Microsoft expands Copilot on Windows with AI agent capabilities, vision and voice wake-up

Microsoft is expanding Copilot across Windows 11 with the global rollout of Copilot Vision, the “Hey Copilot” voice activation, and Copilot Actions, which brings the assistant’s AI agent capabilities to the desktop. Copilot Actions allows the assistant to perform tasks directly on the device through voice, such as opening apps, typing, scrolling, or completing multi-step actions like editing files or sending emails. It runs inside a secure “Agent Workspace” with restricted permissions and always requests user confirmation before accessing local files or making changes.

The “Hey Copilot” wake-up feature also lets users activate the assistant hands-free by saying the phrase, followed by a chime and a small microphone window ready for commands. The feature is off by default and must be enabled manually, requiring a microphone and internet connection.

Copilot Vision is now available in all markets, letting users share their desktop or apps for real-time analysis and contextual help. It can interpret on-screen content, answer questions, and guide users through actions using the Highlights feature. When paired with Microsoft 365 apps such as Word or PowerPoint, Copilot can analyze full documents, not just visible sections. Microsoft is also testing a “Text-in Text-out” mode for typed interactions during Vision sessions.

Lastly, Windows 11 now integrates Copilot into the taskbar’s search bar, opening quick results for apps, files, and settings along with shortcuts to Copilot tools. Microsoft notes that these additions don’t give Copilot direct access to personal data, as they rely on existing Windows Search APIs and only interact with files or apps when users give explicit permission.