Google doesn’t want its AI to just answer your questions anymore. It wants it to do your job.
The company unveiled Gemini agent on Thursday at its Gemini at Work 2026 event: a universal AI assistant built to handle real knowledge work, from answering questions and handling tasks to creating images and media, writing code, and running it. You hand it an objective, not a step-by-step brief. It plans the work, picks the right tools, connects to your company’s systems, and delivers the finished product. It keeps working in the cloud after you close your laptop. Google says longer tasks can run for hours or days without another briefing.
The pitch is timed to a crowded field. OpenAI launched always-on agents called “dots” in September that chase user goals across apps on their own. Meta released Muse, a personal AI agent that shops, books travel, sends emails, and makes payments. Microsoft revamped its own agentic offering last month. Now Google is in, and Wall Street’s reaction told the story: Alphabet shares rose just 0.2% in late-morning trading before surrendering most of the gain. Barron’s put it dryly: the muted move suggests investors figured it was only a matter of time.
Gemini agent for work: one agent, your whole stack
The technical bet is that one agent should handle everything instead of a fleet of narrow bots. Gemini agent plans the work, uses tools, connects to a company’s business systems, and returns finished work in documents, email, and developer environments, Google said. It works inline inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, carrying the same memory, skills, and controls everywhere.
One detail that stands out: the agent isn’t locked to Google’s own apps. It can also be used through Microsoft 365 and Slack, connects to development tools like Git and Jira, enterprise platforms like Salesforce and ServiceNow, and databases including BigQuery, Databricks, Postgres, and Snowflake. A company can even register its own tools and custom skills in an enterprise registry. The tools layer also reaches files on a user’s own desktop and any Model Context Protocol server inside or outside the company network.
Memory is the other big piece. Google says Gemini keeps four kinds of memory (session, semantic, procedural, and episodic) and onboards itself “the way a new hire would,” learning the user, the tools, and the team before starting work. Creepy or useful, depending on your inbox.
Coworker agents get their own email addresses
Here’s the part that would make any IT admin blink. Users can create “coworker agents” that act as team members, with their own email address and access only to the information they’re given. In coworker mode, the agent gets a full Workspace identity (email, calendar, Drive, and a presence in the company directory) and acts under its own identity rather than the user’s, following the sharing and membership settings the team already uses.
Google also built three modes into the Workspace experience. A personal-assistance mode arrives already briefed on your calendar, team, and projects. A proactive-delegation mode recognizes delegatable tasks and offers a one-click handoff. Then there’s the coworker mode described above — essentially a digital hire.
Gemini can also spin up temporary sub-agents with their own identities for different parts of complicated jobs, working simultaneously or sequentially. Thomas Kurian, CEO of Google Cloud, wrote in the announcement blog post: “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, or on behalf of a specific role in an organization, like an analyst in your finance department.”
The multi-model twist: Claude is invited
Perhaps the most telling detail is under the hood. Gemini agent picks the best model for each task — and it currently runs on Google’s Gemini models and Anthropic’s Claude models, with other private and open models to be added later. Yes, Google’s flagship work agent delegates tasks to a rival’s model when it fits better. Google frames it as Smart Routing: matching the model to each task for quality and cost, with real-time spend caps to keep CFOs calm.
The enterprise pitch is aimed squarely at the verticals: financial services and legal versions are in preview now, with government, healthcare, and retail coming soon. That matters because the agent race is quickly becoming a compliance race — and Google is foregrounding identity and policy management, authorization and permission controls, secure sandboxing, and network gateways. Enterprises won’t hand an AI its own email address without all of that.
Google name-checked some heavy users to back up the pitch. Sportswear brand On tested the agent’s dynamic model-selection; Shopify blends frontier models across millions of merchants; PayPal routes 10 million multi-model requests a week. Brazilian bank Bradesco cut document review time from one hour to five minutes, and Orange Spain has deployed more than 1,000 custom Gemini Enterprise agents, according to Google’s announcement. Numbers like that are why Google feels confident calling this “headcount.”
Why this matters
This is the moment the AI agent race moves from demos to payroll. Chatbots were a feature; agents are headcount. Google is betting that the company which controls the tools workers already live in — Gmail, Docs, Sheets — can turn that distribution into an army of digital coworkers. The near-80% figure is Google’s, not mine, but its claim is that nearly 80% of all Google Cloud customers use its AI products and almost 90% of Fortune 100 firms use Gemini Enterprise. That’s a runway OpenAI’s apps don’t have.
But there’s a real tension here. Companies rarely have just one suite of software — and Google knows it, which is why Microsoft 365 and Slack support made the announcement. The fight is over who orchestrates the work across all of it. OpenAI has the momentum. Meta has the consumers. Microsoft has the enterprise lock-in. Google has the Workspace distribution and now, a genuinely multi-model agent that doesn’t care whose logo is on the model underneath.
Don’t expect this to settle quietly. The agent wars just went from a skirmish to the main event.
Frequently asked questions
What is Google’s Gemini agent? Gemini agent is a universal AI assistant for work unveiled by Google Cloud on Oct. 8, 2026. Instead of just answering prompts like a chatbot, it plans work, picks tools, connects to business systems, and returns finished results in documents, email, and developer environments.
How is Gemini agent different from ChatGPT or a regular chatbot? Chatbots respond to each prompt; Gemini agent takes an objective and executes the whole workflow on its own. It can create temporary sub-agents for different parts of a job, and coworker agents that hold their own email addresses and work like team members.
Does Gemini agent use only Google’s AI models? No. Google says Gemini picks the best model for each task, currently orchestrating across Google’s Gemini models and Anthropic’s Claude models, with other private and open models to be added later — partly a cost-control move called Smart Routing.
When will Gemini agent be available? Industry-specific versions for financial services and legal work are in preview now, with government, healthcare, and retail versions coming soon. Gemini also works inline inside Google Workspace apps like Gmail, Docs, Sheets, and Calendar.
Sources: Reuters, Unite.AI, Google Cloud Blog.