I learned the word “ontology” a few years ago, from reading about Palantir. What it describes, I have been doing for thirty years.
Palantir was founded in 2003 and initially worked for intelligence agencies and the military. Inside its system Foundry, the ontology is the core: a model of what things exist, how they relate to each other and what properties they carry. Since the company connected its ontology with language models in 2023, Fortune 500 companies now talk about “building their ontology.” The word itself is much older than Palantir’s use of it. Lorhard coined it in 1606, Goclenius in 1613. The discipline behind the word is older still. Aristotle founded it over two thousand years ago as First Philosophy. His question was how what we know hangs together.
That question is the one I have worked on my entire career, long before I had a name for it. Single facts do not explain a market. Knowing who buys what, where and why regulation matters gives you a pile of facts. The understanding starts only when you build the relationships between those facts.
Take an example that has nothing to do with markets, just to show the shape of the problem. We know daylight improves concentration. We know noise reduces it. Neither fact helps you plan an office that is bright and loud. You need the relationship between the two effects: which one dominates, at what noise level, for which kind of work and how that shifts with age, time of day and whether people work alone or in a team. A planner can use that relationship. Neither fact by itself tells them anything they can act on. That is ontology, whether or not anyone calls it that.
I introduced and distributed Power Balance across Europe. I imported health products from Japan. Today I advise American companies entering the European market and build the logistics and distribution structures they need, direct or retail, depending on what makes sense in the specific case. Each time the content changed, while the method underneath stayed the same. In every case, market data and sales and marketing strategy decided the outcome, and in every case there was an enormous amount of data to research by hand, process, understand and communicate across teams. The effort was immense. We were always behind. Nothing ran in real time.
Databases and algorithms improved over the years and the work got faster, but the volume of spreadsheets and data sources remained a specialist’s territory. People running the business day to day rarely touched any of it directly. Every large retail chain ran its own customer systems with its own interfaces and formats. Even small distributors needed serious IT effort just to handle daily operations. Data, content and interfaces dominated everything, and whoever handled them better moved faster. Constant monitoring and controlling was not optional.
Whether anyone truly understood the data underneath all this activity is a question I still cannot answer cleanly. I struggled with it myself. Most people ignored the data and worked from their own read of the situation. The ones who succeeded had the highest hit rate, but luck and chance played a real part. The story of the manager or salesperson with the right instinct holds some truth, but even the best instincts failed sometimes, and the people who owned them were demoted for it, often unfairly.
I had no word for what I was doing with all this. I called it “my system” or “my method”: a kind of ontology, built mostly in Excel, with little elegance. It was the attempt to make visible, in the middle of the chaos, the connections a person could actually base a decision on.
With today’s tools, more people understand what is in the data, clearly more than a decade ago, though still far from everyone. That has changed my own position in the work. I switched sides, from player to coach. Instead of fighting inside the market myself, I now build the systems other people work with. That lets me go deeper and across a wider range than I could before, and it is work I enjoy for the same reason I always have: the pull to keep figuring things out.
Ten years ago, my part of this was still heavy research, spreadsheets and my own feel for the market. Today that volume of information is no longer something a person can process by hand, and most of the work runs through data and AI instead. I build systems for that work, meant to be used and carried to the next client. The structure underneath is still ontology. The work has got better for a reason that has little to do with machine intelligence. Building the ontology forces me to write down, explicitly, the structure that used to live only in my head. Anyone who has to define the relationships between things cleanly finds out fast where they were relying on gut feeling before, and where their own assumptions were wrong.
One question is still open for me. Agentic AI, the term for a system that breaks a task into parts, works out what depends on what and runs the steps in the right order, sounds to me like ontology with a new label. I have not worked through that comparison carefully enough yet to say it with confidence. The next essay takes up that question directly: why AI without structure does not get far, and what ontology changes about that.