Databricks Genie+36 Positive

GILD Review

GILD's call and the buying facts behind it.

From GILD research

Should we adopt Databricks Genie?

Can your data team curate Databricks Genie well enough that business users trust its numbers?

30-Second TakePick Databricks Genie if your data lives in Unity Catalog and a data analyst can own and test each domain's agent.
Best fitCompanies on Databricks with Unity Catalog data, metric views, and a data team ready to curate and benchmark each domain.GILD Review
Consider another workflowIf your data spans several platforms, evaluate Hex. If it lives in Snowflake, evaluate Snowflake CoWork. If your metrics already live in LookML, evaluate Looker's Conversational Analytics.GILD Review

Drawn from 93 sources and 2 operator insights

GILD analysis

Databricks Genie Review

Pilot Genie if your data lives in Unity Catalog and a data analyst can own and test each domain's agent. In 74 public accounts, practitioners agree curated context decides whether Genie answers correctly, and they split on how much review answers need. Every answer runs as read-only SQL under the asker's own Unity Catalog permissions.

AI Data Analysis
AI Adoption

The decision

Best fit: Companies on Databricks with Unity Catalog data, metric views, and a data team ready to curate and benchmark each domain.
Consider another workflow: If your data spans several platforms, evaluate Hex. If it lives in Snowflake, evaluate Snowflake CoWork. If your metrics already live in LookML, evaluate Looker's Conversational Analytics.
Adoption condition: Curate a Genie Agent for each domain and get it passing its benchmark set. Set a Genie budget. Finish both before Genie One reaches business users.

Basis: Databricks documentation and pricing, and public accounts from data teams. These include Adam Gilmore's test of Genie with no curated context and Databricks' Summit pages on the accuracy and latency Atlassian and GSK report. Measure in a pilot: score one agent's benchmark questions before curation and again before launch, and widen access only on a passing score.

Buying & deployment factsWhat business and product leaders should verify before rolling it out across teams.

How you buy it

January 31, 2027

Genie One and Agents are free for named users until then. Then 150 free DBUs a month each.

Source 1, Databricks, Genie pricing: Genie pricing Source 2, Databricks, Budgets and cost controls: Budgets and cost controls

Details

Genie is billed in DBUs with no seat fees, and Databricks prices it for Premium and Enterprise accounts. Genie One and Genie Agents usage by named users is free through January 31, 2027. From February 1, 2027, each user gets 150 free DBUs a month, worth US$10.50 on AWS US East, and pays US$0.070 per DBU beyond that. Databricks has billed Genie Code usage beyond a free monthly allowance since July 8, 2026.

Cost drivers

Warehouse queries

The SQL Genie runs is billed separately and outside Genie budgets.

Source 1, Databricks, Genie pricing: Genie pricing Source 2, Databricks, Budgets and cost controls: Budgets and cost controls Source 3, Databricks, Create and manage a Genie Agent: Create and manage a Genie Agent Source 5, databricks.com, Lakehouse (Databricks SQL) pricing: Lakehouse (Databricks SQL) pricing

Details

LLM usage beyond the free allowance is billed under the Serverless Real-time Inference SKU, and service principals get no free usage. The SQL that Genie generates runs on a pro or serverless SQL warehouse, billed separately and outside Genie budgets. Serverless SQL costs US$0.70 per DBU on AWS Premium, US East.

Where it runs

Your workspace

Agents query Unity Catalog on your SQL warehouses. Model partners retain no prompts.

Source 3, Databricks, Create and manage a Genie Agent: Create and manage a Genie Agent Source 6, Databricks, Genie One: Genie One Source 7, docs.databricks.com, Use Genie on mobile: Use Genie on mobile Source 8, Databricks, Partner-powered AI features: Partner-powered AI features Source 4, Databricks, AI trust and safety: AI trust and safety

Details

Genie Agents query Unity Catalog data on the customer's SQL warehouses inside its Databricks workspace. Genie One runs at workspace or account level on the web, and in the Genie mobile app for iOS and Android, which is not yet generally available. Chat mode uses Azure OpenAI and OpenAI on Databricks. Agent mode uses OpenAI and Anthropic on Databricks. Model partners retain none of the data submitted through these features.

What to verify

Shared credentials

Users share the author's embedded warehouse credentials, while data access follows the asker.

Source 3, Databricks, Create and manage a Genie Agent: Create and manage a Genie Agent Source 8, Databricks, Partner-powered AI features: Partner-powered AI features Source 9, docs.databricks.com, Designated Services: Designated Services

Details

Unity Catalog data access follows the asker, while every user shares the author's embedded warehouse credentials. Genie Agents require partner-powered AI features. Databricks is removing that setting at a later date, and where it is enabled its toggle is already disabled. Agent mode runs in-Geo in the Americas, Asia, Australia and New Zealand, Europe, and India, and needs cross-Geo processing only in Mainland China and the Middle East and Africa.