Google Gemini Agent: Businesses Get a Universal Work Agent First

Work in Gemini now starts in a single prompt box. On October 8, 2026, at Gemini at Work 2026, Google Cloud introduced the Gemini agent: a universal agent that can take objectives, plan the work, use skills and tools, plug into company systems and bring back something finished.

What it does

• Plans multi-step work from an objective, not only from step-by-step instructions (Thomas Kurian, Google Cloud, via TechCrunch)
• Connects to Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres, Snowflake and any Model Context Protocol (MCP) server
• Picks the best model by default; users can override, including Anthropic’s Claude, with open-source and private models planned later
• Gets its own Workspace account — email, context, calendars, approvers — and can be reached by tag, email, share or group chat
• A tasks inbox shows thinking, subagents, skills, code and progress, with an audit trail attributed to the agent

Where you can use it

iOS and Android, Windows and Mac desktops, the command line, Google Workspace, Microsoft 365, ServiceNow and Slack.

Why businesses first

CEO Sundar Pichai said Gemini has more than 1 billion monthly active users and that nearly 90% of Fortune 100 companies already use Gemini Enterprise. Google is solving security, scale and performance with enterprises before a consumer rollout. Early testers include On, Shopify and PayPal; Google also cites customers such as BNP Paribas, Merck and Ulta Beauty. New spend caps, smart routing and multi-model orchestration aim to keep costs under control.

Rocks take

An agent with its own corporate identity is a big step past chat bots. Home Assistant and local-AI users will want to know how far MCP can stretch into self-hosted tools — Google says any MCP server, inside or outside the network, is fair game.