I'm setting up a process involving extracting fields from invoices sent to us by vendors, but I'm finding that the prebuilt invoice model in AI Builder doesn't work well for a select few vendors, but works very well for other vendors. Because of this issue I'm looking into setting up a custom model for those problem vendors, ideally I would like to just augment the existing prebuilt model, since it handles a decent number of our invoices well already.
I'm wondering if when I create a custom model and select 'Invoices' as the document type, does that retain the training of the prebuilt models? Like can I create a new model if the invoice type can I train it on only the problem vendor invoices and expect the same performance on other invoices? Or will my new model only work for the problem vendors?