Explains a collaborative approach to choosing AI projects that combine user value, manageable risk and achievable technical performance, drawing on work at Carnegie Mellon’s Human-Computer Interaction Institute.