Google Cloud launches Gemini agent for enterprise work

Google Cloud Chief Executive Thomas Kurian said Gemini can answer questions, handle knowledge work, create images and media or turn ideas into finished products. “You give it objectives, not instructions. You delegate an outcome and come back to finished work,” Kurian said. He added that for an agent to do that, it must be connected to personal workflows, systems of record and enterprise controls.

The agent is designed around persistent execution. It maintains a single set of memories, context and personalization across all communication channels and with all its sub-agents, the report said. It can act as a teammate that remembers what users tell it and can create smaller coworker agents with their own identities, expertise, @agents.company.com email addresses and persistent storage. Those sub-agents can access only the context provided by the user or team members.

Gemini agent also provides model flexibility. Each job runs on a model suited to the task, with simple tasks potentially handled by Gemini Flash and long-horizon work by a flagship frontier model such as Argon. Anthropic PBC’s Claude models are available today, and Google Cloud said other leading private and open models will come later.

Kurian described a system intended to understand business context ranging from pricing and product portfolios to departmental norms. According to the report, Gemini observes and contains institutional knowledge and grounds answers through company data. It connects securely to collaboration tools including Confluence, Microsoft Office, Teams, Slack and Workspace; development tools such as Git and Jira; enterprise platforms including Salesforce and ServiceNow; databases such as BigQuery, Databricks, Postgres and Snowflake; and files on desktops.

Users can build reusable skills that serve as instructions, knowledge and workflows teaching agents to perform specific multistep tasks. Gemini ships with a global library of skills, but users can add and publish custom skills to a shared company registry, the report said. They can also ask an agent to convert a task into a reusable template. The system learns from every question and objective, with session memory that can run for days, communicate with people, work with other agents and read documents. Google said it continues to update its understanding of how users work with it and what it should produce.

Google Cloud is extending Gemini with specialized skills, starting with data. For data scientists and engineers, a user can describe an outcome in plain language and Gemini can generate PySpark code, provide a notebook to edit and test it, train a model and troubleshoot issues on its own, according to the report. For business users, Gemini can use operational BigQuery reporting skills and the Knowledge Catalog to construct and save queries, allowing real-time operational reports to be created simply by asking. Once saved, teams can run reports without incurring additional token costs and produce consistent, verified answers, the report said.

Google Cloud also emphasized cost controls. Per-token prices have fallen by about 98% since 2024, but enterprise AI volume has greatly increased, the company said. Google announced flexible spending options in August, and those are coming online today. Internally, multimodel orchestration selects the right model for a job or breaks complex jobs across quick models for small tasks and frontier models for difficult parts. Smart routing automatically triages enterprise workloads so each runs on the model that delivers maximum performance at the lowest possible cost, according to the report. Users can set hard limits on AI spending in the Cloud Billing Console. Gemini monitors token usage, and if a spending cap is triggered, the agent pauses; the user can resume work in the console with approval.

Editor's Summary Google Cloud introduced Gemini agent, an autonomous AI assistant that works across devices, apps and enterprise tools with persistent memory and reusable skills. The system offers model choice, including Gemini Flash, Argon and Anthropic Claude models, and supports spending caps in the Cloud Billing Console. Google said per-token prices have fallen about 98% since 2024 while enterprise AI volume has risen.