Google has launched a new AI agent for businesses that can plan and carry out tasks across workplace applications, pushing Gemini beyond its role as a chat assistant. Announced at a Google Cloud event on Thursday, the agent will first be offered to enterprise customers before reaching consumers. Sundar Pichai said Gemini has over a billion monthly active users, and nearly 90% of Fortune 100 companies already use Gemini Enterprise.
Thomas Kurian, CEO of Google Cloud, said the agent works toward defined goals rather than following one instruction at a time. It can plan tasks, use custom skills and connect to internal systems, including Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres and Snowflake, as well as MCP servers. Gemini picks a suitable model for each task by default, though users can choose their own. Anthropic’s Claude models are among the first third-party options, with open-source and other private models to follow.
To keep work transparent, Google is adding a task inbox where employees can track the agent’s progress, code, skills used and delegation to subagents. The agent gets its own Workspace account and email address, plus context on team structures, time zones, calendars and approval rules. Staff can tag it, email it, share files or add it to group chats, and its actions are logged in an audit trail under its own identity.
Early testers included On, Shopify and PayPal, while Gemini Enterprise customers include BNP Paribas, Merck, Ulta Beauty and Wesfarmers. Google will also offer spending controls such as multi-model orchestration, smart routing and real-time limits to help companies manage AI costs. Pichai said starting with businesses will help address security, scale and performance challenges.
