HAT model reveals when AI replaces human workers in organizations

A new paper on arXiv (2607.20781) by Banerjee and Singh presents the Human-AI Task Allocation (HAT) model, a rigorous analytical framework to predict when AI will replace human employees in hierarchical organizations. The model's central insight is the formal encoding of an economic asymmetry: humans acquire skills through costly, slow learning, while AI capabilities scale with dramatically lower marginal costs. Using this asymmetry, the authors derive the Human-AI Substitution Principle, a precise condition that determines replacement thresholds based on risk-adjusted costs, organizational depth, deployment scale, and strategic adaptation. The model predicts abrupt, not gradual, workforce transitions as AI crosses cost-efficiency thresholds.

The HAT model yields several actionable predictions. Middle-management roles exhibit the highest vulnerability to automation because their tasks combine routine coordination with moderate skill requirements — a sweet spot for AI substitution. Highly skilled workers face risk only above a certain skill threshold, which itself depends on organizational depth and baseline cost differentials. Interestingly, risk heterogeneity across tasks can sustain hybrid human-AI teams without requiring a minimum human fraction. The paper also predicts flatter organizational structures with wider spans of control as AI absorbs middle layers. This work bridges automation economics, organizational theory, and AI governance, offering a quantitative tool for workforce planning and strategic AI adoption.