Gemini Agent for Work Gives AI Coworkers Their Own Email and Identity

Google introduced the Gemini agent on October 8, 2026 at its Gemini at Work event. The headline is a single agent for work, but the part that matters to anyone building agents is the identity model: an agent can get its own email address, calendar, Drive and directory entry, and every action it takes is logged under its own name instead of yours.

This post covers what Google announced, how its governance pieces fit together, and which ideas are worth copying when you design agents for a small team.

What Google actually announced

According to Thomas Kurian's keynote write-up on the Google Cloud blog, the Gemini agent answers questions, does knowledge work, creates media and writes and runs code from one prompt box. You give it objectives instead of step by step instructions, and it comes back with finished work in the tools you already use.

It runs in the cloud, so memory and context follow you across web, mobile, desktop, the command line, Google Workspace, Microsoft 365 and Slack. For bigger jobs it creates temporary sub-agents, each with its own identity, and coordinates them across parallel and sequential steps that can run for hours or days.

Coworker agents and agent identity

The coworker agent is the most interesting piece. You describe a role, such as an events coordinator, and Gemini creates an agent with its own Workspace account. Colleagues add it to a Chat space or mention it in a document comment, and its edits show up in version history under its own name.

Google says a coworker agent acts under its own identity rather than yours, and it sees only what you share with it. Access follows the sharing and membership your team already uses. Technobezz's coverage adds that these agents can reach only material shared with them through existing team membership and permissions.

The four governance questions

Google frames agent governance as four questions every company should be able to answer.

• Who is the agent, and what identity does it have?

• What is it allowed to do?

• What did it do, and where can I see it?

• What should it never touch?

The answers are identity, role-based permissions, an audit trail and a component called Agent Gateway. Agents run inside a sandbox with their own network boundary, and all traffic in, out and between agents passes through the gateway, which Google describes as an AI network firewall. You write a policy once, for example that agents may not open documents classified as Need to Know, and it applies to every agent.

Cost controls and model choice

Two details are easy to miss. First, the agent is separate from the model under it. Today Gemini orchestrates across Google's own models and Anthropic's Claude models, and Google says other models will follow. Reworked's report notes that it can pick a model per job to balance accuracy and cost.

Second, admins can set a hard spending cap per project. When the cap is reached the agent pauses until someone resumes it, and spend is tracked per project so costs can be charged back to departments. If you have ever watched an agent loop burn through a token budget overnight, you will see why this matters.

What small teams can copy

• Give every agent its own identity and credentials, never a shared human login.

• Start with least privilege and widen access only when a task needs it.

• Log agent actions separately from human actions so audits are readable.

• Put a budget limit on each agent project before it runs unattended.

• Keep a human approval step for anything that sends money, deletes data or emails customers.

We covered the failure side of this in why fully autonomous AI agents still fail and what works instead, and the always-on model from OpenAI in our post on Dots.

What is still unclear

Access is limited at the moment. Reworked reports that a limited number of companies already have it, and that Google's spokesperson told VentureBeat it will be built into Gemini Enterprise at no additional charge where Gemini Enterprise is available. Industry versions for financial services and legal are in preview, with government, healthcare and retail to follow. Pricing, licensing details and how well it works outside Google's own apps will only be clear after wider testing.

Frequently asked questions

What is the Gemini agent?

It is Google's universal agent for work, announced on October 8, 2026. It handles questions, knowledge work, media creation and coding from a single prompt box and runs in the cloud.

What is a coworker agent?

A coworker agent has a persistent role on a team, with its own Workspace account, email address, calendar, storage and directory entry. It sees only the content that people share with it.

Can it work outside Google Workspace?

Yes. Google says it also works with Microsoft 365 and Slack and connects to tools such as Salesforce, ServiceNow, BigQuery and any Model Context Protocol server.

Which models does it use?

Today it orchestrates across Gemini models and Claude models from Anthropic. Google says other models will be added.

Is it available to everyone?

Not yet. A limited number of companies have access, and wider availability depends on Gemini Enterprise.