OCR for really basic documents might be "hello world", but as usual real world practical applications have all sorts of interesting things that make this a harder problem than it seems.