Scottish Water has turned one of its most common data headaches into a conversational service, allowing project teams to ask questions about its Capital Investment programme in plain English and get answers in Microsoft Teams within seconds.
The utility’s challenge was not a lack of information. Its project status, financial performance, milestones and risks were already captured across a wide range of reports and tables. The problem, according to a Databricks case study, was...
Continue Reading This Article
Enjoy this article as well as all of our content, including reports, news, tips and more.
By registering or signing into your SRM Today account, you agree to SRM Today's Terms of Use and consent to the processing of your personal information as described in our Privacy Policy.
finding the right figures quickly enough for them to be useful. Teams often had to search through multiple reports, while analysts were repeatedly asked to extract information that already existed elsewhere.
To tackle that, Scottish Water built SPARK, its internal name for a natural-language interface to project portfolio data powered by Databricks Genie. Databricks says Genie is designed to let users ask questions in natural language while grounding responses in governed organisational data through Unity Catalog. In Scottish Water’s case, the tool was embedded in Microsoft Teams, so users did not need to leave the environment where they were already working.
The workflow is deliberately structured. A question is entered in Teams via Copilot, passed to a supervisor agent, and then routed through the Model Context Protocol to a Databricks Genie Space. Genie converts the request into a query, runs it against governed data, and returns the result to the user. Databricks says this approach keeps the interaction simple for business users while preserving control over the underlying data.
The questions being asked are practical and specific: which project risks are expiring in August, what the live risk score is for a particular project, which risk has the highest exposure, or who the future contractor will be. Rather than searching through SharePoint, reporting hubs and individual dashboards, teams can now ask directly and receive a tailored response.
That shift has also reduced friction. Scottish Water says a lookup that once took several clicks plus dashboard load time can now begin with a single question in Teams. The company estimates that replacing repetitive report hunting with conversational access could save many hours a week across a broad user base, while also reducing dependence on specialist support.
Trust was central to the design. Databricks’ Genie documentation says answers should be grounded in governed data and shared definitions, and Scottish Water adopted that principle by curating the relevant information into the gold layer before exposing it to users. It also built a semantic layer based on metric views so that measures, dimensions and business terms are used consistently.
The system was further tuned to Scottish Water’s own language and processes. Curated instructions were added to reflect internal rules such as fiscal period definitions, milestone sequencing and handling of project identifiers. The team also used example questions and SQL mappings to train the system on how staff actually ask for information, and created a benchmark suite to test accuracy whenever the space or source data changes.
Scottish Water has also put monitoring in place to track how the service performs in production. Usage, answer quality, conversation length, query volume, slow-running queries and cost per user are all being watched, while recurring questions are being reviewed to see whether some should be turned into visualisations or standard reports.
The deployment model was built for repeatability as well as trust. According to the Databricks blog, the solution is packaged with Databricks Asset Bundles, separated into development, test and production environments, and deployed through Azure DevOps with approval gates. Access and credentials are handled through Microsoft Entra ID and Azure Key Vault.
Allan Mason, Programme and Project Delivery Manager for Business Analytics, said SPARK would change how portfolio and project teams interact with data, moving them from static reports to real-time conversations with information and enabling faster, better-informed decisions.
For Scottish Water, the broader value is clear: less time spent hunting for reports, less reliance on specialists, and more governed insight available directly in the flow of work. Databricks presents SPARK as an example of how conversational analytics can be made reliable when the data, rules and controls are designed together rather than bolted on afterwards.
Source: Noah Wire Services
