Surgical pathology remains the cornerstone for the diagnosis, staging and prognostic stratification of cutaneous melanoma. The emergence of digital pathology and artificial intelligence (AI), notably deep learning, is transforming this field. Convolutional neural networks trained on large repositories of digitized histological slides demonstrate diagnostic capabilities comparable, and in specific tasks superior, to board-certified dermatopathologists. Current applications encompass the automated classification of melanocytic lesions, including the challenging differentiation between benign melanocytic naevi and early melanoma. Beyond morphological diagnosis, AI algorithms extract virtual biomarkers directly from haematoxylin-eosin-stained slides, predicting underlying molecular alterations, including BRAF mutational status, and offering a transformative approach to prognostic stratification, enabling the extraction of morphological and spatial features that are imperceptible to the human eye. Challenges remain, including dataset bias, inter-laboratory standardisation of whole slide imaging workflows, model interpretability, and medico-legal accountability. The regulatory landscape is rapidly evolving: the European Union Artificial Intelligence Act classifies medical AI as high-risk. The 2024 FDA de novo authorization of DermoSensor, an adjunctive AI-enabled device for assessing suspicious skin lesions, illustrates broader progress in dermatologic AI, although it does not concern surgical pathology. This review provides a comprehensive overview of the current state and future perspectives of AI in management of cutaneous melanoma. These new tools will augment the central role of the pathologist in precision oncology, in refining adjuvant treatment decisions and surveillance strategies in routine clinical practice.