Microsoft Scout Explained: What an Always-On AI Agent Means for Your Workday

What Microsoft actually announced

On June 2 at Build 2026, Microsoft introduced a new category of AI agents it calls Autopilots: always-on agents that work autonomously, carry their own identity, and act on your behalf inside the permissions your organization sets. ^[1] The first one is Microsoft Scout, and it lives inside the Microsoft 365 apps most of us already spend our day in.

Scout connects to Teams, Outlook, OneDrive, and SharePoint, plus the data underneath: chats, email, calendar, and contacts. You talk to it in Teams, and a desktop app extends its reach to your browser and local files. ^[1] Under the hood it runs on OpenClaw, an open-source agent framework Microsoft has adopted and is contributing back to, with a context engine called Work IQ that learns how you work over time. ^[1,5]

The same week, Microsoft also announced new in-house AI models and made its Work IQ APIs available to developers, which means third-party tools will be able to build on the same intelligence layer. ^[3,4] Taken together, it was the clearest statement yet of where Microsoft thinks office work is going.

Copilot vs Autopilot: the difference that matters

Names aside, here is the distinction in plain English. Copilot answers when you ask. You type a prompt, it responds, the exchange ends. An Autopilot does not wait to be asked. It stays active in the background, watches the flow of your work, and takes action on its own, then reports back. ^[1]

That is a bigger shift than it sounds. The mental model for Copilot was a very fast intern sitting next to you. The mental model for Scout is closer to a chief of staff: it has its own login, its own permissions, and a standing brief to keep your work moving even when your attention is elsewhere.

If the phrase “AI agent” still feels fuzzy, our plain-English guide to AI agents covers the foundations. Scout is that idea, shipped by the company that owns your inbox.

What Scout actually does day to day

Microsoft’s launch post lists concrete behaviours rather than vague promises, which I appreciated. According to Microsoft, Scout can: ^[1]

• Coordinate meetings proactively, including finding times across time zones, without you driving the back-and-forth.
• Flag important meetings and generate prep materials before you walk in, pulling from your own files and threads.
• Spot upcoming deliverables and block focus time on your calendar so the work actually gets done before the deadline.
• Surface risks early, like a decision that has stalled in a thread for a week, so you can unblock it before it becomes a fire.

Notice what these have in common: none of them are creative work. They are coordination, the connective tissue that eats hours without ever appearing on a to-do list. That is a smart target. It is also exactly the territory Microsoft has been moving toward with its agent products all year (we covered Agent 365 and Copilot Wave 3 as those pieces landed).

Who can use Microsoft Scout right now

Here is the honest part: probably not you, and not this month. Scout is rolling out as an experimental release to a select group of customers in private preview and to organizations enrolled in Microsoft’s Frontier program. Access requires Frontier enrollment, Intune policy configuration by your IT team, an opt-in attestation, and a GitHub Copilot license to install the experience. ^[1]

That is a long checklist, and it is deliberate. Microsoft has been testing Scout internally with its own employees first, and is clearly pacing the rollout while the trust and governance pieces mature. ^[1]

So treat today’s announcement the way you would treat a concept car at an auto show: you cannot drive it home, but it tells you exactly what next year’s models will look like. Always-on agents are coming to the standard Microsoft 365 experience the same way Copilot did: preview first, broad rollout later.

Two open questions worth tracking. Microsoft has not announced general availability timing or pricing for Scout, and history says both will shape who actually benefits: Copilot launched as a premium add-on before working its way into standard business plans. And because Scout reports into Teams, organizations that barely use Teams will feel the value later than Teams-heavy ones. If your company lives in Google Workspace, watch Gemini’s agent features instead; the direction is identical even when the logo is not.

The permission question your IT team will ask

The most consequential design choice in Scout is not a feature. It is identity. Every Scout agent operates under its own governed Entra identity (Entra is Microsoft’s system for managing who can access what), rather than borrowing yours or hiding behind a shared service account. ^[1]

In practice that means three things a non-technical manager should understand. First, everything the agent does is attributable: there is a record of which agent did what, under whose authority. Second, it can only touch resources you and your organization have explicitly approved, and sensitive actions can require a human sign-off before they go through. Third, existing data protection rules, like sensitivity labels on confidential files, still apply to the agent. ^[1]

Can it still go wrong? Of course. An always-on agent with calendar and email access is a new kind of risk surface, and no permission system removes the need for judgment about what you delegate. But “agent with its own audited identity” is a meaningfully better answer than “bot logged in as you”, and it is the answer your IT team will be evaluating.

Three ways to get ready before it reaches you

You cannot install Scout today. You can absolutely build the skills that will make you good at it the day it arrives. Three moves, all free:

1. Practice delegation with the agent features you already have. ChatGPT’s agent mode and Copilot’s existing tools both reward the same skill: writing a brief clear enough that someone else could execute it. Pick one recurring coordination task this week, like meeting prep, and write the instruction you would give an assistant. (Our guide to ChatGPT agent mode is a good practice ground.)
2. Decide your delegation list now. Write down the five recurring tasks you would hand to an always-on assistant tomorrow, and the two you never would. Knowing that boundary before the tool arrives beats discovering it afterward.
3. Get fast at reviewing AI output. The bottleneck in an agent workflow is not the agent. It is how quickly you can check its work and correct course. Skimming a draft agenda and spotting the wrong attendee in ten seconds is a real skill, and it compounds.

One more thing worth doing if you manage a team: start the delegation conversation now, before the tool forces it. An always-on agent raises questions a team should answer together. Which meetings can an agent schedule without asking? Who reviews what it sends? What happens when it blocks focus time over something a human considers urgent? None of these are technical questions, and all of them are easier to settle calmly in June than in a hurry on rollout day.

The professionals who did this with Copilot two years ago are the ones running circles around their colleagues now. Scout will reward exactly the same preparation. The watch list from here: a general availability date, pricing, and how much control admins get over what Scout can touch. We will cover each as Microsoft confirms it. Start practising before you have to.

Sana Mian, Co-Founder of Future Factors AI

Sana is an AI educator and learning designer specialising in making complex ideas stick for non-technical professionals. She has trained 2,000+ learners across corporate teams, bootcamps, and keynote stages. Future Factors offers AI Bootcamps, Corporate Workshops, and Speaking & Consulting for businesses ready to adopt AI without the overwhelm.

More about Sana →