AI Is Good Enough. The Edge Is Domain Expertise.

There is a familiar debate in technology right now about whether AI is ready for serious, regulated, professional work. In the built world — civil engineering, permitting, compliance — the honest answer is yes, with a supervisor. The frontier models have crossed the threshold where they can do real work: reading codes, interpreting regulations, drafting applications, analyzing data, and producing engineering-grade explanations of their reasoning.

We say this from use, not from hope. We point these models at actual permitting and civil engineering problems daily. They are not perfect, and they should not be trusted blindly — but neither should a junior engineer, and we have built entire professions around supervising junior work. The model is now a capable junior. That is a profound change, and most of the industry has not internalized it yet.

Here is the part that matters for where value accrues: if everyone can reach the same frontier models, the model is not the moat. Raw capability is becoming a commodity. The durable edge is not who has AI — it is who knows what to do with it.

That edge is domain expertise. Knowing that a construction stormwater permit triggers at one acre of disturbance; that a Section 404 permit applies only when you touch waters of the United States; that a delegated state runs its own program while a non-delegated one routes to the EPA — and knowing it well enough to catch the model when it is confidently wrong. Knowing which permit a project actually needs is worth far more than the ability to generate a thousand fluent words about permitting.

This is why the two obvious groups of players each hold half the equation. The incumbents in AEC technology have distribution, data, and growing moats — but rebuilding around AI threatens the very workflows their businesses are priced on, and large companies rarely cannibalize themselves on time. The AI-native newcomers have the models and the speed — but most have never sat at a permit desk, run a project, or stamped a drawing, and it shows in what they ship.

We are not especially interested in that contest. We are interested in the narrow, valuable space neither side occupies well: people who have the domain expertise and the willingness to build AI-native from the first line of code. That is a small group. We are in it.

Our work in permitting and compliance reflects exactly this. The accuracy that matters does not come from a clever prompt — it comes from an engineer deciding which agency truly has jurisdiction, which form is current, which threshold applies, and encoding that judgment so the AI carries it forward for everyone. The model does the volume. The expertise does the deciding.

This is why we believe it is our time — not because we hold a secret model, because nobody does, but because the moment rewards the exact combination we have spent careers assembling: deep knowledge of civil engineering, construction, and regulation, paired with the ability to build software and sell it. The tools have finally caught up to the people who understand the problem.