Back to articles

AI Moats: It's Not About Data Volume, It's About Structure

•2 min read•By Ioannis Zempekakis
AIDataStrategyToqanProduct

A lot of conversations about AI moats start in the wrong place. I often hear people point to data volume as the advantage. In most cases, that's not really a moat.

If you have data about traffic jams from Amsterdam to Rotterdam, everybody can go to Google Maps. That data exists. It's useful. But it's not defensible.

What Actually Matters

What matters much more is how your data is structured. And how well it's connected.

A good moat is not just having data. It's having data that actually tells you something. High-signal data. Data that points to potential value. Not just what already happened.

There's also a difference between data that exists and data that compounds. If the same data can be accessed or recreated by others, it's very hard to build a lasting advantage on top of it.

Beyond the Data

And… data alone is still not enough. What we're seeing is that real defensibility also depends on the system around it.

A tech stack that allows you to:

  • Test effectively
  • Experiment rapidly
  • Catch errors before they reach production

That testing capability is what makes it possible to build new value on top of systems that already work, instead of just trying to replace them or make them cheaper.

The Common Mistake

That's often where companies think their moat is, but it isn't. They assume scale equals advantage.

In practice, structure, connectivity, and the ability to safely iterate matter much more than volume.

And that's what we do with Toqan at Prosus.

The Real Question

So when we talk about AI advantage, the question isn't how much data you have.

It's whether your data and systems together can tell you something others can't.

That's usually where a real moat starts.

Originally published on LinkedIn