Context-ready agents to automate work

Genie agents are expert-curated AI agents that deliver trusted answers on specific, high-priority domains of your business. They reason across governed tables, unstructured documents in Unity Catalog volumes and enterprise tools in a single conversation, and they use your organization's definitions, metrics and terminology to give accurate, cited answers.

Genie Agents is available to all existing Databricks Lakehouse (Pro and Serverless) customers and requires your data to be managed in Unity Catalog.

Genie Agents is available in Databricks Lakehouse — $DBU rates apply.

Genie Agents is designed for business and nontechnical users who want to analyze and visualize data without relying on expert data analysts. Subject matter experts curate each agent with example queries and instructions so it can answer domain-specific questions accurately for the teams that use it. Because Genie Agents can be embedded into existing business applications like Microsoft Teams, Slack or Glean through its API, users can get answers directly within the tools they already use each day.

You can start by watching a short demo or signing up for a free trial using the express setup, which includes a sample bakery dataset you can explore right away. Existing Databricks Lakehouse (Pro and Serverless) customers with data in Unity Catalog can also create a Genie agent directly in their workspace, adding expert-curated instructions, sample questions and a knowledge store to improve accuracy. For questions about enterprise deployment, you can also contact the Databricks sales team.

Genie Agents is the evolution of Genie Spaces. It’s the same product with expanded capabilities. There’s no migration and no new SKU: existing Genie spaces automatically gain the new features, including unstructured data analysis, MCP connectors and sharing.

Genie One is the AI coworker for every employee. It answers questions across your entire data estate. Genie agents is purpose-built for a specific domain, such as finance, HR or sales, and curated by the subject matter experts on your team. Genie One automatically routes questions to the right Genie agent when it needs deeper domain analysis.

Genie agents are grounded in Genie Ontology, which supplies your business definitions, metrics and table relationships. Subject matter experts add instructions, SQL expressions and example queries. Benchmarks act like unit tests: you define questions and gold-standard answers, score accuracy in Agent and chat mode, and accept suggested improvements. The monitoring tab and Genie Code help you find and close context gaps in production.

Genie Agents can query any data registered in Unity Catalog, including structured tables and unstructured files stored in Unity Catalog volumes. This full integration with Unity Catalog means Genie Agents inherit your organization’s existing access-control policies, so answers stay within the bounds of each user’s permissions. Because of this dependency, using Genie Agents requires that your data already be managed in Unity Catalog.

Yes. With Open Sharing, you can share a governed Genie agent preloaded with your data, instructions and knowledge with partners, suppliers and customers across Databricks accounts. With Genie Clean Rooms, recipients can ask questions and get answers without seeing the underlying data, and your data and IP stay protected.