Google Cloud launches the Gemini agent for work: one agent for questions, documents and code that runs for days, spawns sub-agents and routes jobs between Gemini and Claude

Why it matters

【Key points】

At Gemini at Work 2026 on Oct 8, Google Cloud CEO Thomas Kurian launched the Gemini agent, a single universal agent for work: one prompt box and one API for answering questions, knowledge work, images and media, and writing and running code. Users assign objectives, not instructions, and come back to finished work.

It is reachable from web, mobile, Windows and Mac, the command line, Google Workspace, Microsoft 365 and Slack, can run headless inside third-party apps, and works inline in Gmail, Drive, Docs, Sheets and Calendar. It runs in the cloud with one set of memories across devices, so jobs lasting hours or days continue after the laptop closes. It can spin up sub-agents with their own identities or act as a persistent coworker agent with its own @agents company email and storage. The agent is separate from the model: today it routes each job across Gemini models and Anthropic's Claude, with other models later, plus smart routing and real-time spend caps set in Cloud Billing that pause the agent when hit. Financial services and legal versions are in preview. Google says nearly 500 customers each processed over a trillion tokens in the past year and nearly 90% of the Fortune 100 use Gemini Enterprise.

【Verifiable】

Google Cloud blog: https://cloud.google.com/blog/products/ai-machine-learning/welcome-to-gemini-at-work-2026/

【Our read】

1. Google is selling the agent apart from the model and putting rival Claude inside its flagship product. Enterprise agent competition is now about who owns context, permissions and billing, not who has the best model; Argon gets one line in the infrastructure section.

2. Once agents have identities, mailboxes and their own sub-agents, governance shifts from what a model said to whose name an agent acted in, the question U.S. liability drafts and the UK ICO are chasing this week.

3. Real-time spend caps are the underrated piece. For agents that run for days, runaway cost is a risk alongside overreach.