Google launched an autonomous agent for Gemini on Thursday at its Cloud event, designed to execute multi-step tasks across enterprise systems without step-by-step instructions. The agent can generate code, schedule meetings, book travel and manage approvals by connecting directly to a company's internal infrastructure.
Google CEO Sundar Pichai said Gemini serves over 1 billion monthly active users, with nearly 90 percent of Fortune 100 businesses using Gemini Enterprise. The company is rolling out the agent to enterprises first, deferring consumer availability to address security and performance challenges at scale.
Thomas Kurian, CEO of Google Cloud, described the agent as operating on objectives rather than discrete commands. It plans workflows, loads specialized skills, and integrates with business systems to achieve goals. Users can monitor execution through a tasks inbox that shows the agent's reasoning, subagent delegation, and tool loading.
The agent functions as a virtual coworker with its own Google Workspace account and email address. This identity gives it access to company org charts, team assignments, time zones, approval workflows and calendar data—critical context for executing real tasks inside a business.
The agent automatically selects the optimal model for each task; users can override the choice. Initial third-party support includes Anthropic's Claude. Google plans to expand model options to include open source and private alternatives.
Integrations include Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres and Snowflake. The agent also supports the Model Context Protocol, allowing secure connection to any MCP server inside or outside the corporate network.
Google's enterprise-first approach plays to its strength: an entrenched cloud customer base with existing compliance infrastructure. The competitive moat lies in breadth of system integration—the company with the most connectors and the deepest read-access to internal systems wins lock-in. The business model rests on automation of routine work and cost reduction for large organizations, with pricing likely tied to task volume or execution hours.
