This is where Microsoft (and Apple) has a leg up -- they can hook the UI at the draw level and parse the interface far more reliably + efficently than screenshot + OCR.