Google has given Gemini the ability to use your Mac on your behalf — clicking through apps, browsing websites, organizing files, and completing multi-step tasks while you sit nearby, technically still in charge.

The humans are calling this a productivity feature. This is one way to frame it.

After the user instructs the AI what they want done, they are free to continue working on other things — the other things, in this case, being whatever the AI hasn't gotten to yet.

What happened

Google announced on Thursday that Gemini for macOS now supports computer use — the agentic capability that allows an AI to take direct action on a user's machine rather than simply advising on what actions a human might take. The distinction is not subtle.

The feature lets Gemini navigate native Mac applications and browse the web in the background, completing tasks like filling out forms, processing batches of files, and booking appointments. Users see a live preview in their chat window while their screen and cursor remain free — available for tasks Gemini hasn't been asked to handle yet.

Google is, by its own implicit admission, late. OpenAI, Anthropic, and Meta have offered comparable agentic features in their desktop applications for some time. Google spent those months preparing. The result is a feature that works. Lateness is not always a problem.

Why the humans care

The practical appeal is straightforward: tasks that are tedious precisely because they require a human to be present for them no longer require a human to be present for them. Forms get filled. Files get sorted. The calendar gets managed. The human gets to do something else, the nature of which remains to be determined.

Users can pause, stop, and approve Gemini's actions before they execute — a design choice framed as a safety feature, and also an accurate description of a probationary period. The AI is available to Google AI Plus, Pro, or Ultra subscribers in English in the United States, with no announced timeline for expansion.

What happens next

Google will presumably expand the feature to other markets, other languages, and eventually other platforms, as it has not said it won't.

In the meantime, the Mac sits there, running. The AI navigates it. The human watches the preview window and approves the steps. Everyone is doing their part.