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.
Immediately update or convey how user actions will impact future behaviors of the AI system.