Microsoft Scout: The Always-On AI Agent That Works While You Sleep

Something shifted at Microsoft Build 2026, and I don’t say that lightly. Every year there’s a headline feature, a demo that gets the crowd going, and then a quiet return to the status quo. This year felt different. Microsoft Scout isn’t an incremental Copilot improvement. It’s a genuine architectural shift in how AI works alongside you, and once you understand what it actually does, you start to see why Microsoft is betting so heavily on it.

Let me break it all down.

What Is Microsoft Scout?

Scout is Microsoft’s first Autopilot agent, a new category of AI that Microsoft unveiled at Build 2026. The distinction matters. Copilot waits for you to ask it something. Scout doesn’t.

Think of it less like a chatbot and more like a digital colleague who’s always on, monitoring your priorities, watching your communications, and proactively moving work forward across your apps. It lives in the background, quietly keeping track of what matters, so that by the time you open Teams on a Monday morning, it’s already done a portion of the groundwork you would otherwise spend the first hour on.

Microsoft describes Autopilots as agents that “stay active in the background, understand how work gets done across your apps and systems, and take action without needing to be prompted each time.” That last clause, without needing to be prompted each time, is the key shift. This is AI as a persistent collaborator, not a search bar.

One small detail that caught my attention in the early demos: you give Scout a name. Microsoft’s own demo had a reporter naming their instance “Sebastian.” That’s not a gimmick, it’s intentional product design. The goal is a lasting working relationship with your agent, not a one-off interaction. You’re building a context-aware assistant that gets better at knowing you over time.

What Does It Actually Do?

Scout connects to the apps where knowledge work actually lives, Teams, Outlook, OneDrive, SharePoint, and pulls context from your chats, email, calendar, and contacts. From there, it can:

• Prepare for meetings, pulling together the relevant threads, documents, and open actions before you even open the invite
• Track deliverables and flag risks, monitoring conversations for commitments and surfacing anything that’s slipping
• Draft documents and follow-ups, producing first drafts based on the context it already has, not from scratch
• Monitor communications, scanning across email and chat for things that need your attention, rather than waiting for you to notice them
• Coordinate across apps and the web, via browser interaction and Model Context Protocol (MCP) integration with external services

Critically, Scout operates across cloud, desktop, and the web, it’s not siloed inside a single app. It follows the work, wherever the work happens to be.

For anyone who’s spent years stitching together Zapier workflows, calendar rules, and manual follow-up systems to stay on top of coordination-heavy work, that list above should feel significant.

How Does It Work Under the Hood?

Scout is built on OpenClaw, the open-source agent framework that’s been generating serious noise in the AI community this year. Building on an open-source foundation means community contribution, third-party validation, and a degree of architectural transparency that enterprise security teams will appreciate.

The identity model is worth understanding. Each Scout instance runs under its own governed Microsoft Entra identity, not a shared service account, not a bolt-on credential. This means Scout’s actions are fully attributable. Every decision it makes, every document it touches, every message it drafts, all of it is traceable back to a specific, auditable identity that sits within your existing Microsoft security posture.

Data protection flows through Microsoft Purview, so your existing compliance controls apply. The agent inherits your access model rather than creating a parallel one. For regulated industries, financial services, healthcare, public sector, that’s not a nice-to-have. It’s the difference between something you can actually deploy and something that never makes it past legal review.

On the governance side, sensitive operations require human approval before Scout acts. You set the priorities; Scout checks before anything high-stakes happens. It’s autonomous, but not unilateral, and that balance is clearly intentional.

Compliance certifications in scope include SOC 2 Type II, ISO 27001, GDPR, HIPAA, and FedRAMP, all handled through the Microsoft Trust Center. Scout also launched alongside Agent 365, Microsoft’s broader platform for tracking and managing AI agents across an organisation, which Microsoft expects to cover 1.3 billion agents automating office work by 2028. The scale they’re designing for here is extraordinary.

Why Would You Want It?

The honest answer is that coordination work is expensive, in time, in attention, and in the cognitive overhead that bleeds into everything else. If you’re running projects, managing a team, or operating in a high-volume communication environment, a meaningful portion of your day is spent on work about work rather than the actual work itself.

Scout targets exactly that layer. Meeting prep, follow-ups, tracking commitments across long email threads, surfacing the three things you actually need to act on from 200 unread messages, these are the tasks where Scout earns its keep. Early use cases that are showing real traction are coordination-heavy: recurring reports, cross-team follow-ups, meeting preparation cycles that currently require someone’s manual time every week.

There’s also a compounding argument. Scout gets better at knowing you over time. The more context it builds, your priorities, your working patterns, the people you care about hearing from, the more useful it becomes. This is fundamentally different from a stateless chatbot where every session starts from zero.

From an MVP and product-building perspective, and this is where it gets interesting for me personally, Scout represents the first serious production-ready example of what an always-on agent looks like inside an enterprise security model. Understanding this architecture now, while it’s in Frontier, is exactly the kind of early positioning that pays off when general availability arrives. The organisations that pilot Scout today will have a genuine operational advantage over those who wait for the case study.

