Scottish Water Drives Conversational Data Access with Databricks Genie & Copilot

Scottish Water faced challenges with fragmented access to critical capital investment data, requiring teams to navigate numerous reports or rely on data specialists. This often led to duplicated efforts, slow decision-making, and valuable information not reaching non-technical users efficiently.

Scottish Water built "SPARK," a natural-language interface on Databricks Genie, integrated with Microsoft Teams via Copilot. This allows project teams to ask questions in plain English and receive trusted answers, grounded in curated and governed data layers, directly within their workflow.

The new system drastically reduces the time and effort required to access project data, turning multi-step searches into direct question-and-answer interactions. It saves project teams significant hours weekly by enabling faster, tailored insights and reducing dependence on specialist support.

The solution was designed with governance from the start, utilizing Unity Catalog for access control and a semantic layer built on metric views to standardize definitions. It was also tuned for accuracy with curated business rules, terminology, and example questions to align with Scottish Water's conventions.

Scottish Water established comprehensive monitoring for the SPARK solution, tracking adoption, answer quality, and performance in production. This includes reviewing recurring questions for potential visualizations and ensuring the solution can be safely and consistently promoted across environments.