An AI system recommends an action. A person reviews the reasoning, considers the context and decides whether to proceed. That interaction captures one of the central questions facing organizations today: As AI moves from generating answers to taking action, how do we determine how much autonomy to give it?