How to Set Up Google Gemini Coworker Agents with Their Own Email Address

Introduction

On October 8, 2026, Google Cloud CEO Thomas Kurian announced the Gemini agent at Gemini at Work 2026, and the detail that changed the conversation was not the model. It was the identity. A coworker agent gets its own Google Workspace account, including an email address, a calendar, Drive storage, and a presence in your company directory.

This tutorial shows how to set one up: how to get preview access, how to define the role, how to draw the permission boundary around its email and Drive access, and how to govern it once it is running. It is written for Workspace administrators and team leads who will own the rollout. Read it alongside our Google Gemini agent tutorial, which covers the agent itself. As of November 8, 2026, this is not a self-service feature you can switch on in the Admin console.

Quick answer

Describe a role in plain language, and Gemini creates a coworker agent with its own Workspace account, email, calendar, and Drive. Your admin approves its permissions, you share only the folders it needs, and spend caps govern it.

Evidence

Google is pitching the Gemini agent as the front door to all of its enterprise AI, and the scale numbers in the launch announcement explain why the rollout starts with large organizations. In the last year, nearly 500 Google Cloud customers each processed more than one trillion tokens, and nearly 90% of the Fortune 100 use Gemini Enterprise, according to Google’s Gemini at Work 2026 announcement.

The customer results Google cites are concrete. Bradesco, one of Brazil’s largest banks, cut document review time from 1 hour to 5 minutes and reduced risk inconsistencies by 60% after deploying Gemini Enterprise. At Bunnings, an internal agent saved staff half a million hours of administrative work.

The design principle behind the whole feature fits in one line from Thomas Kurian: “You give it objectives, not instructions. You delegate an outcome and come back to finished work.” A coworker agent is the version of that idea with a permanent job description. It keeps four kinds of memory, session, semantic, procedural, and episodic, so it onboards the way a new hire would. That permanence is why the setup matters more than the demo: an agent with its own email address can be emailed by anyone, open files, start meetings, and send messages, so it needs an account, permissions, and guardrails before it touches real work.

Prerequisites

• A Google Workspace organization on a Business or Enterprise plan, with Gemini Enterprise available. Coworker agents are a Gemini Enterprise feature, and the preview is aimed at business customers first.
• A Workspace super administrator, or an admin with permission to manage users, groups, and Drive sharing settings. Only an admin can approve the agent’s permissions and set the spend cap.
• A named human owner for the agent. Google’s governance model assumes every agent has a person who is accountable for what it does and who switches it off when needed.
• Access to your Google Cloud account team or sales representative. There is no public self-service signup for the preview as of November 8, 2026, so preview access goes through your Google contact.
• A written role brief: what the agent does, what it never does, and which folders it may touch.

Step-by-step guide

Step 1: Confirm eligibility and request preview access

The Gemini agent is in private preview for a select group of enterprise customers, and Google has said wider availability is coming soon for Workspace customers on select Business and Enterprise plans, without giving a date. If your organization is already a Gemini Enterprise customer, ask your Google Cloud account team to add you to the coworker agent preview.

Do not expect a signup form or a toggle in the Admin console. As of November 8, 2026, Google has published no public enrollment page for coworker agents, and the Gemini Enterprise product page describes the agent without offering a self-service trial. If anyone on your team finds a third-party page promising instant access, treat it as unverified. While you wait, write the role brief and the permission list from the prerequisites, because preview access arrives with an expectation that you will deploy quickly.

Step 2: Name the owner and write the role brief

Every coworker agent needs one human owner, and the owner is not a formality. The owner writes the role brief, approves the first tasks, reviews the audit trail, and is the person who suspends the agent if it misbehaves. Pick someone who understands the work, not just someone with admin rights. For a first agent, choose a narrow, low-risk role with a clear output, such as an events coordinator that drafts launch readiness documents.

Write the brief as four short lines. First, the role in one sentence: “Events coordinator for the marketing team.” Second, the outputs: “Draft launch readiness documents, coordinate meeting times, and post status updates to the team Chat space.” Third, the limits: “Never email customers, never delete files, never access folders outside the shared marketing Drive.” Fourth, the escalation rule: “Ask the owner before spending money or sharing anything outside the company.” This brief is what you hand to Gemini in step 3, and what your security admin checks against in step 4.

Step 3: Describe the role and let Gemini create the agent

Once your domain is in the preview, creating the agent is the simplest step. You describe the role you need, in plain language, and Gemini creates the agent. According to Google’s announcement, the agent then receives its own Workspace account, including an email address, calendar, Drive, and presence in your company directory. The address sits on a separate agents subdomain, in the form name@agents.company.com, so it is visibly distinct from human mailboxes.

Check the new account the way you would check a new hire’s account. Confirm the directory entry shows the agent’s name and role, the calendar is shared with the right group, and the Drive storage is empty and private by default. The agent arrives with no access to your files beyond what you share with it, so an empty Drive is the correct starting state, not a problem.

Step 4: Approve role-based permissions and share only what it needs

This is the step that makes the difference between a useful coworker and a security incident. Google’s model gives each agent fine-grained, role-based access permissions, approved by your organization’s security administrators, and the agent sees only what you share with it. Access follows the sharing and group membership your team already uses.

Work through the role brief line by line with your security admin. For each folder, mailbox, or system the agent needs, decide whether it gets read access, write access, or nothing at all. Share the specific Drive folders from the brief with the agent’s @agents.company.com address, and nothing else. Add it to the specific Chat spaces where it will work, and do not add it to spaces where it has no job. If the agent connects to an external system, its identity is mapped through industry standards such as OAuth, so it signs in as itself rather than borrowing a human’s credentials. When you are done, write the list down: every folder, space, and system the agent can reach, and who approved each one.

