We present the latest development of an artificial intelligence–powered, fully automated phase contrast microscope (PCM) for the diagnostic of asbestos fibers in air samples. The method follows the recommendation NIOSH 7400 and various country-dependent, derived standards. It consists of identifying fibers longer than a critical length with an adequate aspect ratio over a hundred consecutive fields. Despite its simplicity, the method is prone to a large number of errors, mostly associated with the optics (configuration of the microscope) and the operator. Consequently, a fully automated PCM was developed by motorizing the stage in three orthogonal directions, introducing a virtual phase ring—a liquid crystal display provides both color filter adjustable to the objective optics and the obturator—all being driven remotely with a tablet computer. The system allows for the capture of a stack of images and the two-and-a-half-dimension reconstruction of a single, flattened three-dimensional image. A series of convolutional neural networks is further used to classify the image and score the fibers and debris candidates for classification, thus providing fiber and debris contents for each field. A plane index is finally derived to establish the most likely focus of the next field. This process reduces counting errors and provides a reliable estimate of the measurement confidence interval as compared with reference counts from human analysts. Using the latest mobile phone and tablet technology for both capture and analysis in the GPU, one achieves 100 fields in a minute, providing a considerable gain over manual examination by a skilled operator.