Six years ago BLS staff read and manually classified hundreds of thousands of written descriptions of work-related injuries and illnesses each year. Today, more than 85% of these classifications are assigned by a deep neural network that is more accurate, on average, than trained human workers. In this presentation, Alex will discuss how BLS addressed some of the many challenges inherent in this transition including how to build these systems, how to decide when and how to use them, and how to monitor and maintain them to continually improve performance.