How can employees help identify worthwhile AI use cases?
Ask employees to describe a recent example of work they repeat and would like to change. Find out what triggered it, what result was needed and why it took so much attention. Then check whether the step can be removed or simplified before you add a tool.
The summit notes included asking people about work they do over and over, and I'd start with those accounts. Dislike alone isn't a good enough reason to automate something, so examine the task first. Separate the necessary review from copying, searching or entering the same information again.
Sometimes the best change is a clearer source or an agreement between teams. If a tool does fit, let the employee help define a successful result, including the exceptions they recognize and the work that stays with them.
Return a decision on every proposal, even the ones you don't pick. Say whether it needs more evidence, conflicts with a constraint or sits behind another priority. Asking for ideas creates an expectation of consideration. It doesn't promise that every frustrating task becomes an AI project.