- Experiment
- LAB 002
- Status
- Prototype
- Tags
- AI · Data · Analytics
- Studio
- Man & Tech Lab
01
The question
Can a business ask its own database a question in plain language and trust the answer?
02
Concept
A semantic layer between the language model and the warehouse, so questions resolve against defined metrics instead of raw guesses.
03
Technologies
- SQL
- PostgreSQL
- LLM APIs
- React
04
What we're learning
- A metric layer removes most hallucinated numbers
- Showing the generated query builds more trust than hiding it
- Ambiguous questions need a clarifying step, not a best guess
05
Future direction
Automatic narrative summaries for weekly performance, with anomaly detection that explains itself.
06
Screenshots

Curious about this experiment?
Tell us what you're trying to build and we'll tell you how we'd approach it.
info@manandtech.com