Microscopy for identification of spermatozoa is a labor-intensive task in forensic laboratories.Automation of imaging and image interpretation has the potential to greatly improve throughput.Automation may also increase quality by helping to ensure consistency between operatorsor laboratories.We describe the evaluation, validation and ongoing implementation of MetaSystems Metafer, anautomated microscope system which includes robotic feeding of microscopic slides and automatedimage analysis by artificial intelligence. The system can be loaded with up to 80 microscopicslides, read the slides unsupervised, assign a score to potential sperm in each slide and presentthose to an operator in order ranked by likeness to a sperm cell. The operator can then make ajudgement on whether to accept the presented objects as sperm or not.We have investigated how to optimize specificity and sensitivity with regards to the operatoreffort needed. It was found that imaging in multiple focus planes considerably increases specificityfor spermatozoa. We have also mapped the distribution of scores for sperm and non-sperm objects,allowing the laboratory to decide on a well-informed policy on how many potential spermobjects that must be reviewed to find positive samples with sufficient sensitivity. Whether certainsample types are unsuitable for automated analysis is also investigated. Finally, we have performedan in-house validation, finding that sensitivity using the automated system surpasses thatof manual microscopy.