At Microsoft Build on 2 June 2026, the company introduced a new category of agentic AI it calls "Autopilot." The first agent in this line, Scout, is designed to operate continuously in the background, observing how work happens across applications and taking action without waiting to be prompted each time. The announcement marks a deliberate move beyond the assistive model that defined Copilot toward something more autonomous.
For enterprise buyers and suppliers, Scout is not just a product launch. It is a signal about where the major platform vendors believe agentic AI is heading. The question is no longer whether agents can answer questions or draft emails. It is whether enterprises are ready to delegate ongoing operational tasks to software that acts on its own.
What Scout does
Scout is an always-on agent that connects across cloud, desktop and web environments. According to Microsoft, it integrates with Teams, Outlook, OneDrive and SharePoint, and draws on calendar, email, chat and contact data to understand the context of a user's work.
The agent can perform several tasks autonomously. It can schedule meetings while accounting for time zones and flag meetings it considers particularly important. It can generate preparation materials ahead of scheduled events. It can identify approaching deadlines and block time on a user's calendar for focused work. It can also spot what Microsoft describes as "stalled decisions" and surface risks that might otherwise go unnoticed.
Crucially, Scout operates with its own Entra identity. This means its actions are attributable and can be governed through the same access controls that organisations already use for human users. Microsoft says organisations can set constraints on what Scout is permitted to do, though the full detail of those controls has not been published.
Why this matters for enterprise buyers
The shift from Copilot to Autopilot reflects a broader industry trajectory. Copilot-style assistants require explicit user prompts for every action. Autopilot-style agents are designed to act proactively, making decisions and executing tasks based on continuous observation of work patterns. This is a meaningful architectural difference with meaningful governance implications.
Continuous access raises continuous governance questions. An agent that reads email, monitors chat, watches calendar and acts across multiple systems requires broad and persistent permissions. Enterprises evaluating Scout or similar agents will need to understand how those permissions are scoped, how they can be revoked, and what happens when the agent encounters data it should not see. The Entra identity model is a start, but buyers should expect to ask detailed questions about runtime policy enforcement, audit logging and data residency.
The boundary between helpful and intrusive is thin. Scout's ability to block calendar time, flag meetings and generate prep materials could reduce administrative overhead significantly. It could also create friction if the agent's judgement about what matters does not align with the user's own priorities. Enterprises piloting autonomous agents should plan for feedback loops that let users correct agent behaviour and refine its decision-making over time.
Trust in the underlying platform becomes critical. Scout is built on OpenClaw, the same platform that powers other Microsoft AI features. Enterprises already using Microsoft 365 will need to assess whether they are comfortable expanding their reliance on this stack for autonomous operations. The security and reliability track record of the underlying platform becomes a more significant factor when the software is acting without direct human oversight for each action.
What this means for suppliers
For companies building competing agent platforms, Microsoft's Autopilot category creates both pressure and opportunity.
The pressure is that a major platform vendor is now explicitly marketing autonomous agents as a product category. Buyers will increasingly expect proactive, background-running agents as a standard feature, not a premium add-on. Suppliers that only offer prompt-driven assistants may find themselves positioned as a generation behind.
The opportunity is that Scout is in very limited preview. Only a select group of customers and organisations enrolled in Microsoft's Frontier early-access programme currently have access. Even within Frontier, access requires a GitHub Copilot subscription, which recently moved to usage-based billing. This creates a window for alternative agent platforms to establish themselves with buyers who want autonomous capabilities now, or who prefer not to deepen their dependence on the Microsoft stack.
Suppliers should also note the governance emphasis. Microsoft is leading with Entra identity, access controls and organisational policy constraints. This suggests that enterprise buyers are already asking these questions, and that any competing agent platform will need comparable governance features to be taken seriously in procurement.
What enterprises should consider before piloting
Scout and the Autopilot category are not yet generally available, but the direction is clear. Enterprises planning agent pilots in 2026 should use this moment to review their readiness for autonomous software.
Key considerations include whether identity and access management systems can handle non-human actors with the same rigour as human accounts; whether audit and compliance functions can track agent decisions and attribute them correctly; whether data classification and sensitivity labelling are robust enough to prevent autonomous agents from overreaching; and whether users have clear channels to override, correct or disable agent actions when they disagree with the agent's judgement.
The Agentic Expo angle
Agentic Expo exists because the transition from assistive AI to autonomous agents is happening faster than most enterprise governance frameworks can adapt. Microsoft's Scout announcement accelerates that transition. It makes the question concrete: are you ready to let software act on your behalf, continuously, across your most sensitive work systems?
The buyers who attend Agentic Expo in March 2027 will not be looking for demos of chatbots that answer questions. They will be looking for agents that can run operations, with the governance, security and auditability to earn their trust. Scout shows what is possible. The exhibitors at Agentic Expo will show what is market-ready.