Optical Character Resolution (OCR) has been around for quite some time in the industry. Performing OCR at endpoints has not been encouraged due to the unavailability of adequate computing and memory to cater to the resource-intensive pre-processing steps. This paper explores our approach, the criteria we considered, and the outcome of our endeavour to enable OCR capability at endpoints to achieve adequate accuracy without bogging down the system resources while running in the background as a silent process.