Hi, we use the Invoice Processing capability of AI Builder. Since the performance of the prebuilt model was not good enough for all invoices, we've started labelling and training ourselves. In the credit consumption from the last 4 weeks we now see significant higher consumption than expected. My hypothesis is, that we are now being billed for the (custom) 'Document Processing' model instead of the (prebuilt) 'Invoice Processing' model, due to the fact that we have now customised the model.
1) Is this correct?
2) Since we need the custom training only for a few types of invoices, is there a way we can do this in some kind of 'hybrid' mode, where only the custom trained invoice types are extracted with the expensive custom model and all the other invoices are extracted with the prebuilt model?
Thanks a lot in advance for your help.
