If you don’t have an AI rule yet, your teams already have one: their own. Better to know about it and make use of it than to ignore it.

Use is already established
Ask around, simply. You’ll find that a good share of your people already use an AI assistant, often on a free personal account. To rephrase a tricky email, summarise a long document, prepare a presentation. They don’t always say so, for fear it looks like cheating.
Where the risk lies
That they save time is rather good news. The risk comes from two things. Data going into tools nobody has approved, sometimes customer information or internal figures. And results nobody checks, with mistakes that end up in a document sent out.
Bring the use out of the shadows
The first step is to lift the taboo. Say clearly that using AI is encouraged, within a framework. Run a short session where those already using it show what they do. You’ll often find very good practices there, and a few uses to correct quickly.
Spot what deserves to be rolled out
Among these spontaneous uses, some save a great deal of time and can serve a whole team. The best first use cases are often found here, much more than in vendor demos, with one decisive advantage: your teams have already tested them on their real work.
Then set the framework
Once the use is visible, the policy becomes easy to write, because it starts from what already exists. Which tools, which accounts, which data never, who checks. Your teams will accept it all the more readily for having helped build it.