2026-09-05

How to query a MySQL or PostgreSQL database in natural language

Traditional BI tools expect you to know the schema, join keys, and SQL dialect. Most business questions are simpler: sales this quarter by store, orders stuck in pending, average ticket versus last year.

Text-to-SQL (natural-language querying) turns that sentence into a safe SELECT against your live database. You stay on MySQL, PostgreSQL, Oracle, or SQL Server. Nothing has to move into a cube first.

What you need before you ask

1. A read-only database user (never an admin account in a BI tool).
2. Network access from the BI host to the database (VPN, IP allowlist, or SSH tunnel).
3. A short glossary: what “revenue”, “active customer”, or “store” means in your tables.

Without that glossary, even a strong model will invent joins. With it, the model maps “sales” to the right fact table.

How BI Jedi does it

BI Jedi is a SaaS Business Intelligence layer: you connect the database, open a chat, and ask in natural language. You can also use a visual Builder, templates, or SQL Pro when you want full control. Answers become charts, PDF reports, or live widgets. Your data is not used to train AI models.

If you do not have a single database, the built-in datawarehouse can pull CSV/FTP or other sources on a schedule and keep a query-ready copy.

A better first question than “show me everything”

Start with a bounded question and a time window:

Show current-quarter sales by store, including order count and average ticket.

Then filter (“only Italy”), compare (“versus last quarter”), or pin the result on a dashboard.

That is Business Intelligence with AI without replacing your warehouse or forcing a Power BI migration. Connect a database and try a free workspace on BI Jedi.