Google Unveils Gemini Agent for Businesses With Cross-App Task Automation

Google has introduced a unified Gemini AI agent that can complete workplace tasks, write code and operate across business applications from a single interface. Announced Thursday, October 8, at a Google Cloud event, the new system is designed to complete assignments rather than answer questions.

According to Google’s official announcement, the agent can plan work, select specialised tools, and connect to company systems to complete tasks. Google is prioritising enterprise customers before expanding the technology to consumers.

CEO Sundar Pichai said Gemini now has more than 1 billion monthly active users, while nearly 90% of Fortune 100 companies use Gemini Enterprise. Google plans to use its existing business customer base to address the security, performance and scalability requirements of autonomous AI agents.

Gemini Can Complete Tasks Across Business Applications

Google Cloud CEO Thomas Kurian explained that businesses can assign Gemini objectives rather than detailed instructions. The agent determines the required steps, uses available tools, and delegates work to specialised subagents when needed.

Gemini connects with Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, PostgreSQL and Snowflake. It also supports Model Context Protocol servers, letting organisations connect additional internal and external systems.

The system automatically selects an appropriate AI model for each task, although users can choose one themselves. Alongside Google’s models, Gemini supports Anthropic’s Claude, with additional open and private models planned for future integration.

Employees can attach files, folders and project materials to requests, giving the agent access to relevant information. A dedicated tasks inbox shows progress, tool usage, delegated assignments, and code execution, letting users follow work as it develops.

AI Agents Get Their Own Workplace Accounts

One notable feature gives Gemini coworker agents their own Google Workspace accounts, including email addresses and access to approved company resources. Employees can assign work by emailing an agent, mentioning it in a conversation or adding it to a group chat.

Agents can use organisational context, including team structures, calendars, approval responsibilities, and time zones. Google says it records actions under each agent’s identity, creating an audit trail that distinguishes automated work from employee activity.

Access will extend across iOS and Android devices, Windows and Mac computers, command-line tools and workplace platforms including Google Workspace, Microsoft 365, ServiceNow and Slack.

Google Adds Enterprise Security and Spending Controls

Google says businesses will retain control over agent permissions, data access and activity monitoring. Dedicated identities, security policies and audit records are intended to help organisations manage automated operations across connected systems.

The company also introduced flexible spending controls, including multi-model orchestration, automated model routing and real-time spending caps. These features help businesses manage the computing costs of running AI agents.

Early testing involved companies including sportswear brand On, Shopify and PayPal. Google also identified BNP Paribas, Bradesco, Merck, Orange Spain, Santee Cooper, SOMPO, Ulta Beauty and Wesfarmers among its Gemini Enterprise customers.

Google has not provided a specific timeline for bringing the unified agent to consumers. Its initial enterprise focus gives businesses access to task automation while allowing the company to develop the operational controls needed for broader deployment.