The AI Deskilling Paradox (cacm.acm.org)

AI is triggering a modern wave of deskilling: as tools take over tasks once done by trained professionals, people cede expertise and core judgment. Empirical examples underscore the risk—A 2025 Lancet Gastroenterology & Hepatology study found colonoscopists’ precancerous lesion detection dropped from 28.4% to 22.4% when AI support was removed—and law and education studies report critical errors among users who over-rely on generative systems. A Microsoft Research survey (2025) found knowledge workers feel tasks are cognitively easier with GenAI, but that ease often reflects surrendered problem-solving rather than true skill transfer. Senior experts may gain productivity, while junior workers and wages can suffer, potentially hollowing out long-term institutional knowledge. Researchers urge proactive responses: design AI to build, not replace, human skills; adopt “hybrid intelligence” models that measure both technical output and human capacity (metrics like employee AI self-efficacy and psychological safety); and refocus training on AI-specific critical thinking, verification, and domain literacy. Policy and workplace design should preserve human accountability for consequential decisions and track skill trajectories as rigorously as productivity. The takeaway: AI can amplify human creativity and reasoning—but only if systems, incentives, and training are engineered to prevent erosion of the very expertise we need to steward them.