The Gemini at Work 2026 logo, from Google Cloud’s announcement of the Gemini agent on October 8, 2026. Image: Google Cloud

Google Cloud has unveiled the Gemini agent, which it calls “a single, universal agent for work”: one agent that answers questions, does knowledge work, makes images and media, and writes and runs code, and that a team can turn into a coworker with its own email address, calendar and Drive.

Google Cloud chief executive Thomas Kurian announced it in his keynote at Gemini at Work 2026 on October 8. The pitch is that the agent runs in Google’s cloud, keeps one memory across every device, and carries on with jobs that take “hours or days” after you close your laptop.

“You give it objectives, not instructions”

Kurian framed the agent as something you hand a goal to rather than steer step by step:

You give it objectives, not instructions. You delegate an outcome and come back to finished work.

Thomas Kurian, CEO, Google Cloud

It can be reached on the web, on iOS and Android, on Windows and Mac, from the command line and inside Google Workspace, Microsoft 365 and Slack. Developers can also build it into their own apps through a single API as a “headless” agent with no interface of its own. For bigger jobs it spins up temporary sub-agents, each with its own identity, and runs their steps in parallel or in sequence.

Google says it keeps four kinds of memory (session, semantic, procedural and episodic), plugs into tools such as Confluence, Microsoft Office, Teams, Jira, Salesforce, ServiceNow, BigQuery, Databricks and Snowflake, and can connect to any Model Context Protocol (MCP) server.

An agent with its own inbox

In Gmail, Docs, Sheets, Slides, Chat, Drive and Calendar the agent works inline, and Google gives it three jobs. As a personal assistant it can, for example, set up a meeting with “the usual team of regional event leads” without being given a single name, by working out who they are from a chat space and an old thread, checking calendars and emailing to find a time, external guests included. Workspace will also suggest tasks to hand over with one click.

The third job is the biggest change. Describe a role, and Gemini creates a coworker agent with its own Workspace account: an email address, a calendar, Drive storage and a listing in the company directory. Colleagues add it to Chat spaces or @mention it in a document, where it suggests edits and shows up under its own name in the version history. Google says a coworker agent “acts under its own identity rather than yours” and sees only what people share with it.

Gemini and Claude under the hood

The agent is not tied to one model. Google says it picks the best fit for each job, “orchestrating across our Gemini family of models and Claude models from Anthropic today”, with other private and open models to follow. A Smart Routing tool sends simpler work to cheaper models. That puts Claude inside Google’s flagship work agent, alongside the Gemini 4 Argon model Kurian listed for frontier reasoning.

On control, every agent gets its own cryptographically attested identity, and every action goes into an audit trail credited to the agent rather than a person. Agents run inside an Agent Sandbox, and all their traffic passes through Agent Gateway, which Google calls “an AI network firewall”, so a rule such as “agents may not open documents classified Need to Know” applies to every agent at once. Companies can also set hard spending caps per project in the Cloud Billing Console: when one is hit, that project’s agent pauses until someone resumes it.

Versions tuned for financial services and legal work are in preview, with government, healthcare and retail “coming soon”.

When you can use it

Google’s announcement gives no price, release date or list of countries. According to 9to5Google, the agent is in private preview, with wider availability “soon” for Workspace customers on select Business and Enterprise plans. Google says nearly 90% of the Fortune 100 already use Gemini Enterprise, the business product the agent sits in.

It lands two weeks after Microsoft pitched its new Copilot as an “OS for work” with an Autopilot agent that keeps going while you are away, and a day after Microsoft said it is building a sandbox for AI agents into Windows.

Why it matters

Giving an agent its own email address, calendar and place in the directory treats it as a member of staff, not a feature, and that is the real bet here. It also shows how much of the fight is about distribution: Google is putting Anthropic’s models inside its own product to keep businesses in Workspace. Without prices or dates, though, nobody outside the preview can yet judge whether it works.

Sources: Google Cloud blog (Thomas Kurian), Google, 9to5Google.

Latest Google news

More Google news