indigoChart Menu

Highlights

The client, a prominent non-profit association in the architecture and construction sector, supports thousands of members across education, certification, and community initiatives. As their data estate expanded within a centralized Azure-based Centralized Data Platform, internal teams found it increasingly difficult to retrieve relevant business insights without technical assistance.

The objective of this engagement was to develop a conversational AI solution that would allow users—across accounting, membership, and education teams—to ask business questions in plain English and receive real-time answers. The aim was to enhance decision-making and reduce dependency on technical resources by enabling Natural Language to SQL (NL2SQL) capabilities within Microsoft Teams.

Challenges

• Contextual Complexity of Metadata: With domain-specific semantics such as membership tiers, event types, and revenue products, understanding query intent required more than simple keyword mapping.

• User Friction with using SQL: Business users lacked proficiency in SQL and data querying tools, which slowed access to information.

• Integration Constraints: The client preferred Microsoft Teams as the central interaction point, requiring seamless integration without disrupting workflows.

  
  

Solution: Key components and Functional coverage

indigoChart delivered a fully integrated AI-powered Databot designed to process natural language inputs, convert them into valid SQL, and return contextual data responses—all within Microsoft Teams.

Key components :

• Natural Language Query Engine : Users could ask questions such as:

  “How many new members joined last month?”

  “What’s the revenue from Women’s Leadership Summit 2024?”

  “Which members haven’t renewed their membership yet?”

  
  

  These were converted into accurate SQL queries using prompt-engineered NLP models running on Azure OpenAI, with context reinforced by a structured metadata layer.

  
  

• Metadata Enrichment and Knowledge Modeling: To support meaningful interpretation, indigoChart created a curated metadata repository from the underlying data sources. This included semantic groupings, business-friendly labels, temporal logic (e.g., "YTD", "last quarter"), and calculated fields for key performance indicators.

  
  

• Pre-built and Dynamic Query Templates: Common questions were mapped to reusable query structures that dynamically adapted to new parameters (e.g., date ranges, entity names). Advanced logic allowed for joins across multiple data views, such as revenue, engagement, and education records.

  
  

• Teams integration and Response logging: The bot was deployed in Microsoft Teams as a native app, allowing end-users to chat directly and receive tabular responses in near real-time. All conversations—including original question, interpreted SQL, and response—were logged for transparency, feedback, and model retraining.

Functional Coverage:

The AI bot successfully supported diverse use cases across business units:

• Membership Team:

  • Track member engagement levels.

  • Identify top new joiners or those at risk of lapsing.

  
  

• Accounting Team:

  • Get breakdowns of revenue by source (dues, events, courses).

  • Analyze trends in dues collection over defined periods.

  
  

• Continuing Education Services Team:

  • Retrieve top-performing courses by provider.

  • Measure course-level attendance and revenue.

Additionally, the bot supported ad-hoc and exploratory queries, with the ability to handle complex logic like subqueries, aggregates, date filters, and business-defined flags.

Outcomes:

The Proof of Value was delivered over a four-week sprint and exceeded all success criteria set by the client:

• >90% NLP Query Accuracy: Most test queries were interpreted correctly, including those with nuanced intent or incomplete phrasing.

• Response time under 10 seconds: Real-time interactions with backend data maintained strong performance.

  
  

• Cross-team enablement: Business users across roles were empowered to explore and access insights independently, reducing reliance on data analysts and IT support.

  
  

• Secure and Auditable: All queries and responses were logged, providing transparency and traceability—a critical requirement in regulated environments.

Looking Ahead:

Based on the success of the initial deployment, the client is now exploring:

• Expanded Data Coverage: Extending the bot’s reach to other structured and semi-structured sources beyond the Centralized Data.

   

• Continuous Learning Loop: Using thumbs-up/down feedback to continuously fine-tune prompts, language models, and user experience.

  
  

indigoChart continues to support the client as a trusted AI innovation partner, helping scale their vision of accessible, intelligent data interaction across the enterprise.