Google unveils Gemini agent for long-running enterprise work

Google Cloud has introduced a universal Gemini agent designed to handle enterprise tasks that can span several applications and continue for hours or days. Carl Franzen reports for VentureBeat that the system will work across Google Workspace, Microsoft 365, Slack and other business tools.

Google positions the agent as a single interface for larger assignments. Employees could ask it to research a topic, use company data to build a financial model and prepare a presentation. The agent is intended to decide which steps and connected systems it needs, rather than requiring users to direct each action.

The company also plans a persistent “digital coworker” setup. In this configuration, an agent can receive its own Workspace identity, email address, calendar, Drive storage and place in the company directory. Teams could assign work to an Event Planner Agent, for example, in much the same way they contact a human colleague.

Security controls and model choice

Google says each agent has a cryptographically attested identity, enterprise-managed permissions and audit logs for its actions. Coworker agents can access only information that team members explicitly share with them, according to the company. An Agent Gateway is meant to apply company policies to connections with outside systems, while sandboxes isolate code execution.

The platform supports Google’s Gemini models and Anthropic’s Claude models, with more proprietary and open-weight models planned. Smart Routing is designed to select models according to task complexity and cost. Google also offers project spending caps that can stop agent activity once a budget is reached.

Integrations include Salesforce, ServiceNow, Jira, Git, BigQuery, Snowflake, Databricks and Microsoft Teams. Google says the agent can also use Model Context Protocol servers and organize temporary groups of specialized sub-agents for complex tasks.

However, Google has not disclosed a general availability date, detailed licensing terms or a separate price for the universal agent. It also has not provided independent benchmarks for reliability on lengthy, cross-application workflows. Those gaps leave key deployment and cost questions for enterprise buyers.