatable to corporate risk managers.

Giving an AI broad write-access to your corporate inbox and calendar has historically been an IT security nightmare. By forcing Scout to carry its own verifiable Entra identity and subject every move to audit trails, Microsoft has built a blueprint for deploying autonomous agents inside risk-averse enterprises without terrorizing the compliance team.

For knowledge workers, this shifts AI value from answering prompts to reclaiming calendar sanity. The low-value administrative friction of scheduling, agenda prep, and chasing down stalled reviews eats hours every week. If Scout can take over those chores reliably, it will create an entirely new level of product stickiness.

That said, high friction remains. Tying early access to Frontier, Intune, and GitHub Copilot licenses means this won't hit mainstream desks overnight. And as OpenClaw's own history proved, granting an agent autonomy inside real communication channels always runs the risk of awkward hallucinations. Still, the line has been drawn: the AI race is no longer about who can talk best, but who can be trusted to run the background shift.