The company still wants the resume. They just want the information extracted accurately. OCR may not be involved because it isn’t accurate enough to associate specific chunks on the resume with specific questions. If companies just had the form they would have three accurate info, but would have to generate a resume internally for human use (which isn’t a bad idea necessarily). If they just had the resume uploader then they would have to have a person manually extract the information.

So they ask for both–because they want accuracy and the original document and there aren’t tools around to give them both today.