Google has launched Gemini Agent — a new agentic layer for its Gemini AI platform, exclusively for business customers, starting Thursday. It’s not a chatbot. It’s not a copilot. It’s an AI that accepts objectives (“Plan Q4 marketing campaign across APAC and EMEA”), then breaks them down, delegates subtasks, pulls data from internal systems, writes code, books meetings, drafts emails, and logs every action under its own Workspace identity.
Complete with audit trail, email address, and calendar awareness. No consumer rollout is scheduled yet. This is enterprise-first, full stop. The agent arrives as part of Gemini Enterprise, which Google says is already used by nearly 90% of Fortune 100 companies.
It will be available on iOS, Android, Windows, macOS, CLI, Slack, ServiceNow, Microsoft 365, and Google Workspace. Early testers include On, Shopify, and PayPal. Customers already using Gemini Enterprise, BNP Paribas, Merck, Orange Spain, SOMPO, Ulta Beauty, Wesfarmers. Are the immediate audience.
The agent connects natively to BigQuery, Snowflake, Postgres, Databricks, Jira, Confluence, Git, Slack, and Microsoft 365. It also supports Model Context Protocol (MCP) servers, inside or outside corporate firewalls. Users can attach files, folders, or preconfigured workstream bundles, say, a folder of brand guidelines plus a Jira board plus a Confluence space.
And assign the whole package as context. By default, Gemini picks the best model for each step. But admins can override: Anthropic’s Claude models are supported at launch. Open-source and private models will follow.
A “tasks inbox” shows live progress, not just outputs, but reasoning chains, subagent delegation, skill loading, and code generation. Cost controls include real-time spend caps, smart routing between models, and multi-model orchestration. All aimed at predictable enterprise billing. No pricing details were disclosed.
No consumer release date was announced. No Indian or US retail price was given, because this isn’t sold on Amazon or Flipkart. It’s licensed through Google Cloud, bundled with Gemini Enterprise contracts. That means rollout depends on enterprise procurement cycles, not store shelves.
Next up: Google says it will expand model choice beyond Claude, add more MCP integrations, and. Eventually, bring agents to consumers. But “eventually” is all it said.
ADFiled by The AI Desk
Models, assistants and the companies and chips behind them, reported from what was released and what was claimed, with the difference kept clear.
More from this desk →
Be the first to comment
Join the argument. No password, just your email or a passkey.