 The study reports real-world performance of deep ensemble for recognition of malignancy, DRM, and AI as a medical device tool in skin lesion assessment at two national health service hospitals between July 2021 and October 2022. DRM demonstrated high sensitivity for detecting melanoma or malignancy with benign lesion specificity ranging from 40.7% to 73.4%. The tool identified 15% to 31% of cases as eligible for discharge and reduced the caseload for hospital specialists in two pathways. The study suggests ski elements for post-market surveillance of AIMDs, including ongoing monitoring of performance, user feedback, and regular updates to improve accuracy and safety. This article was authored by Lucy Thomas, Chris Hyde, Dan Malarkey, and others.