What is Databricks Genie?

Summary

• Genie is the Databricks AI experience for users: a single place to ask data questions in natural language, explore AI/BI dashboards, and run Databricks Apps. Every answer is grounded in your organization's data and governed through Unity Catalog.
• The Genie family has three parts: Genie One for business users, Genie Agents that data teams configure for a domain, and Genie Code for developers and technical practitioners.
• Genie Agents use a compound AI system to turn a question into SQL. Generated queries are always read-only, and in most cases they run on the agent's SQL warehouse.
• Data access is evaluated as the end user, so Unity Catalog row filters and column masks are enforced per user.

What is Databricks Genie?

Genie is the Databricks AI experience for users. It gives every user in your organization a single place to ask data questions in natural language, explore AI/BI dashboards, and run Databricks Apps. Every answer is grounded in your organization's data and governed through Unity Catalog. The Databricks product page for Genie One describes it as "an AI coworker for business teams."

What are the parts of the Genie family?

Part	What it is	Who uses it
Genie One	The simplified Databricks UI for business users, a single entry point for interacting with data and AI without navigating technical concepts like compute, queries, models, or notebooks	Business users
Genie Agents	Domain-specific environments where data teams configure trusted data, metrics, and business rules that power Genie One answers	Data teams configure them, and business users ask questions
Genie Code	The AI coding and data assistant for developers and technical practitioners in the Databricks workspace	Developers and technical practitioners

How does Genie answer a question?

Genie selects relevant names and descriptions from annotated tables and columns to convert a natural language question into an equivalent SQL query. It responds with the generated query and a results table, if possible, and it can ask follow-up questions when it cannot generate an answer.
Genie Agents use a compound AI system, which combines multiple interacting components instead of relying on a single large language model. The context Genie draws on includes:

• Unity Catalog table metadata, including names, descriptions, and primary and foreign key relationships
• Column names and descriptions
• The agent's knowledge store, example SQL queries, SQL functions, and instructions
• The prompt and response history of the current chat

In most cases Genie generates a SQL query that runs on the agent's SQL warehouse. Generated queries are always read-only.

What data can Genie use?

A Genie Agent is based on data registered to Unity Catalog, including managed tables, external tables, foreign tables, views, metric views, and materialized views. In Chat mode, Genie works with structured data only. It cannot answer questions about unstructured content such as PDFs or Word documents.

How do data teams improve Genie's answers?

Domain experts configure each Genie Agent with datasets, sample queries, and text guidelines. The key concepts in the documentation are:

Concept	Purpose
Knowledge store	Curated semantic definitions such as agent-level table and column descriptions, synonyms, join relationships, and SQL expressions. It is scoped to the agent and does not change Unity Catalog metadata.
Instructions	Plain-text notes on business rules, preferred terminology, and date formats
Example SQL queries	Reference answers that Genie selects from to generate more accurate responses
Trusted assets	Parameterized queries and SQL functions whose logic an author has verified
Benchmarks	Test questions that measure response accuracy. Genie does not use them to improve its context.

A Genie Agent's behavior does not change based on user feedback alone. Authors use the Monitor tab to review questions, responses, and flagged feedback, then update the agent.

How does Genie handle access and security?

A Genie Agent uses two separate credentials. Compute access uses credentials embedded by the agent author, so end users do not need direct warehouse permissions. Data access is evaluated by Unity Catalog as the end user's own identity, not the author's. Each user sees only the data they are authorized to access, and row filters and column masks defined in Unity Catalog are enforced per user. Each query is attributed to the end user in query history and the agent's monitoring page, and a question about data a user cannot access returns an empty response.

FAQs

Is Genie the same as a Genie Agent?

No. Genie is the family of AI experiences. A Genie Agent is one part: a domain-specific natural-language chat interface where users ask questions of their data and get back SQL queries, results tables, and visualizations.

Can Genie change my data?

No. Genie-generated queries are always read-only.

Does Genie work inside dashboards?

Yes. Published AI/BI dashboards include a Genie Agent by default, which adds an Ask Genie button so viewers can chat with the data instead of relying only on predefined visualizations.

Does Genie support languages other than English?

Yes, for example Portuguese and French. The underlying agent framework wraps prompts in English, so Databricks recommends that agent creators add as much metadata as possible in their language, and responses might sometimes appear in English.

Does Genie learn from user feedback?

Not on its own. Users can mark a response as correct, flag it as incorrect, or request a review. Flagging a response as incorrect can regenerate that one response using the feedback provided, but the agent's behavior does not change from user feedback alone. Databricks recommends collecting feedback, and authors use it to identify improvements to the agent's instructions and context.