Robotic pancreatectomy is a complex procedure where surgical performance critically influences patient outcomes. This project aims to leverage computer vision to analyze robotic surgery by deconstructing it into phases, steps, and instrument-tissue interactions. The first phase establishes an international Delphi consensus with expert pancreatic surgeons to define surgical phases, key steps, critical errors, and performance metrics. With this framework, annotated videos from multi-center robotic pancreatectomies will be analyzed to identify and classify phases, steps, errors, and gestures (e.g., cutting, retraction, suturing), correlating them with outcomes like postoperative pancreatic fistula. Computer vision metrics will be used for objective evaluation of surgeon performance, providing granular insights into surgical performance. We aim to create tools for objective evaluation, real-time feedback, and AI-assisted guidance, enhancing surgeon performance and patient outcomes.