Skill is built in a domain, not in a person

Competence is local, and the evidence that it does not transfer is everywhere once you look.

The everyday model of expertise treats it as a personal attribute, like height. Some people are smart and knowledgeable, others less so, and the smart ones know things generally. Almost everything we understand about how skill forms says otherwise. Expertise is built inside a particular domain, through repeated engagement with that domain's specific problems, and it consists largely of pattern recognition tuned to that domain's regularities. The tuning is the whole asset. Move the person to a different set of regularities and the tuning is not just useless, it can actively mislead, because the patterns that fire will be the wrong ones confidently applied.

This is why the phenomenon of the outstanding practitioner making elementary errors in an adjacent field is so common that it barely registers as news. It is not hypocrisy or laziness. It is the predictable result of a system that was trained on one distribution being run on another. The person has not become stupid. They have simply left the region where their instincts are calibrated, while keeping the feeling of fluency that came with being calibrated, which is the dangerous part.

The practical consequence is that the relevant question is never whether somebody is an expert. It is whether they are an expert in the specific thing you are asking about, defined narrowly. A person with deep knowledge of one process may know almost nothing useful about the process next to it, even if the two share a name, a professional body and a building. When you can, push the scope of the question down until it matches the scope of the demonstrated skill. Where those two do not overlap, you are not consulting expertise. You are consulting an opinion held by someone who is used to being right elsewhere.