Automation promises faster and more scalable work, yet its consequences for human expertise remain unsettled. Systems that improve assisted performance can also reduce task participation, alter legitimate knowledge, intensify reliance, and narrow everyday learning opportunities. This Critical Review Article examines that expertise paradox by integrating evidence on tacit knowledge, skill dependency and deskilling, informal workplace learning, and conditions that may preserve expertise inside automated workflows. The review combines structured evidence identification and screening with critical synthesis of mechanisms, contradictions, alternative explanations, and boundary conditions. The literature does not support a simple conclusion that automation either erodes or augments expertise. Instead, it indicates heterogeneous trajectories shaped by task exposure, epistemic visibility, work design, career stage, opportunities for practice and feedback, human–AI calibration, and the organization of knowledge exchange. The article proposes an Expertise-Resilience Framework that distinguishes erosion pathways from resilience pathways while treating their connections as original integrative synthesis rather than a validated causal model. It also identifies unresolved problems involving unaided competence, knowledge transfer, occupational hierarchy, failure recovery, and multilevel inference. Its contribution is a bounded organizational account of when automated workflows may weaken, redistribute, or help sustain expertise and what evidence would be required to discriminate among these possibilities.
