I love this idea. It would be incredibly useful!

I feel LLMs are great at suggestions that you follow up yourself (if only for sanity checking, but nothing you wouldn't do with a human too).