The interesting part of an AI feature is not the prompt. It is everything around it: what the model is allowed to see, what happens when it is wrong, and how anyone decides whether it is good enough to leave switched on.
We take that engineering view of it. A feature is scoped to one job - summarise this, classify that, pull these fields out of a document - grounded in the business's own data rather than in a general web index, with a defined path for output that is uncertain or refused, and with acceptance criteria agreed before it goes anywhere near a customer.
If a model is not the right tool for the job, we will say so.