How Databricks’ marketers use data 3x more with Genie, an AI analytics assistant

Marketing teams often face delays when seeking trusted data to inform decisions, despite aiming to be data-driven. Databricks addressed this by unifying marketing data in a governed lakehouse and developing Marge, a conversational analytics assistant built on Genie Agents. Marge enables marketers to ask questions in natural language and receive governed answers in seconds, transforming how the team accesses and uses data.

The solution required a structured approach: creating a Marketing Lakehouse as the single source of truth, aligning metrics across systems, and embedding Marge into existing workflows. Unity Catalog provided centralized governance, ensuring role-based access and data lineage. Genie One served as the unified interface, routing requests to specialized agents for domains like web performance and marketing planning, while Agent mode handled complex multi-step analyses.

Trust was prioritized through four mechanisms: clear data models with centralized metadata, verified logic for high-value questions, example question-and-query pairs for common scenarios, and behavioral guidance for interpreting company-specific terminology. These steps reduced ambiguity and improved answer reliability, with users able to provide feedback on responses.

Maintenance remained lightweight, with one BI manager dedicating about an hour weekly to review feedback and monitor performance. This effort reduced the rate of flagged incorrect answers by 25%, demonstrating the system’s effectiveness in sustaining adoption and accuracy.