The line inverts the usual sequence. Most people wait for conviction before starting. The deck needs to feel right. The process needs to make sense. The new tool needs to prove itself before you commit the hour. Mollick is naming what actually changes behavior.

AI adoption follows this pattern exactly. The person who refuses to try ChatGPT until they understand transformers will still be refusing in two years. The person who opened it once, fed it a bad prompt, got a mediocre answer, and concluded "this isn't useful" has locked in that conclusion. The belief hardens before the evidence accumulates.

The only way through is repeated, low-stakes contact. One question a day for two weeks. Not because you believe it will work — because repetition rewrites the prior. By day ten the reflex changes. The thing that felt effortful becomes automatic. You stop asking whether to use it and start asking how.

Run the same sequence on any avoided behavior. You cannot convince yourself to enjoy the gym. You can go three times, hate it twice, tolerate it once, and return next week. The thinking follows the doing. It has never worked the other way.

So today: one hour with the tool you have been postponing. Not to master it. Not to prove anything. To log contact number one. Because conviction is not the price of entry. Conviction is what you earn after fifty mediocre sessions.