TextQL Built a Rosetta Stone for Enterprise Data — and It Actually Works
Most natural language query tools fail when you move beyond your laptop. TextQL built theirs to work with data that doesn't fit on a single computer.
The company spent three years teaching AI how to work with tables at enterprise scale. Not the tables you find in a Google Doc or Excel file. The ones with hundreds of thousands of tables, trillions of rows, and petabytes of data spread across different systems.
Ethan Ding, TextQL's CEO, explained their approach during the 64th IT Press Tour. "First generation AI products were designed for documents that fit on your computer," he said. "Try to put more than 10 gigabytes of data in an Excel file and your computer starts to fall apart."
The Real Problem
Enterprise data lives in data warehouses, ERP systems, CRM platforms, and ITSM tools. Each platform has its own query language, table format, and way of storing data. Moving data between these systems costs millions of dollars and takes months.
The expense isn't accidental. These platforms deliberately make it expensive to leave. They charge 10 to 20 times markup on the underlying infrastructure costs. A CFO who wants to move business logic from SAP to NetSuite faces a contract with Accenture that runs $50 million over five years — and only moves 10% of the actual business logic.
"If your landlord knew it cost you $50 million to move apartments, they could charge whatever they wanted," Ding said. "That's why these platforms have really high lock-in."
The Technical Challenge
TextQL built what Ding calls a "Rosetta Stone system." Instead of translating between every possible combination of query languages and table formats (an n-by-n complexity problem), they created an intermediary layer that makes it a one-by-n problem.
The system can connect to any database and translate queries on the fly. It doesn't require data migration or lengthy setup. You connect your database, ask a question in natural language, and the AI agent writes code, checks results, and tries different approaches when queries fail.
During the demo, Ding showed how the system handles a complex query about flight routes and passenger data. The AI wrote SQL queries, caught execution errors, rewrote the queries, and eventually produced visualizations on a satellite map. All without manual intervention.
"The AI is problem-solving the same way a human data scientist would," Ding explained. "You try something, it fails, you keep working through it."
What Makes It Different
Most natural language query tools work fine with documents and small datasets. TextQL focused on the hard stuff. They can handle petabytes of data across multiple systems with different schemas. The system reconciles when a customer appears as "Fred" in one database, "Fred Frankel" in another, and "Emails are not allowed" in a third.
The company uses Claude and GPT models but runs their own infrastructure layer. "We can spin up a really big computer on the fly," Ding said. "Historically, you had to do this work where the database lives. We can do it on the fly, which means you only pay for minutes in transit."
Their pricing model reflects this approach. They charge based on compute usage, not on tokens or seats. Customers commit to an amount and burn it down as they use the system.
The Hard Road
TextQL started in December 2022. They rewrote their entire codebase seven times. The first version didn't work. Neither did the second. Or the third.
"We lost every customer until January of this year," Ding admitted. "All the code is new as of January. We haven't lost a single customer or pilot since then."
The 15-person team is mostly senior engineers — dropouts from high-frequency trading firms, former researchers, and people with Master's and PhDs. The CTO was the youngest senior engineer ever at Facebook, working on a team led by Andrew Bosworth (the person who got the famous billion-dollar offer from Facebook).
Current State
TextQL raised $5 million and projects $7 to 7.5 million in revenue. They started serious customer conversations in June and closed their first major contracts. Sales cycles in this space typically run 12 months. TextQL shortened theirs to six.
Their target customers are CFO offices, financial services firms, and healthcare organizations. The team recently hired experienced enterprise sales executives who've sold to Fortune 500 companies for decades.
"Put us in a room next to competing software," Ding challenged. "Put us in a one-to-one comparison. I will bet on our system every time."
The company's approach works because they focused on the hardest problems first. When data is spread across multiple systems, when queries need to run against trillions of rows, when the cost of vendor lock-in reaches millions of dollars — that's where TextQL operates.
Most companies promise natural language querying. TextQL built a system that can actually deliver it at enterprise scale. That's a meaningful difference.