Microsoft is building dedicated PCs for AI agents, what does it mean?

For years, a PC has been a machine people sit in front of to get work done. But Microsoft thinks that could change as AI agents become a bigger part of the workplace. At Microsoft Ignite 2026, the company announced Windows 365 for Agents, a service that gives AI agents their own Cloud PCs. This means you don't need to buy physical computers. Instead, they are secure Windows environments hosted in Microsoft's cloud, where agents can open applications, interact with screens, click buttons, type and move information between different programmes. The idea is pretty basic: if an AI agent needs to work like a human employee, why not give it a computer of its own?

What exactly is Windows 365 for Agents?

How do you normally use your PC at work? You open a browser, log into your company's software, download files, update spreadsheets and move information between applications. Most of these tasks involve interacting with a screen, rather than simply asking software to exchange information through an API. AI agents can already handle several such tasks, but they often run into problems when an application doesn't have an API or wasn't designed to work with AI. This is where Windows 365 for Agents comes in.

Microsoft is essentially giving these agents their own Windows-based Cloud PCs. An agent can see what's on the screen, click on buttons, type into fields and move data between applications, much like a person would. This means it can also work with older business applications, browser-based tools and on-premises Windows applications, including those that don't offer APIs.

Since these Cloud PCs are hosted remotely, organisations don't need to provide a separate physical computer for every agent. They can provision and manage them through the same systems they already use for employee devices. Microsoft says Windows 365 for Agents Cloud PCs are connected to Microsoft Entra, its identity and access management service, and managed through Microsoft Intune. This means an organisation can treat an agent's computer much like any other PC in its fleet, with centralised management and security controls.

The service is available as a computer-use tool in Microsoft Copilot Studio. Microsoft also plans to bring it to Microsoft Foundry in preview early next year. It is now eligible for Copilot Credits, which brings it into Microsoft's common billing and cost-management system for eligible AI services.

There is another interesting development here. Microsoft is working on persistent Cloud PCs for agents that need to run for longer periods without stopping. These are planned for Frontier private preview, and might be useful for businesses that want agents to handle ongoing tasks rather than simply respond to individual requests.

Microsoft says its own business operations team is already using Windows 365 for Agents to process orders. The company claims this has reduced processing time by 60%. That's a useful example of where the technology could make a difference, particularly in businesses that handle large volumes of repetitive transactions.

What happens when an AI agent gets access to your computer?

Giving an AI agent its own PC is one thing, but what happens when it starts accessing files or applications it shouldn't? An agent could make a mistake, open the wrong file or access information that it was never supposed to see. This is something Microsoft needs to address if businesses are going to trust AI agents with everyday work.

Microsoft is working on this with a set of security features designed to control what AI agents can access and do on a Windows PC.

One of these is Microsoft Execution Containers (MXC), which has been available on Windows 11 since October. Think of it as giving every AI agent its own restricted workspace. Businesses can decide which files, networks and system resources an agent can access, instead of giving it free rein over the entire PC. So, if an agent is supposed to work with files in one particular folder, it shouldn't be able to wander into other folders without permission.

At Ignite, Microsoft announced a few more additions to MXC. Process container improvements are now in preview, while learning and permissive modes are generally available. These modes let IT teams first see how an agent behaves and which system resources it needs before deciding how strictly to restrict it. This way, they can set security rules without accidentally blocking tasks the agent needs to perform.

Microsoft is also testing something called Local Micro VMs. These use hardware-based virtualisation to keep an agent's work separate from the rest of the system. This is meant to offer an additional layer of protection, particularly for tasks that need tighter security. However, the feature is still experimental.

There's also the question of knowing who did what. If an employee updates a file, it's easy to identify them. But what if an AI agent does it? Microsoft is introducing agent identities so businesses can tell the difference between actions performed by people and those carried out by AI. This also helps organisations decide which applications and resources an agent is allowed to access. Support for using these identities with on-premises applications in Windows 365 for Agents is currently in preview.

Microsoft also wants businesses to manage these agents using tools they may already be familiar with, including Entra, Intune, Defender, Purview and Agent 365. This should make it easier for IT teams to keep track of what their agents are doing without having to start from scratch.

Microsoft says OpenAI has enabled MXC by default in Codex, while Meta is using it across tens of thousands of Windows devices. More than 20 developers, including Anthropic, Box, Perplexity and Replit, are also building on the platform. That's a sign that Microsoft isn't looking at this as a feature for its own AI tools alone. It wants other AI companies to use the same security framework too.

So, will AI agents become a regular part of our PCs?

Windows 365 for Agents is still a new idea, and it will take time before businesses start using it widely. But the concept is interesting. Imagine having an AI agent that can log into your work applications, update records, move files around and finish routine tasks while you focus on something else.

Of course, businesses will have to be careful about what these agents can access and what they're allowed to do. But Microsoft is clearly betting that AI agents will become a regular part of how people work. And if that happens, having a separate Windows PC running in the cloud for an AI agent might not sound so unusual after all.

Aman Rashid

Aman Rashid is the Senior Assistant Editor at Digit, where he leads the website along with the brand’s YouTube, social media, and overall video operations. He has been covering consumer technology for several years, with experience across news, reviews, and features. Outside of work, Aman is a sneaker enthusiast and an avid follower of WWE, Dragon Ball, and the Marvel Cinematic Universe.

View full profile