Gemini could soon act more like a coworker than a chatbot

Google’s Gemini AI has mostly been used for writing emails or summarizing documents. However, a recent leak hints that big changes may be coming soon.

TestingCatalog has spotted Google testing a new “Agent” tab in Gemini Enterprise, signaling the company’s intention to move toward more complex, multi-step workflows. This change also puts Google in direct competition with Anthropic’s Claude Cowork. More importantly, it’s a strong indication that the next phase of AI is about managing your apps and handling complex tasks for you.

At present, most people use tools like ChatGPT or Gemini for single tasks. You enter a prompt, get a response, and that’s it. With agentic AI, you can set a broad goal. The agent will then break down the task and use local files and connected services to complete everything on its own, without needing your constant input.

Early previews shared by TestingCatalog reveal that Google’s upcoming workspace is built specifically for this kind of heavy execution. There’s a new Agent tab next to the usual chat interface, which includes a dedicated “Tasks” section and an inbox. On the right, you’ll find panels for your “Goal,” “Agents,” “Connected apps,” and “Files.”

It appears, though, that Google isn’t letting these agents act freely on your computer. A “Require a human review” toggle suggests the system will stop and wait for your approval before doing anything sensitive. This feature shows that Google is getting ready for real desktop-level task management, where mistakes could have serious consequences.

There are also signs that this setup will include features like reusable Skills and repeating schedules. This means you could set up ongoing workflows that run in the background while you work on other tasks. With Google I/O on the horizon, the broader strategy is coming into focus.