Co-engineered a generative-AI analytics platform with DataChat that lets non-technical users interrogate enterprise data warehouses in plain English — and, critically, see exactly what the AI did to get the answer. The system decomposes each conversational question into a transparent, reproducible sequence of analytical steps, runs natively across Snowflake, BigQuery, Databricks, and Redshift, and supports iterative drill-down without ever surfacing SQL to the user.
The Challenge
In most enterprises, business users wait in line behind a small team of analysts who can write SQL, and the analysts spend their days answering ad-hoc questions instead of doing deeper data work. Traditional BI tools narrow the gap with dashboards but still demand technical fluency, while first-generation natural-language query tools have been brittle and opaque — when the answer is wrong, no one can see why. DataChat needed a generative-AI analytics layer that was genuinely usable by non-technical staff and trustworthy enough for enterprise data teams to stake their name on.
Our Approach
We co-engineered the generative-AI core that translates a plain-English question into an executable, multi-step analytical workflow over the customer's data warehouse. Each conversational turn is decomposed into discrete, named operations — load, filter, join, aggregate, visualize — and rendered to the user as a transparent, editable sequence rather than one opaque LLM call. Every result is reproducible and auditable, and follow-up questions refine the existing workflow instead of starting over. The execution layer abstracts SQL dialect and schema differences across Snowflake, Google BigQuery, Databricks, and Amazon Redshift, and the interaction model supports real-time drill-down on results already in context.
Results
DataChat shipped a no-code conversational analytics product that lets business users ask data questions in plain language and get answers their data team can actually defend, because every step the AI took is visible and reproducible. The platform connects natively to the four leading cloud data warehouses, dissolving the analyst bottleneck without sacrificing governance or auditability. The company was subsequently acquired.
