Announced at Gemini at Work 2026, the agent plans tasks, connects to company systems and returns finished work inside the tools employees already use
Google Cloud has introduced the Gemini agent at its Gemini at Work 2026 event. It describes the agent as a universal agent for work, built to answer questions, handle knowledge work, create images and media, and write and run code from a single prompt box. Google Cloud CEO Thomas Kurian made the announcement in his keynote, saying work now starts in the prompt window.
How it works
Users give the agent objectives rather than step-by-step instructions. It plans the work, picks the skills and tools it needs, connects to company systems and brings back finished output inside the documents, inbox and developer environments people already use. It can be assigned work, scheduled tasks or set to respond to events.
Key design points include:
Runs in the cloud: The agent keeps one set of memory and context across devices and channels, including web, mobile, desktop, command line, Google Workspace, Microsoft 365 and Slack. Long jobs continue after a laptop is closed.
Sub-agents and “coworkers”: It can spin up temporary sub-agents for multi-step jobs. It can also act as a coworker agent with a persistent role and its own identity, including an email address, storage and a place in the company directory.
Model choice: The agent picks the model that suits each job. Today that means Gemini models and Anthropic’s Claude models, with other private and open models planned.
Connections: It links to tools such as Slack, Microsoft Office and Teams, Confluence, Jira, Git, Salesforce, ServiceNow, BigQuery, Databricks, Snowflake and Postgres, plus any MCP server. Teams can publish their own tools and skills to shared company registries.
Inside Workspace and data tools
Gemini works directly in Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar with the same memory and controls. It can handle personal tasks, suggest tasks to delegate and act as a team member. New data and analytics skills let engineers generate PySpark code and train models, and let business users build reports by asking in plain language.
Industry versions
Specialised versions for Financial Services and Legal are in preview, with Government, Healthcare and Retail coming soon. The financial version draws on data from providers such as FactSet, LSEG and S&P Global, and Google says CME Group and Deutsche Bank are already using it.
Security and cost controls
Every agent gets its own identity, least-privilege permissions and an audit trail. Agents run inside a sandbox, and an Agent Gateway enforces organisation-wide policies on all agent traffic. On cost, Google offers multi-model orchestration, Smart Routing and real-time spend caps. When a project hits its cap, its agent pauses until an administrator resumes it.
Adoption claims
Google says nearly 80% of its Cloud customers use its AI products and nearly 90% of the Fortune 100 use Gemini Enterprise. Customers named include BNP Paribas, Ryanair, Verizon, The Home Depot and India’s Tata Steel, which Google says deployed more than 300 AI agents in nine months.
Pricing and general-availability dates for the Gemini agent were not given in the sources.