Google Cloud on Thursday introduced Gemini agent, a unified artificial intelligence assistant that the company says can act autonomously, generate code and complete work across web, mobile and desktop devices.
The agent is designed to be accessible through any channel, including the command line, Google Workspace, Microsoft 365 and Slack. According to Google Cloud, it also operates inside third-party applications without requiring a dedicated interface, part of an effort to make the assistant present wherever a user works.
Thomas Kurian, chief executive of Google Cloud, said the product can answer questions, handle knowledge work, create images and media, and turn ideas into finished products. He described the interaction model as one in which users give the agent objectives rather than instructions, delegate an outcome and return to completed work. Kurian added that achieving this requires connecting the agent to personal workflows, systems of record and enterprise controls.
The launch reflects a broader industry push toward persistent execution. Gemini agent maintains a single set of memories, context and personalization across every communication channel and all of its sub-agents. It functions as a teammate that retains 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 a user or team members.
Google Cloud said the system offers flexibility in model choice, running each job on a model suited to the task. Simple work may run on Gemini Flash, which the company describes as a small, quick model for day-to-day tasks, while long-horizon work may use a flagship frontier model such as Argon. Anthropic PBC's Claude models are available today, with other leading private and open models planned later.
Gemini connects securely to collaboration tools including Confluence, Microsoft Office, Teams, Slack and Workspace, as well as Git and Jira, enterprise platforms such as Salesforce and ServiceNow, databases including BigQuery, Databricks, Postgres and Snowflake, and desktop files. Users can build reusable skills that serve as instructions, knowledge and workflows for multistep tasks, drawing on a global skill library or publishing custom skills to a shared company registry.
Google Cloud also highlighted cost controls, noting that per-token prices have fallen roughly 98% since 2024 while enterprise AI volume has risen sharply. Flexible spending options announced in August are coming online today. Multimodel orchestration and smart routing direct each workload to the model that delivers maximum performance at the lowest cost, and users can set hard spending limits in the Cloud Billing Console. Gemini monitors token usage, pauses the agent when a cap is reached and requires approval in the console to resume, which accepts exceeding the budget. Because tracking is per project, companies can charge AI costs back to specific departments and plan budgets by team.
More AI news from TechManNews.
