Why extraction quality matters more than simply having OCR
An OCR engine that reads 80% of the fields correctly does not save time - it creates review work on every single document. What makes the difference is handling uncertainty explicitly: 4expenses does not just fill in the fields, it flags how confident it is about each one. High-confidence fields move on without any intervention; uncertain ones are highlighted and ask for a one-second confirmation. Human effort then goes exactly where it is needed, instead of being spread across re-reading the whole document.
VAT read by rate, not as a single amount
A receipt from a mixed retailer usually contains items at different VAT rates. If the system only extracts the total, the accountant has to open the document anyway. 4expenses reads the breakdown by rate wherever the receipt shows it and carries it through to the export, which noticeably shortens bookkeeping work and reduces errors when VAT is deducted.
Thermal receipts, skewed photos, poor lighting
Real documents do not look like the ones in a demo. Thermal receipts fade, photos are taken in a car or on a restaurant table, some documents are crumpled. The app corrects perspective, boosts contrast and crops the document out of the background before reading it. When a document is unreadable anyway, the employee gets immediate feedback to retake the photo - while the receipt is still in hand, not three weeks later.
The extracted data, and the original document too
Automatic extraction does not replace the supporting document: the original image stays attached to the expense report, with the date and time it was uploaded. During an inspection or an internal audit you can always compare the recorded amount with the document it came from, and the history shows whether a field was corrected manually after reading, by whom and when.