Knowing what matters
There is an enormous emphasis on technology and I use it like anybody else. At the same time, I notice what AI quietly redefines and underlines: the value of talking, and of people.
Everyone wants to automate knowledge graph construction. Feed it documents, get an ontology out. But here’s what I keep running into: the hardest part of knowledge graph design isn’t extraction. It’s knowing what matters. It’s the wisdom of people — insights that only domain experts carry.
No LLM will tell you that “supplier” and “vendor” mean different things in your company. Or that the relationship your legal team calls “controls” is not the same one your finance team calls “controls.” Or that a concept buried in one paragraph of a 200-page contract is actually the load-bearing idea the whole domain rests on.
That knowledge lives in people’s heads. It surfaces in conversation, usually over a coffee, sometimes over frustration. There is passion. Innovators with visions, creative minds. Failure and adventure.
I’ve learned more about what an ontology should look like from a 30-minute call with a domain expert than from days of iterating on automated extraction pipelines.
AI is a powerful lens, but a lens needs someone to point it at the right thing.