n8n looks extremely powerful and I’m excited to get started with it. But I’m not really sure how we’d go about training it to do what we want. In our case we get what we would call a deal from our customers that we need to process. There are a number of things that could affect how we process the deal such as, does the client have any paid holidays we need to include, or is it in a state with tight regulations on certain industries like healthcare. I don’t have an account yet, so I can’t really get in and poke around but after watching some videos here’s what I was thinking.
Basically we would try to identify all of the attributes about a deal that would factor into processing it - location, client, submitting customer, etc. and creating a database table to hold that info. When a new deal is added by a customer, we’d copy all of that logic necessary for AI reasoning into a new row in that table. I’m hoping there’s a trigger for row added in a database integration that would then kick off the automation where the AI would analyze the deal and based on what it finds, create a summary of things that we need to do to finalize it.
So, then, the main thing I’m wondering about is how to train the AI using RAG on how to reason when analyzing a new deal.