Wellllllllllllll I'm designing a qualification / coaching bot, that will solve advanced problems without relying on chat gpt to just "generally" solve things as it sees fit. This bot will answer either: "Is this client qualified to work with me?" or for a pre coaching call: "Where is this client currently at with their business?" The main workflow talks to the client, and does all the high level analysis of the client problem. When there is either missing information to solve the problem, or a weak answer from the client, the main workflow will go through the analysis process, which is broken down into 3 sub problems. In this case, it's niche, budget, and agency operations. Each folder has a master workflow that analyzes what is known about the sub problem at a high level, and solves it. If it can't solve it, or the answer is weak, I have a series of sub workflows for each "part" of the sub problem that needs to be investigated, that is designed to gather more data from the user, ask probing questions, before passing it back up to the master sub problem workflow. THHHHHEEENNN.... once the whole sub workflow is complete, it sends that back up to the main workflow to see if the problem is solved, and if not, it goes down the next sub problem folder, and rinse and repeat.