AI boosts demand for data plumbing
Charles Chretien, co-founder of Prequel, on the modern data stack’s ROI problem
The real bottleneck in text-to-SQL is not writing SQL, it is knowing what the data actually means. In practice, business datasets are full of company specific definitions, edge cases, and joins that only make sense if someone knows how the tables were modeled and what question the business is really trying to answer. That is why Prequel is more focused on moving clean data into products and warehouses than on replacing the analyst layer with a chat box.
-
Prequel’s broader thesis is that AI increases demand for core data plumbing, because companies need customer data from warehouses to power product experiences. That makes reliable import and export infrastructure easier to monetize than a natural language layer that still struggles with schema nuance.
-
This fits the larger ROI reset in the modern data stack. Horizontal tools that process lots of data were cut when budgets tightened, while more vertical products like CDPs held up better because they could tie data work to concrete outcomes like ad efficiency and conversion lift.
-
There are already products pushing text-to-SQL into the market, including MotherDuck with SQL debugging and text-to-SQL features. But that feature works best when the data model is simple and local, not when an enterprise team needs answers grounded in custom business logic across many tools.
The likely next step is not fully autonomous analytics, but narrower AI workflows built on better structured data and more opinionated products. The winners will be the platforms that first make data definitions, movement, and activation reliable, then layer AI on top where the workflow is constrained enough for the model to be right consistently.