Google Cloud on Thursday announced the Gemini agent, a single, universal AI agent for work that answers questions, handles knowledge work, creates media, and writes and runs code — all from one prompt box and all across the apps where office work already lives. The announcement came from Google Cloud CEO Thomas Kurian's keynote at the company's Gemini at Work 2026 event.
The pitch is a fundamental change in how workers interact with AI. In a blog post adapted from the keynote, Google framed the shift bluntly: with the Gemini agent, users give objectives rather than instructions. You delegate an outcome, and the agent plans the work, uses skills and tools, connects to your systems, and brings back something finished. For more context on this story, see our ongoing AI trends.
One Agent, Every Surface
The Gemini agent works directly inside Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, carrying the same memory, skills and controls it has everywhere else. It is accessible from any device — web, iOS and Android mobile devices, Windows and Mac desktops — and through any channel, including the command line, Google Workspace, Microsoft 365 and Slack. It can also be integrated into third-party applications or operate as a headless agent that requires no dedicated interface at all.
Because the agent runs in the cloud, Google says it maintains a single set of memories, context and one personalization graph no matter which device or channel a worker uses. You never have to re-brief it or remember which machine you assigned a task on. Critically for long-running work, tasks that take hours or days keep running after a user closes their laptop, and the results are still there when they return.
The Verge reported that the agent will be available within the Gemini Enterprise app, allowing users to chat with it and assign it tasks from a single interface, and that it can work with job-specific sub-agents to complete tasks.
A Coworker With Its Own Email Address
The most eye-catching capability is organizational rather than technical. The Gemini agent can dynamically create a roster of sub-agents — temporary, job-specific agents, each with its own identity — to tackle multi-step tasks, coordinating workflows that include parallel and sequential steps running for hours or days.
Beyond those temporary helpers, Google says the agent can act as a coworker: a persistent team member with a defined role and operational presence, such as a project manager within a team or an analyst in a finance department. Coworker agents get their own email addresses on a company's domain — The Verge cites the format @agents.company.com — along with their own persistent storage, and they can only access the context that a user or team explicitly provides.
That design is clearly aimed at the governance concerns that have slowed enterprise agent adoption. By giving AI coworkers bounded, auditable identities rather than sharing a human's credentials, Google is attempting to make delegating work to an agent feel more like onboarding an employee and less like handing over the keys to the company intranet.
The Model Is a Separate Choice — Including Claude
In a notable admission of a multi-model future, Google says the Gemini agent is not tied to Google's own models. The agent runs each job on the model that fits best, orchestrating across the Gemini family of models and Claude, Anthropic's model lineup, according to the blog post.
That architecture turns the agent layer, rather than any single model, into Google's source of differentiation. As frontier models from multiple labs converge in capability, the durable moat becomes the system that knows your data, your tools and your work history — and that can route each task to whichever model performs it best.
The Scale Numbers Behind the Push
Kurian's keynote paired the announcement with usage figures that Google says show enterprises moving past experimentation. Nearly 500 Google Cloud customers each processed more than one trillion tokens in the last year, according to the company. Google also says nearly 80 percent of all Google Cloud customers now use its AI products, and nearly 90 percent of the Fortune 100 use Gemini Enterprise.
Those are Google's own figures and have not been independently audited, but they sketch the competitive context: the enterprise AI market has moved from pilots to production, and the vendors are now fighting over which platform becomes the default operating layer for AI-assisted work.
The Agent Platform Race Is On
The Gemini agent arrived one day after Anthropic launched Claude Dashboards and Motion, new tools that turn Claude into a live data analysis and animation platform for business users. The back-to-back releases show the frontier labs converging on the same conclusion: the chatbot era is giving way to the agent era, and the winner will be whoever embeds agents deepest into enterprise workflows.
For now, Google's advantage is distribution. The Gemini agent meets workers inside the Gmail inboxes, calendars and documents they already use, and extends into rivals' territory through Slack and Microsoft 365 integrations. Anthropic's advantage is the trust it has built in coding and analysis. Microsoft, with Copilot embedded across its own office empire, remains the incumbent neither wants to wake.
What happens next will be decided less by demo videos than by procurement decisions — and by whether enterprises conclude that an AI coworker with its own email address is a colleague they are ready to hire.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
