The best first AI project is rarely the most impressive one. It is the one that removes a small, repeated frustration without making the team feel like a science experiment.
Start by following one piece of work from beginning to end. Notice where information is copied, where the same question is answered for the tenth time, and where a good decision waits for a busy person to return. That is your opportunity map.
A useful first system has a narrow input, a clear human review point and an outcome you can see in a week or two. It might be a triage view for enquiries, a private search across internal documents or a draft that a person can approve before it goes anywhere.
The goal is not to remove people from the process. It is to return attention to the parts of the work that need judgement, empathy and taste.