Microsoft Scout and Autopilots: What Autonomous Agents Mean for Professional Services

At Build on 2 June 2026, Microsoft announced Scout and positioned it as something more than another chat tool. It called it its first real personal assistant and the first product in a new category it calls Autopilots: always-on agents that run in the background, hold their own identity and act on your behalf within set permissions.

Scout follows Copilot Tasks in February and Copilot Cowork, but it is a genuine step change. Those were still prompt by prompt. Scout is autonomous.

For UK professional services firms, the marketing is less interesting than the direction of travel. This piece looks at what Scout actually is, where the honest catches sit and why, despite them, it is a good direction to see.

What Scout actually is

Scout is an always-on agent integrated across Teams, Outlook, OneDrive and SharePoint, grounded in your day to day flow of work. It runs across cloud, desktop and web, with the current release being a desktop app for Windows 11 and macOS 12 or later.

It is powered by Work IQ, the context layer that learns how you work, what you care about and what needs to happen next, so it should get more useful over time. We have written before about why your Microsoft 365 data setup decides how useful Work IQ can be and that point matters even more for an agent that acts rather than answers.

One detail stands out for a Microsoft product. Scout is built on OpenClaw open source technology, with Microsoft contributing policy conformance upstream rather than forking its own version.

How it differs from Copilot

This is the part worth slowing down on.

Copilot lives inside the apps and waits to be asked. Scout sits above them, watches what matters and acts without being prompted each time. Microsoft’s own framing captures the shift well: this is a different kind of AI, where you might get a phone call from your assistant rather than typing into a box.

The desktop app also reaches further than Copilot. It can touch files, the shell, the browser, developer tools and Model Context Protocol servers, not just Microsoft 365 content. That is powerful and it is exactly why the governance question gets louder.

What it can do

The concrete examples are recognisable to anyone running a busy professional services team:

• Proactively schedule and coordinate meetings across time zones and flag the important ones.
• Identify upcoming deliverables, then block calendar time so they actually get done.
• Generate prep materials such as briefs, reports and slide decks ahead of meetings.
• Spot risks like stalled decisions before they turn into blockers.

Microsoft says more than 3,000 of its own employees already use Scout for scheduling, paperwork, booking travel and filling in forms. That is a useful signal of where the value lands first: the administrative friction that surrounds skilled work, rather than the skilled work itself.

The governance and security story

This is the section that matters most for firms with client confidentiality obligations and Microsoft has clearly thought about it.

Every Scout agent runs under its own governed Entra identity, not a shared service account, so every action is attributable to a known actor. Credentials are scoped to the task, redacted from logs and managed like any first party Microsoft service.

Microsoft Purview controls apply in the moment. Sensitivity labels and data loss prevention are enforced before anything is sent or written. Sensitive actions can require a human to sign off and by default auto approve is off, so every action prompts for confirmation.

Importantly, Microsoft treats the OpenClaw layer as untrusted. It is sandboxed with no access to secrets or Microsoft 365 data, controlled through Agent 365, Purview and Defender.

It is a serious architecture. It is also, by Microsoft’s own admission, new.

The catches and the honest reality

A credible read of Scout has to include the parts the press release glosses over.

Availability is genuinely tiny. At launch it is limited to Frontier-enrolled organisations and a small group of private preview customers. Getting access means jumping through hoops: Frontier enrolment, Intune policy configuration, an opt-in attestation and GitHub Copilot licensing on top of Microsoft 365 Copilot.

There is a real data consideration too. The preview includes data paths outside Microsoft 365 through GitHub-based inference, which is exactly why the admin attestation is mandatory. Any firm handling confidential client information should flag that clearly before going near it.

And the trust question is open. Microsoft’s own chief executive once said that for the company to launch OpenClaw “would be considered Microsoft launching a virus” and Scout is now built on it. Even sympathetic coverage notes the safeguards are unproven in the wild. Handing an agent this much autonomy still takes a leap of faith.

This is an experimental preview with no general availability date. It is a watch and prepare story, not a deploy now one.

What this means for UK professional services firms

The direction of travel is clear. Agents are moving from assistants you prompt to colleagues that act. The right time to think about governance is before the technology lands, not after.

The prerequisites tell you who benefits first. Entra identity, Intune policy, attestation and layered licensing all assume a tidy foundation. The firms that get value early will be the ones whose Microsoft 365 setup and policies are already in good shape. The same readiness that makes Copilot work with client data safely is the readiness that makes an Autopilot safe.

There is a competitive angle as well. Google is pushing Gemini Spark into Workspace, so this is shaping up as a race to own the enterprise personal assistant. That competition tends to accelerate the timeline, which is another reason not to be caught flat footed.

For a Colchester accountancy practice or a regional law firm, the practical question is not whether to buy Scout this quarter. It is whether your identity, sensitivity labelling and data loss prevention could safely support an autonomous agent the day it reaches general availability. Most firms are not there yet.

Why this is a good direction to see

It is fair to be sceptical of an experimental preview, but the underlying shift is a healthy one.

For years, capable professionals have lost hours to coordination: chasing diaries, assembling packs, remembering the deliverable that has gone quiet. That work is necessary and almost entirely undifferentiated. Moving it to an accountable, governed agent frees skilled people to do the judgement work clients actually pay for.

Just as important, Microsoft is leading with identity, attribution and human sign off rather than raw capability. An agent that acts under its own Entra identity, inside Purview controls, with auto approve off by default, is a far more defensible model than a free roaming bot. If autonomous AI is coming either way, and it is, this is roughly the shape you would want it to take.

The future Scout points to is not AI that replaces the professional. It is AI that clears the runway in front of them, under rules the firm sets and can audit. That is worth preparing for properly.

How FiveForward helps

Autonomous agents are only as safe as the governance underneath them and most firms are not yet ready on identity, sensitivity labels and data loss prevention.

As a Microsoft 365 consultancy focused on professional services, we help firms get there deliberately rather than chasing the hype. A practical first step is a readiness assessment that checks whether your Entra, Intune and Purview setup could safely support an Autopilot like Scout when it reaches general availability, alongside the wider Copilot agents groundwork.

If you want to be ready for autonomous agents before they land rather than after, book a consultation and we will tell you honestly where your foundations stand.