Step 5: Wire the agent into the team’s daily flow

Colleagues work with a coworker agent the way they work with anyone else, by adding it to a Chat space or mentioning it in a document. That is the whole point of giving it an email address and a directory entry: it shows up in autocomplete, it can be emailed a task, and its replies land in the thread under its own name. Start with one channel. Add the agent to a single team Chat space and give it one recurring job, such as posting a weekly status summary.

Show the team how to use it, and what not to do. To assign work, they email the agent’s address or mention it in a Chat space with a clear outcome, a deadline, and the documents it needs, the same way they would brief a colleague. To check its work, they look at the document or the thread, where the agent’s edits appear under its own name in version history. Tell them explicitly not to forward it confidential threads, not to ask it to email customers, and not to share a folder with it just because it is faster than asking the owner.

Step 6: Set the guardrails: spend cap, gateway policy, and audit review

Google gives admins three controls, and you should set all three before the agent’s first real task. First, set a hard spending limit per project in the Cloud Billing Console. Gemini enforces the cap itself, and when the cap is hit, that project’s agent pauses until someone resumes it with a single click. Because tracking is per project, you can charge AI costs back to the department that owns the agent. Pick a cap low enough that a runaway task stays an annoyance.

Second, write one policy in Agent Gateway, the AI network firewall that all agent traffic passes through. Google’s own example is a policy that agents may not open documents classified as Need to Know, and you write the policy once and it applies to every agent in your company.

Third, agree on an audit review rhythm with the owner. Every action the agent takes is written to an audit trail and attributed to the agent rather than to a person, and its identity is stamped into any virtual machine it spins up to run code. The owner should review that trail weekly at first, looking for actions outside the brief, shares the agent received that were not approved, and tasks that cost far more than expected.

Troubleshooting

The agent cannot see a document the team shared with it. This is the most common first-week problem, and it almost always means the document was shared with the wrong identity. Check that the folder or file was shared with the agent’s @agents.company.com address, not with the human owner’s address and not with a group the agent is not a member of. If the file lives in a shared drive, check the shared drive’s membership separately.

The agent’s edits show up under a human’s name, or a human’s edits show up under the agent. Check who was actually mentioned. If a colleague tagged the agent in a document comment, the agent’s reply and suggested edit appear under its own name in version history. If a colleague pasted the agent’s draft into the document themselves, the edit is theirs, not the agent’s. Train the team to let the agent make its own edits rather than copying its output into a document by hand.

The agent pauses in the middle of a long task. First check the spend cap. When a project’s AI spend hits the hard limit set in the Cloud Billing Console, that project’s agent pauses, and the owner can resume it with a single click. If the cap is not the cause, check whether the task hit a policy block in Agent Gateway, such as a document classified Need to Know that the agent tried to open. A policy block is not a bug. It means the guardrail worked.

The agent is not available in your Workspace at all. This is expected for most organizations right now. Coworker agents are in private preview for a select group of enterprise customers, and Google has announced no general availability date as of November 8, 2026. If your admin console has no agent settings and your Google contact has not confirmed preview access, you cannot set one up yet, and no workaround will change that. Use the waiting time to write the role brief and the permission list, so you are ready when access arrives. For the broader agent setup, our Google Gemini agent tutorial explains what the personal agent can do today, and our Gemini skills team registry guide shows how to share reusable skills across a team.

FAQ

Can I set up a Gemini coworker agent today? Only if your organization is in the private preview. Google announced the Gemini agent on October 8, 2026, and coworker agents are available to a select group of enterprise customers. Wider availability is promised soon for select Workspace Business and Enterprise plans, with no general availability date as of November 8, 2026.

What email address does a coworker agent get? Its own address on a separate agents subdomain, in the form name@agents.company.com, along with its own calendar, Drive storage, and an entry in your company directory. It acts under that identity, not under the identity of the person who created it.

Does a coworker agent cost a Workspace seat? Google has not said whether a coworker agent consumes a paid Workspace seat. Google told the analyst firm Futurum that it does not intend to charge a per-seat license for coworker agents in Workspace.

Can a coworker agent read all of my team’s files? No. A coworker agent sees only what you or your team members share with it, and access follows the sharing and group membership your team already uses. Share a Drive folder with the agent’s Workspace account and it can work with those files. Do not share them, and the agent cannot see them.

How do I stop a coworker agent or take away its access? Treat it like an employee account. Its Workspace account can be suspended or removed by an admin, its role-based permissions can be revoked, and its spend can be cut off with a hard cap in the Cloud Billing Console.

How is a coworker agent different from a personal Gemini agent? A personal agent works for one person, under that person’s identity. A coworker agent has a persistent, defined role on a team, its own identity and email, and works across days and sessions for the whole group.

Conclusion

A Gemini coworker agent is less like installing software and more like hiring someone. The email address, the calendar, and the Drive storage are the easy part. The real setup is the role brief, the permission boundary, and the guardrails, and those are decisions only your team can make. Get them right before the agent’s first task, and you get an agent that works under its own name, inside limits you set, with a record you can audit.

If your organization is not in the preview yet, do the paperwork now. Write the role brief, list the folders the agent may touch, and pick the owner. When access arrives, the setup takes an afternoon. For a broader automation setup, see the complete n8n guide.

Frequently asked questions