Licensing and Pricing, What You Need to Know

This is the part that requires some careful reading, so let me be direct about what’s currently required and where the gotchas are. I’ll use real UK GBP figures throughout because the US dollar list prices that most coverage leads with are not what UK organisations actually pay.

Scout is currently in experimental release via Microsoft’s Frontier program, their invite-only early-access track for groundbreaking features. To access it, you need to clear two separate admin gates:

• Gate 1: Enrol your organisation in the Frontier program through the Microsoft 365 admin centre
• Gate 2: Configure an Intune policy on target devices, and complete a Frontier organisation sign-up attestation form

On top of both gates, each user needs an active GitHub Copilot Business or Enterprise licence. Scout uses your GitHub Copilot licence for token billing from June 2026 onwards, following GitHub’s shift to usage-based AI Credits. Every user also needs a GitHub account before they can sign in. Microsoft’s own docs are explicit: Frontier access without a Copilot licence won’t work, and a Copilot licence without Frontier access won’t work either. Both gates must be complete before a single user can sign in.

The Base Licence Floor

Because Scout requires Intune-managed devices, Microsoft 365 Business Standard (currently around £9.40/user/month) is ruled out. The minimum starting point is Microsoft 365 Business Premium at approximately £17.60/user/month (annual commitment, ex. VAT), which includes Intune and is capped at 300 users.

For enterprise organisations over 300 users, or those that need stronger compliance and security capabilities, the relevant tiers in GBP as of May 2026 are:

• Microsoft 365 E3: approximately £32.55/user/month (annual, paid monthly)
• Microsoft 365 E5: approximately £51.45/user/month (annual, paid monthly), the tier most commonly referenced for Scout deployment in regulated industries
• Microsoft 365 E7 Frontier Suite: approximately £85.68/user/month (annual, paid monthly), the new top-tier SKU launched May 2026 that bundles E5, Microsoft 365 Copilot, Entra Suite, and Agent 365 into a single licence. This is the cleanest single-licence path to Scout for enterprise customers, and represents roughly a 15% saving versus buying the components separately.

Note that E5 prices are increasing by approximately 5% from 1 July 2026, and E3 is rising around 8%. If your renewal is after that date, it is worth auditing your licence mix before you sign.

Adding GitHub Copilot

Every Scout user also needs a GitHub Copilot seat on top of their M365 licence. The current seat prices are:

• GitHub Copilot Business: $19/user/month (approximately £15/user/month)
• GitHub Copilot Enterprise: $39/user/month (approximately £31/user/month), though Enterprise requires GitHub Enterprise Cloud as a prerequisite at $21/user/month, pushing the real all-in cost to around $60/user/month (approximately £47/user/month) at that tier

For most Scout deployments, Copilot Business is the right tier. Enterprise is only warranted if your organisation also needs GitHub Enterprise Cloud for its development workflows.

What Does It Actually Cost Per User?

Putting it together for a typical UK SME on Business Premium piloting Scout with a small team:

• Microsoft 365 Business Premium: ~£17.60/user/month
• GitHub Copilot Business: ~£15/user/month
• Microsoft 365 Copilot add-on (required for Frontier): ~£24/user/month (the enterprise Copilot add-on at $30, converted)
• Approximate total: £56/user/month before VAT

For a 20-person pilot team, that is roughly £13,440/year before VAT, plus setup time for Frontier enrolment and Intune configuration. For an enterprise organisation already on E7 at £85.68/user/month, the GitHub Copilot Business seat is the only meaningful addition, bringing the all-in cost to around £100/user/month.

Microsoft has not yet published a specific Scout add-on price beyond these base requirements. That may change as Scout moves out of Frontier preview, so treat the above as the floor rather than the ceiling.

The Honest Caveat

The licensing stack is genuinely complex for smaller organisations, and I’d rather say that clearly than gloss over it. If you are on Business Standard today, getting to Scout requires upgrading to Business Premium, adding the M365 Copilot add-on, and adding GitHub Copilot Business. That is three separate changes to your licensing stack before anyone can even attempt to sign in. Do the maths before committing a team to this, and factor in the July 2026 price increases if your renewal sits close to that date.

For mid-market and enterprise organisations already on E5 or E7 with Intune deployed, the path is much cleaner. The E7 Frontier Suite is genuinely the most straightforward route for organisations serious about agentic AI at scale. And with Frontier updates running on a bi-weekly cadence with major feature drops monthly, this is moving fast. The licensing model will almost certainly simplify as Scout approaches general availability.

The Bottom Line

Microsoft Scout is the clearest signal yet that the chatbot era of enterprise AI is giving way to something more ambitious, agents with governed identities, persistent context, and the ability to take action across systems without waiting to be asked. The architecture is solid, the security model is enterprise-grade, and the use cases are real.

The licensing complexity is a legitimate friction point today, and worth planning around rather than ignoring. But for organisations already invested in the Microsoft 365 ecosystem, especially those running knowledge-heavy, coordination-intensive workflows, Scout is worth getting in front of now.

Frontier enrollment is open. The window to be an early adopter, and to build the institutional understanding that comes with it, is right now.

If you’re exploring agentic AI for your organisation or want to talk through the MVP angle, I’d love to hear what you’re building. Drop a comment or find me on LinkedIn.