Raw LLM Responses

I do not think that it is secret but looking at your comments I think that you are way overhyping those work flows. First of all those "chat bots" you call as primitive absolutely do use agentic work flow under the hood these days. Furthermore you talk about bleeding edge use cases which I categorically disagree with. Because use case actually assumes it is being used. If it was actually used in such a way human engineers would be obsolete by now. Multi agentic work flow is not rocket science either, you just have many, many agents talking to each other burning millions of tokens doing so. Not only is it not guaranteed to bring expected results (althought there are big hopes and money in it), it is not even guaranteed to be cheaper than humans were those results achieved.