The company highlights user feedback and invites readers to try Agent Mode and provide input.

Databricks unveils Genie Agent Mode, an AI-driven assistant that mimics a human data analyst by planning, hypothesizing, and iterating explanations to tackle complex data questions rather than just returning single-query results.

Launched as part of Genie spaces during Week of Agents, Agent Mode enables planning, exploration, and reasoning over data to answer business questions.

It delivers actionable recommendations to improve performance and supports sharing reports within the platform or as PDFs.

Agent Mode provides real-time, actionable insights on issues like churn spikes, campaign spend optimization, and revenue impacts from supply disruptions.

The approach uses iterative hypothesis generation, querying, and reflection to reach well-supported conclusions, producing a report with visuals, quantified findings, and SQL references.

Potential benefits include reduced reliance on specialized data teams, higher customer value realization, and lower churn as adoption scales within organizations.

In customer-support scenarios, Agent Mode verifies spikes, identifies contributing factors, and uses Genie context, including Unity Catalog metadata and user semantics, to focus analyses.

Agent Mode improves accuracy across simple and complex questions, validates data understanding for straightforward queries, and scales reasoning to handle more complex tasks.

The feature is available now, with plans to add API support and unstructured document analysis in future updates.

The mode plans an approach, tests hypotheses, and iterates toward explanations instead of returning a single query.

Hypotheses are evaluated through multiple data queries guided by Genie’s knowledge store semantics, with transparency so users can verify accuracy.