Help the user understand how often the AI system may make mistakes.
Engage in disambiguation or gracefully degrade the AI system’s services when uncertain about a user’s goals.