A commonly acknowledged barrier for the adoption of new computer-assisted orthopedic surgery (CAOS) technologies relates to a perceived long and steep learning curve. However, this perception has not been objectively tested with the consideration of surgeon-specific learning approaches. This study employed the cumulative sum control chart (CUSUM) to investigate individual surgeon's learning of CAOS technology by monitoring the stability of the surgical process regarding surgical time. Two applications for total knee arthroplasty (TKA) and two applications for total shoulder arthroplasty (TSA) provided by a modern CAOS system were assessed with a total of 21 surgeons with different levels of previous CAOS experience. The surgeon-specific learning durations identified by CUSUM method revealed that CAOS applications with “full guidance” (i.e., those that offer comprehensive guidance, full customization, and utilize CAOS-specific instrumentation) required on average less than ten cases to learn, while the streamlined application designed as a CAOS augmentation of existing mechanical instrumentation demonstrated a minimal learning curve (less than three cases). During the learning phase, the increase in surgical time was found to be moderate (approximately 15 min or less) for the “full guidance” applications, while the streamlined CAOS application only saw a clinically negligible time increase (under 5 min). The CUSUM method provided an objective and consistent measurement on learning, and demonstrated, contrary to common perception, a minimal to modest learning curve required by the modern CAOS system studied.