Key Facts
- What: Scottish Water launched "SPARK," a conversational AI interface for capital investment data.
- Technology: Built on Databricks Genie and integrated with Microsoft Teams via the Model Context Protocol (MCP).
- Governance: Data is managed through Databricks Unity Catalog to ensure consistency and security.
- Impact: Estimated savings of 520 to 1,300 hours per year by reducing manual data retrieval and report navigation.
Scottish Water has deployed a new natural-language interface called SPARK to transform how its teams interact with capital investment data. Powered by Databricks Genie, the tool allows project managers and delivery teams to query complex project portfolios using plain English directly within Microsoft Teams.
The initiative aims to solve a persistent "access problem" where critical information on project status, financial performance, and delivery milestones was trapped within a vast estate of static reports and underlying data tables. By moving from manual data extraction to a conversational model, Scottish Water intends to accelerate decision-making across its Capital Investment (CI) function.
Solving the Data Access Bottleneck
Prior to the implementation of SPARK, Scottish Water faced significant friction in its data workflows. Although the organization possessed ample data, non-technical users often struggled to find specific insights. This led to a cycle of duplicated efforts, where analysts frequently created new reports to answer questions that existing documentation had already addressed.
According to Databricks, this lack of visibility meant that delivery teams often waited for data specialists to manually extract figures from underlying tables. The result was a bottleneck that prevented important project data from reaching decision-makers in a timely manner. Analysts, meanwhile, spent a disproportionate amount of time reproducing work rather than performing high-value modeling or strategy.
The SPARK Architecture: Conversational Intelligence
The solution, branded internally as SPARK, serves as a natural-language bridge to Scottish Water’s project portfolio data. The system is designed to meet users where they already work, integrating the Databricks Genie experience into Microsoft Teams.
The technical workflow involves several layers of orchestration:
- User Interaction: A user submits a natural-language question in Microsoft Teams via Copilot.
- Orchestration: A Copilot supervisor agent manages the request.
- Connectivity: The agent connects to the Databricks Genie Space using the Model Context Protocol (MCP).
- Execution: Genie translates the English query into a technical query, runs it against governed data in the Databricks Unity Catalog, and returns the result to the Teams interface.
This architecture ensures that users do not need to switch between multiple software interfaces to find answers. By using the Unity Catalog and a shared semantic layer, Scottish Water ensures that the answers provided by the AI are grounded in governed data and aligned with established business logic.
Real-World Applications and Queries
The SPARK interface is designed to handle specific, practical business questions that previously required manual dashboard navigation. Examples of queries now handled by the system include:
- "List all open project risks that are expiring in August, including project name, risk description, and owner."
- "What is the current live risk score for project X?"
- "Who is the future contractor for project X?"
Each query returns an immediate response derived from governed metric views. This shift allows teams to spend less time "report hunting" and more time acting on the insights provided.
"SPARK is going to completely change how our portfolio and project teams interact with data," said Allan Mason, Programme and Project Delivery Manager for Business Analytics at Scottish Water. "It moves us from static reports to real-time, intelligent conversations with our information, empowering our people to make quicker, better-informed decisions."
Quantifying the Impact
The transition to conversational analytics has yielded measurable efficiency gains. Scottish Water reports that a typical data lookup, which previously required approximately eight clicks and significant dashboard load time, can now be completed with a single question.
For report-based inquiries, users no longer need to navigate through SharePoint, the Reporting Hub, or various report categories. Scottish Water estimates that if 100 users ask just three questions per week, the organization will handle roughly 300 information requests weekly. With a conservative estimate of saving two to five minutes per request, the system is projected to save between 520 and 1,300 hours annually.
Beyond time savings, the conversational model provides tailored answers rather than broad, static reports designed for a general audience. This reduces the organization’s dependence on specialist support and makes data-driven insights accessible to a wider range of non-technical staff.
Governance and Trust
A critical component of the rollout was ensuring that users could trust the AI-generated answers. Scottish Water integrated governance into the system by default. Because SPARK is grounded in the Databricks Unity Catalog, all responses are consistent and explainable. The use of a semantic layer ensures that the AI uses the same definitions and logic as the company's official financial and project reports.
Databricks Genie, which recently became generally available, is part of a broader trend in the industry toward "AI/BI"—merging traditional business intelligence with generative AI to make data accessible to every corner of a business. Scottish Water's implementation represents a significant use case of this technology within the utility sector.
What’s Next
Scottish Water has launched the SPARK interface to its project teams, focusing on capital investment data. While the company has not yet announced specific timelines for expanding the tool to other departments, the current implementation is designed for scale and continued integration within the Microsoft Teams environment. No further roadmap for additional features was provided in the initial announcement.