In Microsoft Copilot Studio, you can now create AI agents that can work with any applications or sites that have a graphical interface.
The new feature, called Computer use, will allow agents to treat websites or computer programs as tools. They will be able to click any buttons, select menus, enter text into on-screen fields, and perform other actions with systems where there is no available API to connect directly to the system.
Agents created in Copilot Studio will be able to automatically adapt to changes in applications and websites. According to information on the Microsoft website, they will be adjusted in real time thanks to the built-in reasoning function, allowing them to independently solve problems without interruption of work.
Such agents can be used for various tasks, such as automating data entry, market research, or invoice processing.
Incidentally, Microsoft previously introduced Actions for regular Copilot users, which allows you to do roughly the same things. Thanks to Actions, users will be able to ask a chatbot to perform a variety of tasks on many websites in the browser. OpenAI also introduced a similar feature earlier this year.