I’ve been building with the cutting edge AI tools and building AI into products for over two years now. Learning, reading, and building, repeat. I’ve also been a product manager working with Fortune 500 clients for the past three and a half years. Through both experiences the similarity I’ve seen is that the hardest part of the process is understanding the intent and turning that into working software. The difference now is that software has become exponentially easier to build. What remains is figuring out what to build.

Part of that equation, especially in B2B context, is understanding what I call the ontology. Ontology is the mapping of relationships, entities, processes, rules, edge cases, and implicit knowledge within an organization — the stuff you can only learn by sitting with someone, watching them work, and asking “why do you do it that way?” Bridging the gap between this knowledge and the code is the key to any applied AI product.

I learned this building a custom platform for a construction subcontractor. The AI already works — I can build agents that extract scope from 400-page specs, source datasheets from the web, draft submittal packages. The models are capable. But the AI doesn’t know the business. A response can pass every schema validation and still be useless because the model didn’t know that this subcontractor never bids on Division 10 work, or that “approved equal” means something different in public vs. private projects. To solve this the answer is to fix context, not code. That mindset shift has been one of the biggest unlocks in how I build. And it’s why the most valuable thing I’ve built isn’t the platform — it’s the ontology artifact. The code will change. The models will get better. But the codification of how the business works will endure.

The idea behind this product is that agents already do a great job at synthesizing and finding connections. What if there was an agent that could do this for the ontology? What if you could do the first discovery session with an agent that would get you up and running — not only doing the first pass, but identifying the highest ROI opportunities for applying AI to solve whichever problem it identifies? This is not meant to replace the human interaction because there is much implicit knowledge that can’t be extracted merely through what’s digitally available; but this would be a first pass that structures the remaining 20% so the follow-up conversations are specific, contextual, and immediately actionable.

Right now I see the start of this being a sort of free consultation for what an applied AI consultancy would do in their first meeting. In the future this could work by continuously running for in-flight projects — keeping the ontology artifact current as new edge cases surface and business rules evolve. The underlying principle is the same for both: the fastest cycle from conversation to structured understanding to working software that delivers value wins.

The people that will win in the post-AI world are those that can go super deep on a topic and stay agile enough to respond to their customers’ needs as the customer needs it. AI enables the “niche-ification” of business — the barriers to building software for many historically underserved businesses, like a 50-person subcontractor construction firm, used to be too high. Market too small, domain too specialized, development too expensive. That’s changing. And the ones who’ve encoded the deepest understanding of their customer’s domain will build the winning products. It’s shortening the cycle between feedback and delivery of new value. That’s what I’m building and that’s what I’m building towards.