When not to use it at all
When checking costs more than doing, when the writing was the thinking, when somebody is owed a person, and when you cannot judge the answer.
A tool this general invites you to use it for everything, and the cases where it costs more than it gives are specific and predictable enough to list. Knowing them is worth more than another prompt technique.
When checking costs more than doing
The first and most common case. If verifying the output would take as long as producing it, you have not saved anything, you have added a step and a risk. This is why asking for a short factual answer you then have to look up anyway is a worse use of a minute than looking it up first.
When the point was the thinking
Some documents exist to produce the document. Others exist because writing them forces you to work something out: your pricing, your positioning, the decision about whether to take a client. Handing those over gets you the artefact and skips the part that was the value. You end up with a plan you have not thought about and will not follow.
The test is whether you would have changed your mind while writing it. If yes, write it.
When somebody is owed a person
- A complaint, especially one where you were at fault. A generated apology reads as exactly what it is.
- A refusal. Saying no to a customer, a supplier or an applicant is a sentence somebody should have taken responsibility for writing.
- Anything about somebody's employment.
- Condolences, congratulations, anything personal. The effort was the message.
When you cannot judge the output
Using it in a field you know nothing about feels like the biggest win available and is the most dangerous thing on this list. A confident explanation of a subject you cannot evaluate is indistinguishable from a correct one. If the output concerns law, tax, medicine, safety, or anything where being wrong is expensive and you would not spot it, the tool is for preparing questions for a professional, not for replacing one.
When the volume is the problem
If the plan involves the word more, stop. More articles, more posts, more emails, more pages. Production was never the constraint on any of these; attention was. Doubling output on a channel that was not working doubles the ignoring, and on search specifically, Google's spam policies name mass-produced pages made to rank as a violation in their own right.
There is a plainer version of this. If the only reason a thing is being made is that making it became cheap, nobody asked for it.
When it is quietly deskilling you
The hardest case to notice, because nothing goes wrong. You hand over a kind of work you used to do, it comes back good enough, and over a year you lose the ability to judge whether it is good. Now you cannot evaluate the output and cannot do the job yourself either.
This matters most for the skills your business is actually sold on. If customers buy your judgement about something, that judgement is the asset, and it is maintained by exercising it. Handing over the first draft of everything you are supposed to be expert in is the one efficiency that can cost you the business it was meant to help.
The practical defence is unglamorous: keep doing a portion of it by hand. Not all of it, and not for nostalgia, but enough that you can still tell a good version from a passable one. The test is whether you could correct the output if it were subtly wrong, and that is a skill which decays silently.
Work to do
- Name the one skill your business is actually sold on, and check how much of it you have handed over in the last year.
- Write down the three kinds of message in your business that will always be written by a person, and tell anyone who writes on your behalf.
- Find one place you are using it in a field you cannot check, and change the job from answering to preparing questions.
- Look at anything you are producing more of than last year, and ask what evidence there is that anyone wanted the first lot.
Sources