Google has unveiled a new Gemini agent for the workplace that is designed to handle tasks more independently and work across the tools employees use every day. Announced on Thursday (October 8, 2026) at Google Cloud’s Gemini at Work 2026 event, the new agent can answer questions, conduct knowledge work, create content, write and run code, and coordinate tasks across business systems. Google described it as a 'single, universal agent for work' that employees can give objectives to rather than detailed instructions.
The announcement reflects Google’s broader push to move workplace AI beyond chatbots that simply respond to prompts. The company wants Gemini to take on assignments, work for extended periods, and return completed results while remaining subject to enterprise security and administrative controls.
The new agent is designed to work directly across Gmail, Drive, Docs, Sheets, Slides, Chat, and Calendar. It retains the same business context and permissions across these applications, allowing employees to delegate tasks that span multiple tools.
For example, a user could ask Gemini to research a market, build a financial model in Sheets, and prepare a presentation from the findings. The agent can also help arrange meetings by identifying relevant colleagues, checking calendars, and coordinating schedules.
According to Google, the agent can keep working after an employee closes their laptop, letting assignments that take hours or even days run in the background.
One of the more significant features is the ability to create 'coworker agents' with defined roles. These agents can have their own Workspace account, email address, calendar, and Drive storage, letting employees interact with them like other team members.
For example, a company could create an agent to coordinate events or handle a specific business function. The agent can be added to a Chat space, mentioned in documents, and assigned work, while operating under its own identity rather than a user's account.
Also Read: Google Nano Banana 2.1: 50% Cheaper AI Image Generation
Google said that the system is designed with enterprise permissions and governance in mind. Agents access only the information users or teams provide, while companies can manage identity, authorization, policies, and other controls.
Gemini can also divide complex assignments among specialized sub-agents, allowing different parts of a project to be handled in parallel or sequence. Google says the system can choose between different AI models depending on the task, including models from Google and Anthropic.
The launch marks another step in Google’s effort to make workplace AI less about answering individual questions and more about delegating complete assignments. For businesses, the appeal will ultimately depend on whether these agents can reliably complete useful work while maintaining the security and oversight required in corporate environments.