The tool works and the person still won't use it: separate can't from won't
When a team does not adopt a working AI tool, the first split is between people who want to but can't yet (a training problem) and people who simply won't (a different problem). The practical move: stop training the unwilling, find the one person with a task they genuinely hate, make the agent do that one thing well, and let one visible win spread further than ten training sessions.