Google Cloud Launches the Gemini Agent, a Single AI Agent for All Your Work
Gemini agent for work: Google Cloud introduced the Gemini agent on Thursday at its Gemini at Work 2026 event — a single, universal AI agent for work that the company says can answer questions, handle tasks, create content and write code. The pitch is that work now starts in the prompt window: the agent holds an organization’s business context, plans the work, uses tools and skills, connects to a company’s business systems, and returns finished work inside the documents, inboxes and developer environments people already use.
What was announced
The agent works directly inside Google Workspace apps — Gmail, Docs, Sheets and Calendar — but Google is not keeping it in-house: it can also be used through Microsoft 365 and Slack, meeting employees inside the tools they already live in. Crucially, it picks the best model for each task, running today on Google’s own Gemini models and Anthropic’s Claude models, with more models to be added later.
Perhaps the most provocative detail is “coworker agents”: custom agents that act as team members, with their own email addresses and access only to the information they are given. Google says the agent ships with built-in cost controls plus the security, administration and governance that enterprise customers demand. Specialized versions for financial services and legal work are now in preview; versions for government, healthcare and retail are coming soon.
Why this matters
The timing is not accidental. OpenAI launched always-on agents called “dots” in September that chase user goals across apps on their own; last month Meta released Muse, a personal AI agent that can shop, book travel, send emails and make payments. The industry’s center of gravity has shifted from chatbots that answer to agents that act — and Google just planted its enterprise flag.
Google’s distribution advantage is real. Workspace is embedded in countless organizations, and an agent that lives inside Gmail and Docs doesn’t need to win a new behavior — it just needs to be good enough at the job. Extending it to Microsoft 365 and Slack is an admission of how people actually work: nobody’s job lives in one vendor’s suite. If the agent is as capable as claimed, the moat is no longer the model — it’s the workflow.
There is also a defensive logic. Every query a worker delegates to an agent is a query that no longer starts in search or in a competitor’s copilot. Owning the prompt window is owning the front door of enterprise AI.
The model-routing bet
Best model for the job
Letting the agent choose between Gemini and Claude — and eventually more models — is Google’s answer to the hardest enterprise AI question: which model do I use for what? Different tasks favor different strengths, and enterprises hate picking wrong. Google is effectively selling model choice as a feature, not a decision. Paired with built-in cost controls, it’s a direct pitch to the CFO: you get frontier performance without frontier waste.
The Anthropic irony
The strangest line in the announcement is that Google’s flagship work agent runs, in part, on Anthropic’s Claude — Google’s fiercest rival in the model race. In practice it confirms where the industry has landed: Anthropic’s models are so good at agentic work that even Google can’t launch its agent story without them. It also tells enterprises something reassuring: this agent is model-agnostic, not a lock-in play. Whether Google’s shareholders see it that way is another question — Alphabet shares were up only marginally in early trading on the news.
Who wins, who loses, what critics say
Google Cloud wins the enterprise narrative. The company already claims nearly 80 percent of its cloud customers use its AI products and nearly 90 percent of the Fortune 100 use Gemini Enterprise. The agent turns that footprint into a daily habit.
Anthropic wins distribution it didn’t have to sell. Every enterprise that adopts the Gemini agent becomes a Claude customer by default. Rivalry aside, that’s revenue.
Microsoft loses some shine. Copilot lives inside Microsoft 365; now Google’s agent lives there too, running on some of the same data. Point-tool AI vendors — the ones selling single-purpose workplace assistants — lose hardest of all, because a general agent that connects to business systems eats their lunch.
Critics see the real risk. Coworker agents with their own email addresses and access to internal systems are a social-engineering and data-exfiltration surface unlike anything enterprises have granted software before. Google promises governance and scoped access, but every IT department hearing “an agent with its own email” will ask who audits what it read — and who pays when it acts wrong. The security story, not the demo, will decide how fast this spreads.
What the numbers actually say
Nearly 500. The number of Google Cloud customers that each processed more than one trillion tokens over the past year — scale that has moved enterprise AI, in Google’s telling, from experimentation to operations.
About 80 percent. The share of all Google Cloud customers now using its AI products, per the company.
About 90 percent. The share of the Fortune 100 that Google says uses Gemini Enterprise — the installed base the new agent will sell into.
10 million a week. The multi-model requests PayPal routes through the platform every week — a live stress test of the routing story.
One hour to five minutes. What Brazilian bank Bradesco cut document review time down to — the kind of concrete productivity number enterprises actually buy on.
Over 1,000. The custom Gemini Enterprise agents telecom operator Orange Spain has already deployed — coworker agents at scale, before they even had the name.
What happens next
Watch availability. The agent starts in private preview; wider access for select Workspace Business and Enterprise plans is promised soon. The speed of that rollout — and the price tag — will say more than the keynote.
Watch the verticals. Finance and legal versions are in preview; government, healthcare and retail are next. Each is a compliance gauntlet, and the first regulated customer to deploy at scale is the real launch.
Watch the agent race. OpenAI’s dots, Meta’s Muse and Google’s Gemini agent are three different answers to the same question — who does the work. The winner won’t be decided by demos but by which agent enterprises trust with their inboxes, their code and their money.
The chatbot era sold answers. The agent era sells finished work — and Google just made its opening bid for the enterprise.