Why Every Trade Needs Its Own AI, Not a Generic Chatbot
Why Every Trade Needs Its Own AI, Not a Generic Chatbot
A frontier model can pass professional exams. So why build a narrow assistant for a single trade?
Because passing the exam and being useful on the job are different problems.
The four gaps
The citation gap. A generalist will answer a code question fluently and may well be right. But it cannot reliably tell you which article backs the answer, and it will produce plausible-looking references that do not exist. In regulated work, an uncheckable answer is not usable.
The interface gap. Professional questions have structure. Sizing a duct means knowing CFM, velocity limit, and duct type. A chat box makes the user supply all of that in prose, every time. A purpose-built tool asks three questions and computes.
The context gap. A trade assistant knows that "the split" means temperature split, and that a number near 400 CFM per ton is meaningful. Generalists ask clarifying questions about things a tradesperson considers obvious, which reads as incompetence.
The consequence gap. This is the deepest one. A generalist is tuned to be helpful and will answer almost anything. A professional tool must be tuned to be careful, and to decline when the honest answer is "go measure this."
What narrow actually buys you
Being narrow is not just a marketing position. It changes engineering decisions:
- You can curate the corpus completely rather than hoping the training data covered it
- You can evaluate exhaustively because the question space is bounded
- You can build real tools, not just chat — calculators that are exactly right, offline
- You can tune refusal behaviour to the actual risk profile of the work
None of that is available to a general assistant, however capable the underlying model.
The counter-argument, honestly
The strongest case against vertical tools is that frontier models keep absorbing capability. Why build a moat that gets flooded?
Our answer: the moat is not model capability. It is the corpus, the evaluation discipline, the calculators, and the interface. A better base model makes our products better — it does not make them unnecessary, any more than a faster processor eliminated the need for specialised software.
The trades will not be served by one assistant that is adequate at every trade. They will be served by tools that know one trade completely and are honest about the edges.
That is the thesis behind our field products — Ampora for electricians, Ventora for HVAC.