Google's Gemini Agent Wants to Be Your Coworker — Email Address and All

THURSDAY, OCTOBER 8, 2026 — Google wants to give you a coworker. Not a chatbot, not a sidebar plugin — an actual agent that holds your company's business context, plans the work, uses your tools, and hands back something finished inside the documents, inbox, and developer environments where you already live. That was the centerpiece of Google Cloud's Gemini at Work 2026 event on Thursday: the Gemini agent, a single, universal AI agent for work, built around one prompt box. The pitch is simple enough to fit on a sticky note: give it objectives, not instructions. Delegate an outcome, come back to finished work.

02 One Prompt Box, Everywhere You Work

The ambition here is that the agent never asks you to come to it. Gemini works directly inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar — inside the email thread, the document, the chat space — carrying the same memory, skills, and controls it has everywhere else. And Google is notably not gatekeeping it inside its own walls: the agent also works through Microsoft 365 and Slack, and can run headless, integrated into third-party apps, without any dedicated interface at all. Web, iOS, Android, Windows, Mac, command line — it doesn't care where you are.

The part that'll sound like science fiction to anyone still hand-feeding AI a single prompt at a time: it can run for hours or days. Work keeps running in the cloud after you close your laptop, and it's still there when you come back. You can assign it work, schedule tasks, or have it respond to events. This is the same territory Apple was thinking about this week when it tightened macOS security around AI agents — once an agent can act for hours unattended across your systems, the stakes of what it can touch get real in a hurry. Google's answer is the enterprise pitch: identity and policy management, authorization and permission controls, secure sandboxing, and network gateways — the security, administration, and governance checklist companies demand before they'll hand an AI the keys.

03 It Picks the Best Model for the Job — Even a Rival's

Here's the flex that sets Google's pitch apart: the Gemini agent isn't loyal to Gemini. The agent and the model underneath it are separate choices. Each job runs on the model that fits best — orchestrating across Google's own Gemini family and Anthropic's Claude models today, with other leading private and open models to be added later. Google's logic is refreshingly blunt: the best model for a task isn't always the largest one. Matching the model to the work raises accuracy on the hard jobs and lowers cost on the simple ones — and since the leading model changes every few months, keeping the choice open means your context, skills, and data stay put when the leaderboard flips.

That multi-model pitch already has early believers. Sportswear brand On tested the dynamic model-selection capability to accelerate its speed to market. Shopify blends frontier models for millions of merchants, and PayPal routes 10 million multi-model requests every single week. And because this is enterprise software, Google paired the routing with built-in cost controls — multi-model orchestration, Smart Routing, real-time spend caps — the boring-but-essential stuff that keeps the CFO from having a heart attack when an agent starts thinking for eight hours straight.

04 Coworkers With Email Addresses

The part everyone will be talking about: "coworker agents." You can create an agent that acts as a team member — its own email address, its own calendar, its own Drive, a seat in the company directory — with access to only the information it's given, according to Google's announcement. Colleagues @mention it in a chat space like anyone else. As Google Cloud CEO Thomas Kurian put it in his announcement: "Gemini can function as your personal assistant or as a team member, where it works on behalf of a group of people, like a project manager within a team."

It gets weirder in the good way: these agents get their own persistent storage, their own @agents.company.com-style identities, and they can delegate and coordinate sub-agents themselves — temporary, job-specific helpers that tackle multi-step work in parallel or in sequence. A marketing manager can ask an events-coordinator agent in a group chat to draft a launch-readiness document and find it posted back to the group when it's done. Tag it in a document comment and it replies in the thread, appearing under its own name in version history. That's not a tool you use; that's a colleague you delegate to. Google's other AI story today was a school district blocking its AI answers — the coworker pitch is the opposite end of the same moment: AI that doesn't just answer, it joins the team.

05 The Receipts: 500 Customers, Trillions of Tokens

Kurian didn't just pitch the vision — he brought numbers. In the last year, nearly 500 Google Cloud customers each processed more than one trillion tokens. Nearly 80% of all Google Cloud customers are now using Google's AI products. And nearly 90% of the Fortune 100 use Gemini Enterprise. Whatever you think of the agent hype cycle, this is the part where you notice the floor has already moved: organizations aren't experimenting anymore, they're running the business on it.

The customer scoreboard has some genuinely eye-popping entries. Brazilian bank Bradesco cut document review time from one hour to five minutes. Orange Spain has deployed more than 1,000 custom Gemini Enterprise agents across HR, IT, sales, and customer service — with non-technical staff building them on their own. And Google says industry-specialized versions for financial services and legal work are now in preview, with government, healthcare, and retail versions coming soon. If the general-purpose agent is the platform, the industry versions are where Google thinks the real money is.

06 When You Get It — and the Race Around It

Here's the fine print: the Gemini agent arrives in private preview, with wider availability for customers on select Workspace Business and Enterprise plans coming soon. No public launch date yet, no pricing details. So the headline is real, but most of us will be watching the pilot customers for a while before we get a prompt box of our own.

The timing is no accident. This lands in the middle of the most intense agent land-grab the industry has seen. OpenAI launched always-on agents called "dots" in September that chase user goals across apps on their own. Earlier last month, Meta released Muse, its personal AI agent that can shop, book travel, send emails, and make payments. Google is now staking out the enterprise as its battleground: not the personal assistant for everyone, but the coworker that plugs into your company's systems with your company's controls.

Investors, for now, are mildly interested. Alphabet shares were up marginally in early Thursday trading, per Reuters, and up 0.2% by late morning, per Barron's. That's the market saying: nice announcement, show us the customers. The coworker with an email address is a great story. Whether companies actually hire it — and what it does on its first day — is the part that matters.