Google has brought Gemini to the Mac with a feature that lets it act directly on the computer. The assistant can use apps, files and websites to carry out multi-step tasks while the user keeps working in the foreground.
The agent clicks, types, moves through menus and fills in forms on the user's behalf. Each operation runs in an isolated window, and Gemini's chat shows a live preview so the work can be followed step by step. The user chooses which apps the assistant can touch and is asked before the most sensitive actions.

Image: Google Gemini — https://gemini.google/desktop/
Google points to typical uses: batch-editing photos, preparing listings for resale platforms, turning data and screenshots into a presentation. These are repetitive chores, exactly the ground on which the company wants to show the value of an operational agent.
The feature arrives as Apple tightens the rules on app access. With macOS 27.3, the system will more clearly warn about what data and actions an AI agent can see and perform. Users will be able to limit access to specific apps and folders, require permission each time, and see an indicator when one app controls another. Apps already granted permissions will not need them granted again after the update.

Image: Google Gemini — https://gemini.google/desktop/
Computer use is available in the U.S. to Google AI Plus, Pro and Ultra subscribers, with plans starting at $4.99 a month. It requires a Mac with Apple silicon. Google has not said when it will reach other countries or other platforms.
The real test will be reliability with complex, messy apps rather than tidy demos.

No comments yet. Be the first!