TL;DR
Google Cloud announced the Gemini agent today at its Gemini at Work 2026 event: a single universal agent that answers questions, does knowledge work, creates images and media, and writes and runs code — all from one prompt box, inside the apps you already use. The headline move: AI “coworker agents” that get their own email address, calendar, and Drive, and take assignments like a member of the team.
One agent, everything work
Today, at Gemini at Work 2026, Google Cloud announced the Gemini agent: one agent for questions, knowledge work, content creation, and coding. Google CEO Thomas Kurian, in a keynote post adapted for the Cloud blog, put the thesis in a line: “Work now starts in the prompt window.”
The agent carries your business context — your tools, data, and work history — and works from a single prompt box. You delegate an outcome, and it plans the work, uses tools and skills, connects to your company’s systems, and brings back something finished inside the documents, the inbox, and the developer environments you already work in.
It works inline inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar — and outside Google’s walls too: Microsoft 365, Slack, web, iOS, Android, Windows, Mac, and the command line. A task can run for hours or days in the cloud, so you can close your laptop and come back to finished work.
The coworker with an email address
The sharpest idea in the announcement is the coworker agent: a persistent digital team member with its own identity — its own @agents.company.com email address, its own calendar, its own Drive storage, and its own presence in the company directory. It only has access to the context you or your team give it.
Working with it looks like working with a colleague. A marketing manager asks an events-coordinator agent in a chat group to draft a launch readiness document, and it posts the finished file back to the group. In a Doc, it suggests edits under its own name in version history.
Gemini also orchestrates teams of temporary sub-agents for complex work — each with its own identity — running parallel and sequential steps over hours or days, and it can schedule tasks or respond to events. Google says four kinds of memory keep it coherent: session memory for the task at hand, semantic memory built from documents and people it has worked with, procedural memory for how jobs get done, and episodic memory of everything it has done before.
The model underneath is a choice, not a marriage
Notable in a race where everyone sells their own model: the agent orchestrates across Google’s Gemini models and Anthropic’s Claude models today, with other private and open models coming later. The point is matching the job to the right model — a big one for hard tasks, a cheap one for easy ones — with Smart Routing and real-time spend caps so costs stay controlled.
Google is pitching the enterprise controls hard: identity and policy management, authorization and permission controls, secure sandboxing, network gateways. Industry-specific versions for financial services and legal teams are in preview, with government, healthcare, and retail coming later, according to Reuters.
The context for all this: OpenAI launched always-on agents called dots in September, and Meta released its personal agent Muse last month. Everyone is now racing to turn the chatbot into a long-running digital employee.
Past the pilot stage, by Google’s numbers
Google is arguing organizations have moved past experimentation: in the last year, nearly 500 Google Cloud customers each processed more than one trillion tokens. The company says nearly 80% of all Google Cloud customers use its AI products and nearly 90% of the Fortune 100 use Gemini Enterprise.
The customer examples have real numbers attached. Bradesco cut document review time from one hour to five minutes and reduced risk inconsistencies by 60%. Bunnings saved staff half a million hours of administrative work. Orange Spain deployed more than 1,000 custom agents across HR, IT, sales, and customer service. Japanese insurer SOMPO built over 10,000 custom agents across 34,000 employees.
The caveats are the usual ones for a launch-day announcement: these are Google’s chosen customers and Google’s chosen numbers. The coworker-agent idea is genuinely new territory, though — identity, access control, and audit for AI teammates are the problems to watch next.