Document recognition

This is only a survey, or a test, to see if this is possible within a fair budget. My question is for you to make optical recognition of documents coming from scanned images or photos or PDF's from various documents. You will have an original invoice that is the base-document. From scanned images, photos or PDF's the software will have to recognize the details of the document and link it to this base original invoice. Sorry if I can't explain it better. What I essentially need is a program that places all sorts of documents in an accounting application - with extracted data. And places it right, with correct links. My first question - is it possible to make a algoritm/recognition app, that can extract data out of various documents optically. If so, please make a test. And let's take it from there.