The Offloading Paradox: Business Simulations and Managerial Judgement in the Age of Generative AI

As generative AI increasingly performs the analytical work of management education, this paper asks what developmental work must remain with the learner if managerial judgement is to form. It answers by proposing a conceptual theory of offloadability, distinguishing between managerial work that can be delegated to generative AI and work whose educational value is altered when delegated. Grounded in recognition-primed decision-making, the theory is developed through a running example of a postgraduate business simulation module built around SimVenture Evolution, used as an illustrative designed object rather than empirical evidence. The paper identifies two sources of non-offloadability. The first is coupling: some managerial situations require a synthetic reading of interacting and competing objectives, and judgement develops through the learner performing that reading. The second is answerability: some work requires a responsible party who can own, defend, and learn from consequential decisions. Neither source is primarily a claim about what AI can compute; both concern what learners must do for judgement to develop. The theory therefore rests not on current model limitations but on the educational conditions through which managerial judgement is formed. The paper reframes AI integration as a curriculum and assessment design problem: the issue is whether learners still undertake non-offloadable work.