In complex public organizations, job-rotation policies are routinely used to build flexible workforces, yet their impact on adaptive performance remains unclear. This thesis investigates how the frequency and volume of rotation reshape organisational memory and, in turn, influence adaptation. Drawing on literature in organisational learning, complexity theory and contingency theory, it develops a conceptual framework that links memory loss, knowledge diffusion and re-learning to adaptive fit. The framework is operationalised in an agent-based extension of Epstein’s “Growing Adaptive Organizations” model, calibrated for a stylised public bureaucracy. Four task-environment regimes – stable, peaks, evolving and flipping – were simulated across nine combinations of rotation interval and volume, with stochastic hand-over accuracy and memory persistence. Results show that, within the model, any rotation schedule underperforms a no-rotation baseline; the best-performing rotating configurations lag by 1–18 percentage points in reliability, depending on environmental volatility. Rotation volume drives outcomes more strongly than interval: moving larger shares of staff consistently erodes performance, whereas shortening intervals only matters when large peaks coincide with rotations. Stable environments buffer memory loss, while environments with sharp, predictable shocks can benefit when rotations are timed to ‘prime’ units just before demand spikes. Excluded factors such as motivation boosts, informal networks and long-run career gains suggest the model is conservative about rotation’s upside. The study refines the memory-versus-flexibility debate and proposes four evidence-based interventions: align rotations with forecast shocks, cap simultaneous moves, use stable departments as learning incubators, and formalise hand-over protocols. Overall, job rotation is neither intrinsically adaptive nor maladaptive; its value is contingent on environmental cadence and design precision.