You don't need to be an expert in SQL to talk to your data

Discover Inpla's Text-to-SQL service, which converts natural language questions into instant queries using artificial intelligence. Inspired by Google and Stanford's Chase SQL, we optimize accuracy and agility to drive your business forward.

Picture this: you have an urgent question about your company's data. You need that information NOW, not in eight hours. But today, if you don't know SQL, you depend on the data team to translate your query and deliver the answer. Waiting doesn't just slow down decisions, it can also slow down your business.

That's why, at Inpla, we set out to solve this problem with a service that converts natural language questions to SQL queries in seconds. Inspired by Chase SQL, developed by experts at Google and Stanford, we decided to take it to the next level.

Why SQL?

Because according to a study, 70% of large companies in Latin America use SAP HANA, that is, they use databases that need SQL to extract information.

The Chase SQL path and how we

improve it

Chase SQL proposes three ways to transform questions in SQL:

1. Divide and conquer: Break the query into small parts and then join them into a final query.
2. Step-by-step thinking: Similar to human reasoning, tables, columns, and filters are identified, slowly advancing until the final query is assembled.
3. Synthetic examples: Similar examples are generated that help the AI understand how to create the final query.

Although Chase SQL is powerful, its trust rate was around 70%. At Inpla we incorporate an additional mechanism: we automatically evaluate the confidence of each path and correct errors as we go. This increased accuracy by up to 90%, making the system much more robust.

Humanizing AI

The key to achieving this was simple: we think like humans. If you were a data scientist, what would you do to solve a complex query? You'd probably break it down into easier chunks, check step by step what you're doing, and use previous examples to make sure you're on track. We put that logic within Inpla.

Here is a video we made that explains this service: https://youtu.be/1ssuDCNC9o0

What's next for Inpla?

We are working to continue improving our service, even aiming to compete directly with models such as Google's in international benchmarks. The idea is simple: that no one else has to wait hours for information that can be obtained in minutes.

At Inpla we don't do magic, we just make the AI really help.