Google announced a universal Gemini agent for enterprise work at its Gemini at Work event on Thursday.
The agent embeds a persistent, multi-model assistant directly into enterprise workflows, competing with standalone assistants by letting users assign objectives rather than step-by-step instructions. Google introduced it in a Cloud blog post adapted from Thomas Kurian's keynote, and The Verge reported it is available only to enterprise customers in private preview.
The release follows a consumer-focused Gemini Spark agent that Google launched in May to work across Workspace and the web through a separate interface. The enterprise agent consolidates those capabilities inside a single Gemini Enterprise app, where users can assign work, ask questions and generate code without switching contexts.
The Gemini agent runs in the cloud and maintains shared memory and context across devices, so tasks continue after a laptop is closed, the blog said. It can spin up temporary sub-agents for specific jobs or operate as a persistent “coworker agent” with a dedicated identity, including an @agents.company.com email address and isolated storage. Google said it orchestrates models based on the job, using its own Gemini family and Anthropic's Claude models today, with plans to add other private and open models.
The company cited early testers such as sportswear brand On for dynamic model selection, Shopify for blending frontier models across merchants, and PayPal, which routes 10 million multi-model requests every week. It provided no independent benchmarks.
The agent connects to collaboration tools such as Microsoft 365, Slack and Confluence, development platforms like Git and Jira, databases including BigQuery and Snowflake, and any MCP-compatible server, the blog stated. Teams can publish reusable skills and tools to company-wide registries, and the system enforces security through identity management, permission controls, secure sandboxing and network gateways.
Google also announced industry-specific specializations for financial services and legal teams, plus cost controls that include real-time spend caps and “Smart Routing” to send simpler tasks to smaller models. It said nearly 80% of its cloud customers already use AI products and nearly 90% of the Fortune 100 use Gemini Enterprise, but it did not disclose pricing for the new agent.
The agent is offered only in private preview with no public launch date or pricing. The blog post acknowledged that the capabilities described come from customer use cases that participants volunteered.