What Is Databricks Genie?
Databricks Genie is the natural-language AI experience in Databricks that lets business users ask questions of their data in plain language and get answers backed by SQL, as tables or charts. A team sets up a Genie Agent, formerly called a Genie space: a curated set of Unity Catalog tables, views and metric views, plus instructions and example queries for a subject area such as sales or finance. Genie writes SQL against those sources, runs it and returns a table or chart, and the generated SQL can be reviewed.
Genie is built for the people who ask questions, not the people who write queries. The setup work happens once, by someone who knows the data, so the questions can come from anyone.
Genie One, Genie Agents and Genie Code
Databricks now uses Genie as the name for a family of AI experiences. Genie Agents, formerly called Genie spaces, are the curated, subject-area environments this page describes. Genie One is the chat experience for business users. Genie Code is for developers and technical practitioners. Databricks is still changing these names, so check its documentation for the current terms.
How a Genie Agent Works
- Data sources. Unity Catalog objects, including tables, views and metric views. Keeping the set focused on one subject area helps Genie choose well.
- Knowledge store. Curated table descriptions, join relationships and SQL expressions that apply inside the space.
- Instructions. A general instructions block, example SQL queries and SQL functions that show Genie how the business asks and answers questions.
- Trusted assets. Parameterized example queries and functions for well-established questions. When Genie answers through one, the response says so, and people can open it to check the query.
- Benchmarks. Test questions with expected answers, used to measure how well the agent performs as it changes.
How to Build a Useful Genie Agent
- Pick one subject area. A finance agent and a sales agent each do better than one agent that holds everything.
- Start from defined metrics. Add metric views for the figures people ask about, so Genie chooses measures instead of writing aggregations.
- Describe the tables and columns. Business names and plain descriptions help Genie match the words in a question to the right data.
- Write instructions in business terms. Fiscal calendar, currency, which statuses count as open and what common abbreviations mean.
- Add trusted assets for recurring questions. The questions leaders ask every month deserve an approved query.
- Test with benchmarks before a wider rollout. Run real questions, including tricky ones, before the agent goes to a broad audience.
What Makes Genie Answers Reliable
Genie writes SQL, so the quality of its answers depends on how much business meaning its sources carry. Pointed at raw ERP tables, it has to guess which columns and codes make up a figure like Total Sales. Pointed at metric views that define the measures, it chooses a governed calculation instead. Clear descriptions, focused scope, good instructions and benchmarks built from real business questions do the rest.
Genie and Text-to-SQL
Genie is a managed form of text-to-SQL: it turns a question into a query. The difference is the curation around it. Instructions, trusted assets and metric views narrow the choices, so Genie spends less effort guessing at structure and more on matching the question to a defined answer.
Frequently Asked Questions
What is Databricks Genie used for?
Genie lets business users ask questions of data in Databricks in plain language and get answers as tables or charts, without writing SQL. Teams set up Genie Agents for subject areas such as finance, sales or operations.
What is a Genie space?
Genie space is the former name for a Genie Agent, a configured workspace for natural-language questions. It holds the data sources Genie can use, plus descriptions, instructions, example queries and benchmarks that guide its answers.
What are trusted assets in Genie?
Trusted assets are parameterized example SQL queries and SQL functions for questions that come up often. When Genie answers through one, the response shows it came from that approved logic.
Does Genie replace dashboards?
No. AI doesn't replace your dashboards. It's a different way to get at the same governed data. Dashboards suit the views people check routinely, and Genie suits the follow-up questions.
Databricks Genie and QuickLaunch's Approach
Microsoft 365 Copilot, Claude and Databricks Genie answer from the same trusted definitions as your reports. On Databricks, the same definitions live as Unity Catalog metric views, so Databricks Genie answers from them too. For JD Edwards, those metric views are built, along with Databricks dashboards, and every QuickLaunch Application Pack is available on Databricks. See our Databricks page or the Databricks definition.
More plain-English definitions of data, analytics and AI terms are in the QuickLaunch Learning Center.