I suspect 90% of the processing is done manually with ReceiptBank. If you run a receipt through OCR it will return a blob of unstructured text, if you have a receipt with sub totals then it’s very difficult to intelligently parse out the amounts, i.e. item totals, VAT, net and gross totals.
Even isolating the supplier name is a challenge, some receipts display a company logo rather than the actual company name. If you have any crinkles that skew the text, the OCR accuracy drops significantly, it’s a bit like how they use images of skewed text for the purpose of blocking robots, they are intended not to be machine readable.
It would be a nice feature to have, but I think for it to be useful we’d need to look at utilising some sort of back office team. Of course if Receipt Bank change their mind, we’d be happy to look at some form of integration too.