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 access an explanation of why the AI system behaved as it did.