For the purposes of an llm "reading" a pdf, it just renders it as an image. The file format does not matter. Let's say you have documents that already exist, a robust ocr solution that can handle tables and diagrams could be very valuable.