Overview
- Google introduced the Gemini agent on Oct. 8 as a cloud-hosted, “universal” workplace AI that can accept objectives, plan multi-step jobs, create sub-agents, and return finished documents or code.
- Businesses can give agents persistent coworker identities with their own email, calendar, Drive storage, and directory listing so agents can act on behalf of individuals, teams, or roles.
- Gemini routes tasks to the most suitable model today by using Google’s Gemini models and Anthropic’s Claude, offers Smart Routing and real-time spend caps to control cost, and lets customers pick models when needed.
- Google built in governance tools including cryptographically attested agent identities, an Agent Sandbox, an Agent Gateway network firewall, and audit trails that log actions under the agent’s identity rather than a person’s name.
- The agent is in private preview with financial and legal industry previews available, no pricing or general-availability date disclosed, and customers are weighing integration, permission and cost trade-offs as Google leverages Workspace adoption against rival agent offerings.