Todays organizations receive large amounts of written communication either sent via fax, snail mail or digital channels. To process these documents through manual means is not feasible or scalable to meet customer or partner service quality expectations. While several automation tools provide document processing capabilities, many of these systems focus on OCR and less so on extracting actionable information. These systems tend to rely on approaches that are brittle to changing document formats and are expensive to maintain in the long term. In this talk we share our best practices that organizations can use when evaluating out of box products, and when developing their own machine learning and computer vision based solutions with the goal to deliver scalability and providing measurable value to the servicing their vendors and customers.