It's like how bank accounts are always used as an example when describing transactional systems, despite not working that way in reality (two near-simultaneous transactions can result in you having a negative balance, and this is a feature not a bug). Naive assumptions about how the world works can certainly produce software that is clear, logical and wrong.

However, what I hear there is "oh sure, you might be able to write software for all of these other things, but not this thing that I'm an expert in, that's way too complicated". Many people have been wrong about this in other domains, and it seems to me that essential complexity can only delay the point at which software can usefully automate things, not prevent it entirely.