Firm Shrinkage and Niche Expansion: Artificial Intelligence, Endogenous Demand Differentiation, and Human Labor Demand

Can AI-induced firm entry change what consumers learn to value and thereby reverse an initial decline in human labor demand? This paper develops a dynamic free-entry model in which active product variety raises future demand differentiation through consumer learning. AI lowers operating costs and human labor per firm. Entry responds immediately to lower costs, while the demand state adjusts through experience and supports further entry. The impact elasticity of sectoral human labor demand is φ/2 − η, where φ and η are the AI elasticities of operating-cost reduction and firm-level labor reduction. With a variety-to-learning elasticity α ∈ (0,1), the steady-state elasticity is φ/(2 − α) − η. When φ/2 < η < φ/(2 − α), labor demand initially falls and eventually exceeds its pre-shock level. This delayed niche reinstatement operates through viable product niches without requiring new human tasks within firms. A stylized allocation of finite learning attention provides a foundation for the demand transition. Extensions characterize how production labor, purchasing power, and market-access costs limit the mechanism. The model identifies a lagged entry-learning-entry sequence that distinguishes demand feedback from a one-time reduction in entry costs.