Optical Character Recognition
Bank statements can be scanned and turned into PDF files. They can then be processed by optical character recognition. Two obvious features of bank statements are that they look just right for transfer to a spreadsheet, and that there is a running total which can be used as a check on the scanning. We have software which can do all this, and then analyse the results automatically for inclusion in the accounts. Our system generates accountant’s working papers on the side.
Our current OCR system is a very basic system which is subject to continuous improvement. It is already productive enough to allow us to relax a bit and to concentrate on improving the basic system first.
No OCR system is currently able to read cheque book stubs. If there are a lot of cheque payments then these will have to be typed in by a human. If you type in a lot of narrative on Excel, you will see it helping you by “autocompleting” what you type, which we take advantage of.
However, the autocomplete system needs a unique trigger in order to kick in. If there are two narratives almost the same, like “International Trading” and “International Products”, then autocomplete is of less use. Our software provides a separate toolpad which continually looks for awkward long-trigger narratives and lists them separately, so we just click on a button to enter them. We call this super-autocomplete. Apart from this, there is another toolpad which just lists the commonest narratives typed in so far. A lot of the work is repetitive, and this helps.
We also have autocomplete for dates. If you type in just the day, then the month and year are copied down from a previous entry automatically.