Ontologies: The Keystone of AI

By Rob Buller, Founder and Managing Partner, Cyberhill Partners, Forbes Tech Council Contributor

Every week, I speak with three or four new companies about their AI programs. Some are global manufacturers. Others are healthcare networks with tens of billions in revenue. Most are already a few years into their AI journey.

And when I ask them what kind of ontology they’re using, the response is almost always the same: “What’s an ontology?”

Let’s talk about it.

At its root, ontology is a philosophical concept—a state of being. But in the world of AI, it’s something more practical and foundational: a domain-specific structured framework that informs a computer how concepts relate to one another.

Think of it as a way for human knowledge to be translated into machine understanding.

In finance, a “stock” means one thing. In retail, it means something else. In agriculture, it might mean livestock. Ontologies make these distinctions explicit. They help models understand not just words, but relationships.

Here’s a simple example from the airline industry: United Airlines flies from San Francisco to Chicago. That’s a flight. But behind that are dozens of relationships—airports, airlines, airplane types, passengers, luggage, ticket types, aircraft age, etc. Ontologies map these connections into a living knowledge graph that machines can interpret.

This is where true insight begins.

Relational Databases Aren’t Enough

Many companies today are running large language models on top of traditional relational databases. That can produce some results—it’s analytics. But it’s limited.

Without an ontology, there’s no real memory. No context. No semantic reasoning. The system doesn’t know that a jaguar can be an animal or a car. It doesn’t know a person has one birthday and many flights—not the other way around.

Why Ontologies Matter

I believe ontologies are the keystone of enterprise AI. Just like in a Roman arch, the keystone is what holds the structure together. If you pull it out, everything else falls.

When built properly at the beginning, a good ontology does the following:

• Enables traceability: Without an ontology, you get zero traceability in your AI decisions. With one, you can follow decisions all the way back to the source data. That’s essential in industries like finance, defense, manufacturing, and healthcare.
• Provides structure for growth: Ontologies evolve. You start with a trunk, then build out branches over time. As your business changes, so does your model of the world—and your AI stays aligned.
• Supports future-state AI: We’re moving from generative AI to agentic AI—systems that reason, plan, and act. These systems demand more context and precision. Ontologies are what make that possible.
• Becomes intellectual property: Over time, your ontology reflects the unique language, workflows, and logic of your organization. That becomes a strategic asset—IP you own, defend, and build on.
• Delivers competitive advantage: Better insights. Smarter models. More reusable knowledge. A mature ontology gives you all of that—and more.

Don’t Wait

The best time to build an ontology is at the start of your AI journey. The second-best time? Now.

There are excellent tools available, and in many cases, you can build a baseline ontology in a day. The power is in how you refine and expand it over time—especially when combined with industry ontologies like SNOMED in healthcare.

The companies that treat ontology as foundational, not optional, are the ones that will win with AI.

If your team is building AI, but hasn’t tackled ontologies yet, I’d be happy to talk. We’ve helped organizations across industries stand up scalable, explainable, and future-ready frameworks that actually evolve with their data.

Schedule a time to discuss here.

🎥 Watch the full segment of Rob Buller’s talk on ontologies from the TechEx Big Data & AI conference below.