Google introduces Gemini Agents for autonomous tasks

Google has outlined a future in which AI agents can take on extended tasks with limited day-to-day direction. Its next generation of Gemini Agents is designed to let users provide an objective while the agent handles work such as answering questions, creating documents and assets, and organizing meetings.

The agents can run in the background on Google Cloud infrastructure, including when users are away from their desks. Google describes them as both personal assistants and digital colleagues. For workplace use, an agent can be assigned its own email address, calendar, and Drive storage allocation.

Gemini Agents will also work with Anthropic's models, and Google says it plans to add other models in the future. Google Cloud CEO Thomas Kurian described the platform as one that can run anywhere, execute any task, support a user's choice of models, and apply enterprise control.

The agents are also designed to work with third-party platforms including Microsoft 365 and Slack, and on iOS, Android, macOS, and Windows devices. They are not generally available yet; some customers can access them through a private preview. Google has not yet shared full details of their use cases, which are expected as general availability approaches.