Table Lens

An AI-native conversational BI tool. Ask questions about your data in plain English — an agent discovers your schema, understands what it means, and answers by querying it directly.

The database behind this instance is a property & casualty insurance book — customers, policies, agents, claims, underwriting, fraud flags, call-center interactions, and audit trails — modeling how a mid-size insurer's data actually looks spread across departments, not one tidy spreadsheet.

It's messy, the way real operational data is: 50 tables, 2,215 columns, and over 5 million rows, ranging from small lookup tables to a 500,000-row audit log. Column types and null rates vary wildly table to table, denormalized monthly snapshots sit next to raw transactional logs, timestamps are stored without timezone, some tables run 15+ columns wide, and a few ambiguous table names as well. This is the kind of database Table Lens actually has to reason about — not one built to make discovery look easy.