◇ Could this help me?
Yes, and the two mental models are worth keeping permanently. The doorman fallacy is the single most useful correction available to anyone estimating what a tool can replace, because the mistake it names is structural rather than careless: the visible task is the one that gets automated and the invisible ones are discovered afterwards, and that happens repeatedly regardless of the technology. Gell-Mann amnesia applied to AI predictions is equally sharp, since it explains why confidence in automation is always highest about other people's work. General practice, with the honest note that the supporting specifics, the retired inventory tool, the executive quotes, and the reading of public offering incentives, are all secondhand and worth verifying before being repeated as fact. The frameworks stand on their own either way.
I want to pressure test an assumption that a task of mine can be automated, using the doorman fallacy as the lens. Here is the task, described below as I would describe it to someone else. Now help me find everything the description leaves out: what does doing this well actually involve that is not the visible main action, what happens around it that I have stopped noticing, and who else depends on something I do incidentally while doing it. Then tell me honestly which parts are genuinely automatable, which parts would fail quietly if automated, and what evidence I would see first if the automation were going wrong.