Google Cloud wants Gemini to stop being something employees chat with and start being something they hand work to. At its Gemini at Work 2026 event, Google Cloud CEO Thomas Kurian announced the Gemini agent. It is a single assistant that combines answering questions, creating documents and media, carrying out tasks and writing code, all from one prompt box and one API.
The pitch is delegation, not conversation. A user sets an objective, and the agent plans the job, picks its tools and skills, and returns finished work. Because the agent runs in the cloud, it keeps one memory across web, mobile, desktop, the command line, Workspace, Microsoft 365 and Slack. Long jobs keep running after the laptop closes. For bigger tasks it can create temporary sub-agents, each with its own identity, to work in parallel or in sequence.
The more striking idea is the “coworker agent.” Describe a role, and Gemini creates an agent with its own Workspace account, email address, calendar, Drive and directory listing. Colleagues can add it to a Chat space or tag it in a document comment, and its edits appear under its own name in version history. Google says it sees only what the team shares with it.
The most telling detail is that Gemini the agent is now separate from Gemini the model. Google says the agent sends each job to whichever model fits best. Today that means its own Gemini family plus Anthropic’s Claude models, with other private and open models to come. That amounts to an admission that no single model wins every job. It is also a sensible hedge for customers who worry about lock-in while the model leaderboard keeps changing.
Inside Workspace, the agent works directly in Gmail, Docs, Sheets, Slides, Drive, Chat and Calendar. When a manager emails asking for a status deck, Google says the agent can spot it as a task to hand off with one click. New data skills let business users build BigQuery reports in plain language. Those queries can be saved, so running a report again doesn’t use more tokens.
Governance gets heavy emphasis, and for good reason. Every agent gets a cryptographically verified identity, role-based permissions and an audit trail that names the agent rather than a person. Each agent also runs in a sandbox, and all of its traffic passes through Agent Gateway, a firewall that applies company policies to every agent at once.
Google backs the push with adoption figures. It says nearly 80% of its cloud customers use its AI products and nearly 90% of the Fortune 100 use Gemini Enterprise. Customers in the region include Ooredoo Qatar and Qatar University, where staff have built more than 2,000 custom agents.
The announcement doesn’t cover pricing or rollout timing. A bigger open question is how agents with their own inboxes will behave once the demos end and real, messy office work begins.
