Help the user understand what the AI system is capable of doing.
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.