Rohan Parikh

Forecasting sales at more than 7,000 retail stores generates frequent inquiries into forecast drivers, and traditional methods like driver dashboards or SHAP fail to give field leaders clear narratives. Data scientists were spending days writing ad‑hoc SQL to answer basic “why” questions. In partnership with Databricks, we built an AI/BI Genie-powered Research Agent that delivers fast, plain‑language forecast explanations using governed data. The agent joins core datasets, computes variances and accuracy metrics, correlates key drivers and links every answer to underlying SQL, cutting analysis time from days to minutes and revealing model biases. This session covers how we used Unity Catalog as the semantic backbone and encoded business rules in the Genie Knowledge Store so the agent reasons like a retail analyst. We’ll demo the agent with Deep Research and share key lessons and a repeatable blueprint for operationalizing Genie.