Google wants its AI to do the work, not just answer questions about it. At its Gemini at Work 2026 event on October 8, Google Cloud introduced a “universal” Gemini agent that combines chat, autonomous task execution and coding in a single interface.
“Work now starts in the prompt window,” Google Cloud chief executive Thomas Kurian said as he presented the product.
What the Gemini agent does
Users give the agent a goal rather than a single instruction. It then plans the steps and carries them out across the tools a company already uses, including Google Workspace, Microsoft 365 and Slack.
- Helper agents: For large jobs, it spins up temporary sub-agents, each focused on one task.
- A digital “coworker”: The agent can have its own Gmail address, calendar and file storage, and it appears under its own name in document edit histories.
- Persistent memory: It runs in Google’s cloud and remembers context across tasks.
- Model choice: Customers can use Google’s Gemini models or Anthropic’s Claude models, or let the agent pick a model for each task.
Built-in guardrails
Enterprise buyers worry about AI agents acting without oversight, so Google emphasised controls:
- Every agent has its own identity, role-based access and activity log.
- Admins set company-wide rules once, and an “Agent Gateway” enforces them on all agent traffic.
- Per-project spending caps pause an agent automatically when a budget is reached.
- Simple tasks can be routed to cheaper models to reduce costs.
Integrations
The agent connects to Salesforce, ServiceNow, Jira, Confluence, Git, BigQuery, Databricks, Snowflake and Postgres. It also works with any service that supports the Model Context Protocol (MCP), the open standard for linking AI tools to data and apps.
Pricing and availability
Google says the agent will be part of Gemini Enterprise at no additional charge, including for customers who access it through Workspace business and enterprise plans. For now, access is limited to a small group of companies.
Industry versions for financial services and legal teams are in preview. Versions for government, healthcare and retail are planned.
Why it matters
The launch puts Google in direct competition with Microsoft, Salesforce and AI labs that are all racing to sell “agentic” software to businesses. Analyst Carmi Levy told Computerworld that Google is trying to become the gatekeeper of enterprise AI agents.
The real test will be reliability. Companies will adopt autonomous agents widely only if they complete tasks accurately, stay within budget and leave a clear audit trail.
Explore more AI and gadget news in our Tech section.