Comment by maximilianroos

Have your AI talk to their AI

Then, if the AIs are positive, the human principals can talk

Seems quite reasonable!