From Data to Dialogue: How S&P Global Energy Made Its Structured Data Estate Conversational with Databricks Genie Agents and MCP

This post details how S&P Global Energy partnered with Databricks to transform their structured data estate into a conversational interface using Databricks Genie Agents and the Model Catalog Platform (MCP). The core technical challenge was enabling natural language interaction with complex, domain-specific energy data. This was achieved by developing custom agents that understand energy terminology and can translate natural language queries into executable Spark SQL. The process involved fine-tuning LLMs on domain-specific data, building robust prompt engineering strategies, and integrating with the Databricks SQL engine for query execution. The outcome is a system that democratizes data access, allowing users to query vast datasets conversationally, thereby accelerating insights and decision-making within the energy sector.