Google Cloud launched a unified Gemini agent at its Gemini at Work event on October 8. The agent receives its own Workspace account, including an email address, calendar, Drive storage and a listing in the company directory. Thomas Kurian, CEO of Google Cloud, described the system as able to take "objectives, not just instructions."
The agent plans multi-step work, delegates to sub-agents, and keeps running after the user closes the laptop. By default it selects the model best suited to each task and can route jobs to Anthropic's Claude as well as Gemini models. Administrators control permissions. Every action writes an audit trail under the agent's own identity rather than an employee's.
Sundar Pichai said Gemini has more than 1 billion monthly active users and that nearly 90% of Fortune 100 companies now use Gemini Enterprise. Google is rolling the coworker agent to businesses first so it can address security, scale and performance issues before a consumer release. Early testers included On, Shopify and PayPal. The company has not published pricing or a general-availability date.
Key takeaways
- Gemini agent gets its own @agents.company.com email, calendar and directory seat.
- Takes objectives, plans work, and can route tasks to Claude models.
- Businesses first; nearly 90% of Fortune 100 already use Gemini Enterprise.
What Google's Gemini coworker agent actually does inside Workspace
The agent operates as a persistent coworker. Colleagues can @mention it in Chat, email it, share a file with it, or add it to a group. It sees only what has been shared with it. Edits it makes appear under its own name in version history. It connects to Google Workspace, Microsoft 365, Slack, Jira, Git, BigQuery and other systems, and can talk to any Model Context Protocol server.
Google also announced flexible spend controls, including multi-model orchestration and real-time caps. The launch sits alongside other agent products from OpenAI and Anthropic that aim to move from chat responses to task ownership. For the content-creation side of these tools, see How AI is Reshaping the Content Creator Stack in 2026.
What to watch next
Pricing and a public launch date will determine how quickly the agent moves beyond private preview. Whether enterprises treat the agent's separate identity as a governance advantage or an audit complication will show up in early deployments. The ability to mix Claude and Gemini models inside one workflow is the feature most directly comparable to multi-model setups already in use at other labs.