Automation-Induced Contestability: AI Adoption, Employee Entrepreneurship, and Entry Deterrence

AI adoption can lower an incumbent's labour costs while simultaneously enabling displaced employees and voluntary leavers to establish competing firms. This paper develops a game-theoretic and computational framework for automation-induced contestability, examining how the credible threat of entrepreneurial entry affects incumbent pricing and the retention of automation savings. In the central simulation, the higher-cost incumbent retains 86.0% of its net automation savings. Nearly all the expected loss arises from defensive price reductions that deter entry rather than from realised competition. Expanding the contested market segment from 60 to 90 contracts reduces the retained share to 63.4%. At higher levels of AI autonomy, realised entry becomes more frequent and contributes to the erosion of incumbent profits. The analysis demonstrates that the pricing consequences of credible entry can persist independently of reductions in the incumbent's marginal production costs, subject to the model's specified deterrence conditions. A further political-economy extension examines regulatory capture and the demand for entry barriers. It identifies how incumbents may seek institutional restrictions when technological barriers to competition diminish, while distinguishing legitimate public-interest regulation from disproportionate incumbent influence. This regulatory mechanism is presented as a testable theoretical extension rather than a calibrated simulation result. The study contributes to the literature on AI-driven economic restructuring, employee entrepreneurship, market contestability, limit pricing and entry deterrence. The numerical findings derive from hypothetical parameters and explicitly stated assumptions. They are illustrative computational results, not empirical estimates or forecasts of actual corporate behaviour. Research transparency: The accompanying replication materials provide the computational model, Python code, dependencies, simulated results and figures. Generative AI tools assisted with model development, computational analysis and editorial preparation. The author directed the research and takes responsibility for the final manuscript and its conclusions.