Microsoft has rolled out a new feature in Copilot Studio called ‘Computer use’. It allows AI agents to interact with websites and desktop apps the same way a person would—clicking buttons, typing into forms, and choosing options from drop-down menus.
What is Copilot Computer Use?
The tech giant claims that think of it like this: if a person can use a piece of software, so can a Copilot agent now. And it does not need APIs to do it.
The new feature is part of an early access research preview, and it is designed to help businesses automate tasks that previously needed human hands. Whether it is entering data into an old desktop app or collecting information from a website, Copilot agents can now handle it using just the screen—no behind-the-scenes integration required.
One of the big draws here is that the AI can adjust to changes on the fly. So if a button moves or a page layout changes, the agent can figure it out and keep working without anyone needing to fix it. That is a common headache with older automation tools, especially in robotic process automation (RPA), which often breaks when something even slightly changes.
This all works through Microsoft’s own cloud, so companies do not need to worry about setting up their own servers, suggests the company. Additionally, Microsoft says enterprise data stays put and is not used to train any of its large models, which should ease some of the usual privacy concerns.