They could just use standardized application forms the way they do for research reports they require (the "ISA ###" forms). Those forms are easily parsable by things like pdftk and don't require any OCR.

I don't necessarily disagree with your point (since it makes complete sense), just wanting to point out that they already have a system in place for this using other means (although even there they are moving toward XML instead, likely because of what a pain it is to deal with text that exceeds the area of the input in PDFs)