Gemini Enterprise: Google Unveils New AI Agent for Businesses

Philipp Briel
Philipp Briel · 7 min. read

At its in-house event “Gemini at Work 2026,” Google Cloud unveiled the so-called Gemini Agent: a single, universal AI agent that answers questions, performs knowledge work, creates images and media, and writes and executes code. The agent is an integral part of the Gemini Enterprise platform and will be included at no additional cost in existing Workspace Business and Enterprise subscriptions.

Key Points at a Glance

  • Google Cloud CEO Thomas Kurian unveiled the Gemini Agent on October 8, 2026—a single agent for knowledge work, image generation, and coding.
  • Users provide the agent with goals rather than individual instructions; it plans the work itself, utilizes skills and tools, and delivers finished results directly to documents, inboxes, or development environments.
  • Depending on the task, the agent selects the appropriate AI model on its own—including Google’s own Gemini models as well as Claude from Anthropic.
  • Gemini works directly within Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, and can also be set up as a standalone “Coworker Agent” with its own company email address.
  • According to VentureBeat, the feature is initially launching in a private preview and will be available to customers with existing Gemini Enterprise access at no additional cost.

What is the new Gemini Agent?

Until now, the Gemini offering consisted of several individual components: a chatbot, Workspace integration, and a coding assistant. With the new Gemini Agent, Google is consolidating these functions into a single interface. Instead of formulating step-by-step instructions, Kurian says that in the future, users will simply provide the agent with a goal—and it will handle the rest on its own, including selecting the appropriate skills, tools, and data sources.

Technically, the agent is based on several principles: It is accessible from anywhere (web, iOS, Android, Windows, Mac, Slack, Microsoft 365), runs continuously in the cloud even after the laptop has been closed, and can create temporary sub-agents for more complex tasks that work in parallel or sequentially. Particularly noteworthy: The choice of model is decoupled from the agent itself. Gemini currently orchestrates both Google’s own model family and Claude models from Anthropic—with additional third-party models set to follow.

Gemini in Google Workspace: Personal Assistant, Delegation, and Digital Colleague

Within Gmail, Docs, Slides, Sheets, Chat, and Calendar, Gemini will operate in three ways going forward: as a personal assistant that, for example, independently coordinates meetings with the right group of people without requiring names or email addresses to be specified; as a proactive suggester that identifies delegable tasks in the inbox and hands them off to the agent with a single click; and as a so-called “coworker agent” that takes on its own role within the team. Such a digital colleague is assigned its own Workspace account—complete with an email address, calendar, and storage—and appears under its own name in the company directory and in the version history of documents.

Users who have been using Gemini for some time—for example, via the Gemini app for Windows —will thus gain significantly more autonomy with the new agent than before—including access rights tailored precisely to the respective role.

Data Analysis, Industry Specialization, and Governance

For data and ML teams, Gemini generates PySpark code on demand, creates notebooks, trains models, and resolves pipeline errors independently. Business users without programming knowledge can generate reliable reports by simply asking text-based questions, which are validated through a connection to BigQuery and the company’s own “Knowledge Catalog.” According to Google Cloud, this approach has already improved the accuracy of SQL queries at Bloomberg Media by 63 percent.

In addition, Google is launching industry-specific versions of Gemini in the preview phase: For financial services providers, the agent accesses data sources such as FactSet, LSEG, S&P Global, and SEC filings, and provides confidence scores as well as traceable source citations. For legal departments and law firms, Gemini imports authorization structures directly from document management systems such as iManage or NetDocuments. Industry solutions for government agencies, healthcare, and retail are set to follow.

On the governance side, each agent is assigned its own cryptographically verified identity, complete with role-based access rights. Every action is recorded in an audit trail that is assigned to the agent rather than to an individual. All data traffic also passes through a so-called Agent Gateway, through which companies can enforce company-wide rules centrally rather than on a per-agent basis.

Cost Control and New Models

To keep AI costs from spiraling out of control, Google combines several control mechanisms: an automatic “Smart Routing” feature that distributes tasks to the most cost-effective model, as well as real-time spending limits in the Cloud Billing Console—once these limits are reached, a project agent is automatically paused.

Under the hood, four new model variants are available for this purpose:

Model Purpose
Argon Frontier model for complex reasoning
Flash Optimized for speed and high volume
Omni Generative image and media creation
Gemma Lightweight, open weights, for edge workloads

For images generated using the Omni model, Google continues to rely on SynthID watermarking. However, knowing how to remove such a watermark from Gemini images does not change the fundamental requirement to label generated content.

Google cites the new TPU-8i chip as the infrastructure foundation, which, according to the company, delivers 80 percent better price-performance than the previous generation. A first practical use case outside of traditional data centers: NASA is running the compact Gemma model directly on board a satellite to analyze Earth observation data in real time.

Availability: Who Can Already Use the Gemini Agent?

According to a report by VentureBeat, the Gemini Agent is initially launching in a private preview and will be integrated directly into Gemini Enterprise—at no additional cost to customers who already have access to Gemini Enterprise through existing Workspace Business or Enterprise subscriptions. Google has not yet announced a specific date for widespread availability.

According to Google Cloud, the test customers already mentioned include Deutsche Bank and CME Group in the financial sector, as well as Germany’s Commerzbank, which claims to have automated document quality assurance with Gemini Enterprise and reduced manual review time from 20 hours to one hour.

Context

With the Gemini Agent, Google is responding to a market in which Microsoft has already made its mark with Copilot Studio and Agent 365, OpenAI with ChatGPT Workspace Agents, and Anthropic with its own agent tools. The fact that Google is integrating Anthropic’s Claude—of all things—as a selectable model into its own platform shows one thing above all: In the race for enterprise customers, having one’s own model matters less now than having control over identity, rights, and costs related to the agents. Whether the Gemini Agent will actually close this gap remains to be seen once the private preview is rolled out more widely and independent tests are available.

Frequently Asked Questions About the Gemini Agent

What is the difference between Gemini Enterprise and the new Gemini Agent?
Gemini Enterprise is the overarching platform for enterprise AI on Google Cloud. The Gemini Agent is the new, central interface within this platform that bundles chat, knowledge work, image generation, and coding into a single agent.

Does Gemini now also run on Anthropic’s Claude models?
Yes. Google orchestrates multiple models in the background—including its own Gemini family and Anthropic’s Claude—and automatically selects the model best suited for the task at hand.

Is the Gemini Agent free to use?
For enterprise customers with existing access to Gemini Enterprise or corresponding Workspace Business or Enterprise subscriptions, the feature is expected to be included at no additional cost. For individual users of the standard Gemini app, this announcement does not change anything for the time being.

When will the Gemini Agent be available to all users?
A private preview for enterprise customers is currently underway. Google has not yet announced a date for general availability.

Sources