Google adds the Gemini Agent to Gemini Enterprise
The agent takes a goal, makes a plan, and picks the AI model, which can be Claude from Anthropic.
On 8 October 2026, Google Cloud CEO Thomas Kurian showed the Gemini Agent at Gemini at Work 2026. It is one agent for knowledge work, image work, and code. It is part of Gemini Enterprise. Customers with Workspace Business or Enterprise subscriptions do not pay more. It starts in a private preview, and Google has not given a date for wide release.
What it does
The user gives the agent a goal. The agent makes a plan, uses skills and tools, and puts the result in documents, inboxes, or development environments.
The agent works in the cloud and continues to work after the user closes the laptop. It can make temporary sub-agents for hard tasks.
Model choice
The agent picks the AI model for each task. It can use Google models and Claude models from Anthropic. Google states that more third-party models will follow.
Control
Google gives the company these tools to control its agents:
- Each agent has a verified identity and role-based access rights.
- Each step goes into an audit trail that belongs to the agent.
- An Agent Gateway sends all data traffic through one point where company rules apply.
Spending limits pause a project agent when the limit is reached.
The numbers
Google Cloud states that SQL queries at Bloomberg Media are 63 percent more accurate. Commerzbank states that it cut manual inspection time from 20 hours to one hour.
Sources
Posted
