Hey! Autogen is a whole new of working with agents. I believe you should think more about the logic of the prompt, and the architecture of humans x agents you are using.

For example, "after finishing the letter, ask for the letter reviewer if it is good". And then set a new "human" as a letter reviewer, where the reviewer would have been you.

I recommend trying out autogen on the UI version mentioned in the post, so you will have a better vision on how it work as a "group chat" like slack, discord, whatsapp etc. But with real humans interacting with agents.

Tell me if it worked! Nice you liked the article.