How do we make an acceptable use policy usable during a workday?
A statement about responsible AI is easy to agree with and hard to apply, and people need to see the boundary in the middle of a task. Name what an AI tool is excluded from, in ordinary language: which information stays out, which decisions stay with a person and which external actions aren't permitted.
For one workflow at a time, I'd write each exclusion with a concrete example from the team's own work, and give each one an approved alternative.
Every exclusion needs a way forward. If the tool can't use a confidential customer record, point to the approved route for getting help. If it can't send a response, explain how to prepare a draft for review.
Let employees try the guidance on sample requests and explain the boundary back in their own words. When they disagree, that can point to a policy gap and doesn't have to mean a training failure.
Give the document an owner and a way to ask questions. Keep it short enough to consult, with links to the detailed rules when needed. People can then keep doing the permitted work without guessing and without stopping everything.
What to try this week
Draft the exclusions for one workflow in plain sentences, then add an approved alternative to each. Ask a few colleagues to explain the boundaries back to you and see where they disagree.