Insight
Why a specialist should only do one job
The case against the do-everything assistant, and what narrowness buys you.
The instinct when automating a small business is to reach for one thing that does all of it. One assistant, every task, a single place to ask. It demos well and it fails quietly, because a system that does everything cannot be checked at anything.
Narrowness is not a limitation here, it is the mechanism. A specialist that only books appointments has a definition of correct: the right slot, in the right calendar, confirmed with the right person. That definition can be written down, tested against real work, and measured after the fact. When it goes wrong, there is exactly one place to look.
Widen the same component to also handle invoicing and marketing, and the definition of correct dissolves. There is no single test. Failures become anecdotes rather than defects, and the only available response is to try rewording the instructions and hope.
The trade is real: several narrow specialists take more work to build than one broad one. What you get for it is a system whose behavior you can predict, and which can be corrected rather than merely adjusted.
