Google Cloud announces Gemini agent, a single universal AI agent for work

Google Cloud announced Gemini agent at its Gemini at Work 2026 event on October 8, 2026. According to Google's blog post, the agent draws on an organization's business context to take on tasks ranging from knowledge work and question answering to content creation and coding, all from a single prompt box. It plans the work, calls on skills and tools, links to the company's business systems and returns finished output in the documents, inbox or developer environments employees already work in.

• According to 9to5Google, it can be called by typing @Gemini in Gmail, Docs, Sheets, Slides and Chat, and it can also be reached from the command line, Slack and Microsoft 365.
• For multistep work it coordinates sub-agents in parallel and in sequence, and these workflows can run for hours or days.
• It picks a suitable model for each job; the Gemini and Claude model families are supported now, with other models planned. Google says cost controls are built in.
• The agent is in private preview for now and will be offered more widely soon to Workspace customers on select Business and Enterprise plans. Neither source gives a price.

Why it matters

The announcement can be read as Google trying to bring AI use at work together around a single agent and a single entry point. For Turkish companies that use Google Workspace, this could mean the agent acting for users or teams on email, documents and spreadsheets. For that reason, it would be sensible for IT and legal teams to assess which data the agent can reach, which actions require approval and what it keeps in memory before the product becomes widely available. According to 9to5Google, the agent has four kinds of memory: session, knowledge base, how jobs get done, and past activity.

You can find more on AI setups that use company data in a controlled way on our enterprise AI page.