It would be helpful if we could:
flag suggested conversations that were not helpful (more often than not, there's at least 1 irrelevant suggested similar convo)
have the AI generate more similar conversations if possible (when the user flags one of the suggestions was not helpful, AI could suggest another if relevant)