This paper develops a real options framework to analyze optimal adoption and switching strategies for human–AI collaboration. I model four distinct configurations, human-only, AI-only, hybrid, and human-in-the-loop (HITL), within a unified dynamic decision-making setting, deriving value functions and optimal thresholds under uncertainty. Three key findings emerge: First, the HITL strategy consistently dominates across varying levels of task uncertainty, performance differentials, and volatility. Second, adoption thresholds respond intuitively to parameter changes: higher uncertainty increases the value of flexibility, volatility raises opportunity costs, and superior human performance requires stronger financial justification for AI integration. Third, when extending to dynamic settings with switching options, we find that regardless of the initial operational regime, the optimal switching action universally targets the Human-in-the-Loop (HITL) strategy, though the required demand threshold for transitioning varies significantly across starting strategies. These results align with recent empirical evidence on human–AI collaboration and provide practical guidance for managers designing adaptive technology integration roadmaps.