A tool can work in a demonstration and still struggle in normal use. Access, workload, checking and the way information is handed over all affect whether people can use it consistently.
The software is often fine. What stops it is the extra step it adds for the person already holding a phone in one hand and a clipboard in the other.
The tools that stick remove something those people hated: chasing a sign-off, retyping the same quote, hunting for the right version of a document. The ones that die ask the busiest person on the job to do something new for someone else's benefit.
So the first question in the room shouldn't be "which AI should we buy." It should be "whose day does this have to make easier, and have we asked them?" Get that wrong and the technology doesn't matter. The tool gets set aside, and the conclusion becomes "AI didn't work for us."
That's not a technology problem. It's an adoption problem. It's also the part most strategy decks skip, which is why we start with the people.