when make a request of a person, you go back and forth with corrections and clarifications. I think it is going to take time for people to realize you need to do the same with LLM chatbots