Google launches ‘Gemini at Work’ with a work agent
Google Cloud Blog ● Covered by 5 sources
Google launched Gemini at Work, a single agent meant to handle work across Workspace. It can keep going across apps and devices, with model choice and controls built in.
Based on reporting by Google Cloud Blog — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Google Cloud is pushing Gemini from chatbot to co-worker. At Gemini at Work 2026, the company unveiled a “single, universal agent for work” that can answer questions, do knowledge work, create images and media, and write and run code from one prompt box. The pitch is simple: give it an objective, not a step-by-step script, and let it bring back something finished.
The agent is meant to live everywhere a worker already is. Google says it works inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, and also across web, iOS, Android, Windows, Mac, the command line, Slack, Microsoft 365, and third-party apps. Because it runs in the cloud, it keeps memory and context across devices, so a task can keep going after the laptop is shut.
Google is also leaning hard into the “agent” part of the pitch. Gemini can spin up temporary sub-agents for multi-step jobs, or create a more persistent coworker agent with its own role, storage, and company email. It can connect to tools and data from places like Confluence, Microsoft Office, Teams, Git, Jira, Salesforce, ServiceNow, BigQuery, Databricks, Postgres, Snowflake, and MCP servers, then use skills and context to decide how to finish the work.
The company is pairing that with cost and governance controls. Gemini chooses from its own models and Claude models from Anthropic today, with other private and open models coming later, and Google says the point is to match the model to the job instead of defaulting to the biggest one. It also says the system is built with identity and policy management, authorization, permission controls, secure sandboxing, network gateways, and real-time spend caps.
Google backed the launch with customer examples, and there are a lot of them. BNP Paribas is using Gemini Enterprise for more than 65,000 employees. Bradesco says it cut document review time from 1 hour to 5 minutes. Orange Spain has deployed more than 1,000 custom agents. And Wesfarmers says an internal agent at Bunnings saved staff half a million hours of administrative work.
My take — AI-written commentary, not fact-checked reporting
This is the right bet: make the agent boringly useful, not mystically clever. The real story here isn’t a magic model, it’s Google trying to turn AI into enterprise plumbing, with permissions and spend controls instead of fireworks. That’s the grown-up version, which is exactly why everyone else keeps dressing up a workflow tool as a revolution.
Read more about this at: Google Cloud Blog