Help the user understand how often the AI system may make mistakes.
Make it easy to invoke or request the AI system’s services when needed.
Enable the user to provide feedback indicating their preferences during regular interaction with the AI system.