Start with the purpose
Explain the specific tasks being explored and what is not changing. Avoid telling employees that everyone must become an expert immediately.
Use low-risk practice
Begin with an invented email or public document. Let employees compare the original and revised work, ask questions and reject poor output.
Keep human judgment visible
Demonstrate that AI can be wrong, that employees remain accountable and that asking for help is expected.
What managers should not do
Do not surprise employees with hidden expectations, frame reluctance as failure, use job-loss language to force adoption or measure learning by prompt volume. Managers should state which uses are optional, expected or prohibited and provide a fair route for questions and support.
Practical checklist
Invite concerns before demonstrations
Avoid job-loss rhetoric
Use optional paired practice
Celebrate careful checking, not speed alone
Provide a named person for follow-up questions