Menu
Image

NEWSLETTER

One email a week on what’s actually working in AI: tools, agents, infrastructure and funding. Plus first-hand lessons from my own founder journey, so you can build and scale AI products that actually work.

Follow Us

Digital AI coworker walking through an office, with the Gemini and Google logos

Google’s Newest Employee Has an Email Address and No Desk

Mélony Qin Published on November 5, 2026 0

Did you know Google’s next AI coworker could have its own email address?

That’s right! On October 8, at its Gemini at Work 2026 event, Google Cloud CEO Thomas Kurian introduced the Gemini agent. Google calls it “your new single, universal agent for work.” It lives inside Gemini Enterprise. You give it a goal. Then it plans the steps, connects to company systems and comes back with finished work.

Today, Gemini became an agent.

Thomas Kurian, CEO, Google Cloud, via The Register
Google Cloud Gemini at Work 2026 artwork on colorful bubbles
Image: Google (source)

You see, assistants wait. Agents don’t.

Most AI assistants still wait for your next prompt. Google, however, wants something closer to a colleague. In Kurian’s words: “You give it objectives, not instructions. You delegate an outcome and come back to finished work.”

Because the Gemini agent runs in the cloud, work continues after you close your laptop. It can also run on a schedule or start when something happens. In addition, it can create short-lived helper agents and remember context between sessions.

So where does it live? First, it works inside Gmail, Drive, Docs, Sheets, Slides, Chat and Calendar. It also shows up in Microsoft 365 and Slack. On top of that, it runs on the web, on phones, on desktops and in the command line. Through a tools registry, it can even reach Salesforce, Jira, BigQuery, Snowflake and any MCP server.

Meet the coworker agent

Anatomy of a Google coworker agent

What Google says each agent gets, and the guardrails around it

Identity

AccountIts own Workspace email, calendar, Drive and directory listing

Examplename@agents.company.com

Access

SeesOnly what a user or team shares with it

GuardrailAgent Gateway, an “AI network firewall”

Models

TodayGemini models plus Anthropic’s Claude

DataGoogle says prompts stay inside Gemini Enterprise

Budget

ControlHard spend caps per project

When hitThe agent pauses until you resume

Audit

LogsEvery action attributed to the agent, not a person

SandboxRuns inside its own network boundary

Source: Google Cloud blog, VentureBeat, SiliconANGLE (October 8, 2026). Availability and pricing still being finalized.

Diagram by CVisiona | cvisiona.com

Diagram: anatomy of a Google Gemini coworker agent, by CVisiona

Here is where it gets interesting. A business can describe a role, and Gemini creates a “coworker agent.” That agent gets its own Workspace account, with an email address, calendar and Drive. According to SiliconANGLE, colleagues can @mention it and add it to Chat spaces. Its edits even appear under its own name in a document’s history.

Security is built around that identity. For one thing, each agent acts as itself, not as you. It also sees only what people share with it. Admins approve its permissions, and every action lands in an audit trail under its name.

Google’s Agent Gateway adds another layer. It checks traffic in real time and can enforce rules such as “Agents may not open documents classified Need to Know.” As Kurian put it: “Every Gemini agent has an identity of its own, separate from your identity.”

Google graphic of the Gemini agent, described as a single, universal agent for work
Image: Google (source)

But here’s the twist

The agent does not rely only on Gemini. Instead, Google says it routes work across Gemini models and Anthropic’s Claude, with more models to come. Google also says your prompts and data stay inside Gemini Enterprise, even when Claude does the work.

Futurum analyst Nick Patience calls the Claude option a commercial concession. In his view, it removes an objection for enterprises. However, he notes that the plan also depends on rivals’ platforms. Futurum’s take sums it up: Google is putting the agent ahead of the model.

That tells us something important. Google is not just trying to win the model race. It is trying to own the workplace where those models do useful work.

What it costs, and when you can get it

Google has not published list prices yet. According to VentureBeat, the agent costs nothing extra where Gemini Enterprise is already available. Also, a Google representative told Futurum there will be no per-seat license for coworker agents.

Instead, long-running work is billed by usage, with spend caps for each project. When a project hits its cap, the agent simply pauses. For context, Gemini Enterprise launched at $30 per seat per month in October 2025. If AI bills have ever surprised you, read our breakdown of the hidden costs of AI subscriptions.

Availability is still early. VentureBeat describes a private preview for select customers. Meanwhile, analysts at Constellation and Futurum pointed to a wider launch around the end of October or early November, starting in North America. Versions for finance and legal teams are in preview, while others are “coming soon.”

The race for your next coworker

Google is not alone. Microsoft announced a new Copilot with an “Autopilot” agent. OpenAI launched always-on Dots, and Meta has its Muse agent. Anthropic, meanwhile, just added Claude to Google Docs, Sheets and Slides. We compared three of them in Claude vs ChatGPT Dots vs Meta Muse.

Still, Google’s pitch is scale. As The Register reports, Sundar Pichai said nearly 80% of Google Cloud customers use its AI products. Futurum’s Keith Kirkpatrick, for his part, called coworker agents the most important news of the day. After all, companies can set them up and audit them like employees.

The questions to ask before you hire one

  • Setup: Google has not fully explained how it creates agents, or whether every one gets Gmail, Calendar and Drive.
  • Data: it is unclear how long Google keeps data when Claude handles the work.
  • Region: early access is North America only, which raises data questions in Europe.
  • Trust: there is no independent proof yet that agents can work for days without supervision.

So our advice hasn’t changed. Start with one narrow task and set a budget. Then read why your first AI agent will fail before you give any agent a company email.

For decades, software helped employees do their jobs. Now software companies want to sell something that looks a lot like the employee. Your next coworker might not need a desk. But apparently, it still needs an email address.

Sources & Credits

Enjoyed this breakdown? Stay ahead of the curve with our technical analyses and news. Join our community by subscribing to the CVisiona Newsletter for weekly deep dives, and subscribe to our CloudMelon Vis YouTube Channel for video breakdowns of Cloud-Native, AI, and Tech Enterprise news.

Question for you: Would you give an AI agent its own company email address? Tell us in the comments.

Written By

I'm an entrepreneur and creator, also a published author with 4 tech books on cloud computing and Kubernetes. I help tech entrepreneurs build and scale their AI business with cloud-native tech | Sub2 my newsletter : https://newsletter.cvisiona.com

Leave a Reply

Leave a Reply

error: Protect the unique creation from the owner !