I Built a Tiny AI Tool That Converts Natural Language Into SQL — And My Team Won’t Stop Using It

I didn’t plan to build a tool. I just wanted to stop answering the same three questions.

Manash Pratim

5 min read

·

4 days ago

--

1

Share

Press enter or click to view image in full size

Every other day, someone from the product or marketing team would ping me: “How do I query the new orders table again?” “What’s the column name for transaction status?” “Can you quickly write a SQL script to filter users by signup date?”

At first, I didn’t mind. Helping teammates feels good. It makes you feel useful.

But after answering the exact same questions for the 40th time, I realized something: It wasn’t a people problem. It was a workflow problem.

My teammates aren’t dumb; the database schema is just complex.

So, instead of resigning myself to being the “Human SQL Encyclopedia” forever, I built a tiny AI layer that sits on top of our database and turns English into SQL.

Not a massive autonomous agent. Not a fancy SaaS wrapper. Just a clean, 50-line Python pipeline.

And honestly? It works better than I expected.

The Architecture (In Simple Words)