Google's Gemini Agent Brings Multi-Model Work Agent to Enterprise

Google Cloud announced the Gemini Agent on October 8, 2026, at its Gemini at Work 2026 event, positioning it as a single universal agent for work that handles knowledge work, question answering, content creation, and coding from one prompt box, according to Google's announcement. The move reframes Google's enterprise AI pitch: instead of selling a chatbot that answers questions, the company is selling an agent that is given an objective, plans the work, connects to a company's systems, and returns finished results inside the documents, inboxes, and developer environments people already use.

Google Cloud CEO Thomas Kurian unveiled the agent in the event keynote, describing the core interaction as giving the Gemini Agent objectives rather than step-by-step instructions. According to reporting from Unite.AI, the agent holds an organization's business context, uses skills and tools, connects to Cloud customers' business systems, and hands back completed work where employees already spend their day. The pitch is that work can keep running in the cloud for hours or days after a laptop is closed, with the Gemini Agent spinning up temporary sub-agents for multi-step jobs, reported Open Loop News.

One agent, many models

The detail that got the most attention from the developer press is model choice. The Gemini Agent selects the best model for each job dynamically, and today that selection spans Google's own Gemini models and Anthropic's Claude models, with additional open and private models planned, according to Google's announcement. Neowin reported that the enterprise work agent lets an IT team route between Gemini and rival model families to control costs — the agent and the model underneath it are deliberately separate decisions.

That multi-model stance is a notable shift for a company with its own frontier models. Shopify is already blending frontier models for its millions of merchants, according to Kurian's keynote as reported by Unite.AI, and PayPal reportedly routes 10 million multi-model requests every week through Google's stack. The subtext is that the winning enterprise agent may not be the one with the best single model, but the one that can call whichever model fits each task.

Underneath the agent sits what Google calls a borderless Lakehouse, which lets the Gemini Agent query data sources including Amazon S3 and Azure Data Lake without variable egress fees, reported Neowin. The agent can also read directly from Salesforce Data 360, SAP, ServiceNow, and Workday without copying data, and can connect to tools such as Jira, BigQuery, and Snowflake, as well as any MCP server, according to coverage from TechFeatured and Open Loop News.

An email address, an audit trail, and a spending limit

The Gemini Agent is not confined to a chat window. According to Open Loop News, an administrator can describe a role and the system creates a coworker agent with its own Workspace account — its own email address, calendar, Drive, and a place in the company directory. Colleagues can add it to a Chat space, mention it, or tag it in a document comment. TechCrunch reports that the agent's actions are written to an audit trail attributed to the agent rather than to a person, which matters for compliance once software starts doing real work under a company identity.

Governance is the feature Google is selling hardest. The announcement emphasizes built-in cost controls, and coverage from developer outlets notes that administrators can set spending limits per agent — a direct answer to the compliance conversation that starts the moment an autonomous agent has access to company systems and budgets. The Gemini Agent carries the security, administration, and governance enterprise customers require, according to Google's announcement, and industry-specific versions are being specialized for finance, legal, government, healthcare, and retail with unique skills, tools, and knowledge, reported Neowin.

The breadth of surfaces is striking. The Gemini Agent works inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, and also from the web, iOS, Android, Windows, Mac, and the command line — plus outside Google's walls entirely, including Slack, Microsoft 365, and developer tools like Git and Jira, according to Google's announcement and multiple outlets covering the event. As TechFeatured put it in its coverage, nobody's actual workday fits inside one app, which is why the agent's cross-platform memory matters: research assigned on a desktop can be checked as a finished summary on a phone without repeating background context.

The numbers Google is betting on

Kurian's keynote leaned on adoption figures. Nearly 500 Google Cloud customers each processed more than one trillion tokens over the preceding year, nearly 80 percent of all Google Cloud customers are using its AI products, and nearly 90 percent of the Fortune 100 use Gemini Enterprise, according to Unite.AI's account of the keynote. On the customer side, sportswear brand On tested the agent's dynamic model-selection capability, Brazilian bank Bradesco reportedly cut document review time from one hour to five minutes, and telecommunications operator Orange Spain has deployed more than 1,000 custom Gemini Enterprise agents, reported Unite.AI.

Those figures set up the real contest. The week the Gemini Agent launched, Sierra also announced its Personal Agent Protocol, an open standard developed with Meta, Stripe, Shopify, Walmart, and others for how personal AI agents identify themselves to businesses. Between Google's universal agent, Sierra's identity protocol, and milestones like China Telecom's TeleAgent passing 1.5 million users, the enterprise agent stack of 2026 is being assembled in public — and each piece answers a different question. The Gemini Agent answers what the agent does once it is inside the company. The standards being drafted elsewhere are starting to answer who the agent is and whether it can be trusted with the company's money.

For now, Google's answer is the simplest to demo: give the Gemini Agent an objective in the morning, and find the finished work waiting in the inbox. Whether enterprises hand over the keys — and the spending limits — will be the story to watch through the rest of the year.

Sources: Google's Gemini at Work announcement; Unite.AI's coverage of the Gemini Agent; Neowin's event report.