TopN AnalyticsEnabling Natural-Language Access to a Fleet Company's Full Data Warehouse
A leading pan-India fleet management company operating a multi-domain data warehouse spanning driver lifecycle, financial settlement, vendor payouts, and recruitment.
Where things stood
Getting an answer from the fleet data warehouse required writing SQL directly, or waiting on the data team to write and run a query — a bottleneck for the many non-technical, business-side users (city operations managers, finance, vendor management, recruitment) who needed fast answers to everyday operational questions.
As the warehouse grew to span dozens of interconnected tables across driver lifecycle, settlements, vendor payouts, and recruitment funnels, the schema became too complex for most business users to navigate directly, and every ad-hoc question routed through the data team as a queue.
What we built
The same PostgreSQL analytics warehouse underpinning this portfolio's other case studies sits beneath a WREN AI semantic layer — schema and entity-relationship mapping, plus a reference-SQL library — powering a natural-language query interface used company-wide.
How it works
TopN Analytics built and deployed a custom implementation of WREN AI — a semantic layer and natural-language-to-SQL engine — across the company's entire data warehouse, mapping schema and entity relationships across the warehouse's interconnected domains, curating a library of reference SQL to ground query generation, and testing against real business use-cases before rolling out company-wide.
Business users can now ask questions in plain language and get back an answer grounded in the actual warehouse schema, without needing to know the underlying table structure or write SQL themselves.
What we delivered and learned
The implementation spans the full data warehouse rather than a single domain, meaning it had to reconcile schema relationships across driver lifecycle, financial settlement, vendor payout, and recruitment data models simultaneously — it is live in production and used company-wide, not a pilot or proof of concept.
What it's built on
Want the full picture?
We're happy to walk through the details, numbers, and trade-offs directly.