The defining constraint: trust requires verifiability

In consumer settings, a plausible answer is often good enough. In professional settings, an answer the expert cannot verify is worse than no answer, because acting on an unverifiable claim is a liability. So the organizing principle for domain agents is: every consequential claim must be traceable to a source the expert can check. This single requirement drives most of the design differences below.