Gemini Now Lets Mac Users Delegate Tasks to AI Agent

Google has unlocked a new level of AI autonomy for Mac users. The company is rolling out a feature in its Gemini desktop app for macOS that lets its AI assistant take hands-on control, performing tasks across the system—manipulating native Mac apps, browsing websites, and interacting with both files and the web—on behalf of users. This marks a big step forward for Gemini’s capabilities.

With this update, Gemini can follow multi-step instructions involving local and online workflows. Once a user issues a command, the AI displays its progress in a live preview inside the chat window, while the user’s screen and cursor remain free for other work. Control features are built in: users will be able to approve, pause, or stop actions before Gemini carries them out.

What kinds of tasks are now possible?

The feature is framed as especially useful for repetitive tasks or workflows. For example, users could have Gemini fill out forms, organize files into folders based on criteria they set, or build slideshows using content from dashboards or local storage. It’s well suited to anything from batch file management to automating tedious web-based or app-based actions.

For now, this capability is limited in scope. It’s available only in English, in the U.S., and requires a Gemini app for Mac plus a subscription—either AI Plus, Pro, or Ultra. Google hasn’t yet revealed when it will widen access to more languages, platforms, or global markets.

This comes shortly after Google opened up Gemini’s “Tasks” feature more broadly. That tool lets the AI agent log into websites, make online purchases, or arrange appointments. Google also recently kicked off early access for its revamped “CC” agent, first unveiled in September, which is designed to help users manage email, calendars, chats, and tasks.

How Gemini measures up to rivals

This new agentic control puts Gemini more in line with functionality already seen in competitor offerings. Other models from OpenAI, Anthropic, and Meta have long supported actions directly on users’ computers. Google’s inclusion of such computer-use capabilities brings Gemini up to speed in this respect.

The live preview, action-approval controls, and cross-app functionality all help Google ensure safety and user control remain central. Empowering the AI to move beyond suggestion mode into autonomous action is powerful—but also requires careful guardrails.

What this means and what to watch

Google’s move signals a shift in how AI assistants are expected to function: not just as helpers, but collaborators capable of executing tasks on users’ machines. For Mac users, this promises productivity gains—especially for workflows that are repetitive, complex, or span local and online resources. However, expanded access depends on language support, platform availability, and how transparent and reliable the approval and safety controls prove to be. Apple users and observers should monitor how well Gemini balances empowerment with control—and whether Google matches competitors in extending agentic AI across platforms.