Ebrahimi Kordlar, Ali, Safaei, Mahdi. (1404). The Primacy of Control: A Risk-Centric TCO Framework for Generative AI and the Financial Irrelevance of Productivity. سامانه مدیریت نشریات علمی, (), -. doi: 10.22067/ijaaf.2025.94743.1574
Ali Ebrahimi Kordlar; Mahdi Safaei. "The Primacy of Control: A Risk-Centric TCO Framework for Generative AI and the Financial Irrelevance of Productivity". سامانه مدیریت نشریات علمی, , , 1404, -. doi: 10.22067/ijaaf.2025.94743.1574
Ebrahimi Kordlar, Ali, Safaei, Mahdi. (1404). 'The Primacy of Control: A Risk-Centric TCO Framework for Generative AI and the Financial Irrelevance of Productivity', سامانه مدیریت نشریات علمی, (), pp. -. doi: 10.22067/ijaaf.2025.94743.1574
Ebrahimi Kordlar, Ali, Safaei, Mahdi. The Primacy of Control: A Risk-Centric TCO Framework for Generative AI and the Financial Irrelevance of Productivity. سامانه مدیریت نشریات علمی, 1404; (): -. doi: 10.22067/ijaaf.2025.94743.1574
The Primacy of Control: A Risk-Centric TCO Framework for Generative AI and the Financial Irrelevance of Productivity
Iranian Journal of Accounting, Auditing and Finance
1Department of Accounting, Faculty of Accounting and Financial Sciences, College of Management, University of Tehran
2Ph.D. Student, Department of Accounting, Faculty of Accounting and Financial Sciences, College of Management, University of Tehran
چکیده
Purpose: The prevailing narrative frames generative AI as a tool for productivity enhancement. This paper challenges this view as financially incomplete and potentially misleading for strategic investment. We argue that a new financial evaluation paradigm is required, centered on risk and control.
Methodology: We develop and test a stochastic TCO framework comparing API and fine-tuned OS strategies. The model incorporates probabilistic estimates for operational and risk variables, which are then analyzed using Monte Carlo simulation and Sobol sensitivity analysis to deconstruct the key drivers of financial performance.
Findings: Our findings reveal a fundamental inversion of classical IT value logic. Sensitivity analysis demonstrates that traditional productivity gains are financially irrelevant in determining the optimal strategy. Instead, the model's outcome is almost entirely dictated by the cost and probability of critical errors (λ, P). The simulation, driven by this risk dominance, shows the OS strategy has only a 6.58% chance of being financially superior, as its TCO is heavily burdened by the immense cost of its insourced risk management (the HIL process).
Originality/Value: This paper redefines the financial calculus of AI investment. It provides a formal model that explains why the economics of trust and control supplant the economics of speed. We use this framework to interpret real-world market behavior, such as multi-billion-dollar acquisitions in the AI safety space, as rational, risk-mitigation investments. The framework offers a new and essential lens for corporate governance, risk management, and strategic decision-making in the AI era.