I have noticed that the current implementation of the GHL AI feature only supports FAQs, and it does not allow the use of plain text documents or PDF files for generating responses. I believe that expanding this feature to include support for plain text documents and PDF files would greatly enhance its capabilities and make it even more versatile.