So Management went to ChatGPT and asked, "can you write a launch email about $event_x " and you then go there to ask, "what did management want to say in this email about $event_x"?
How likely is it that the output summary is mostly made of stuff that the LLM made up in the first expansion process? (eg, you're getting summarized noise if the original signal - the prompt - was much shorter than the email)
