The same conversational analytics platform as the Beauty POC was delivered for a post-lease automotive trading client with annual revenue in the tens of millions (USD). Self-service analytics for operations and sales: natural-language questions, SQL-backed answers, and optional auto-selected charts; role-based access for different teams.
Purpose and Goals
- Business: Give non-technical users (operations, sales) direct access to enterprise data without writing reports or waiting on IT; self-service analytics with natural-language questions and instant answers with charts; reduced dependency on IT and shorter time-to-insight.
- Technical: Same end-to-end flow as the Beauty POC: natural-language question → intent routing → SQL generation and execution → formatted answer with optional visualisations; conversation history and optional external context; configurable via the client's environment.
Challenges
Same technical and operational challenges as the Conversational Analytics POC (Beauty): semantic–database alignment, SQL reliability, concurrency and UX (async, timeouts, cancellation), multi-backend support (SQLite/PostgreSQL), pipeline complexity (multi-agent coordination), and deployment (Docker, secrets, networking).
Deliverables
Same architecture and stack as the Beauty POC: LangGraph-based pipeline (routing, clarification, SQL generation/execution, analysis); FastAPI backend with async request manager, Bearer-token auth (Auth0-ready), health checks; Next.js chat application with conversation history, saved and preset questions, internationalisation, theme support; SQLite and PostgreSQL for analytics; Docker Compose and cloud deployment. Tailored for automotive data and roles (operations, sales).
Outcome
Working proof of concept for this client: faster time-to-insight and reduced dependency on IT; foundation for scaling to more users and data sources. Reusable patterns (request queue, multi-agent graph, categorical matching, visualisation pipeline) applied across both clients.
For full technical details, architecture, and stack, see the Conversational Analytics POC (Global Beauty Conglomerate) project.