Productivity, Work, and Leisure: From AI Task Efficiency to Human Welfare

Abstract

Artificial intelligence can accelerate individual tasks without necessarily raising organizational throughput, wages, or workers’ usable time. This paper develops a conditional framework tracing a technology shock through four connected transformations: from task efficiency to quality-accepted organizational capability; from productive capability to the incidence of gains among output, prices, profits, pay, and usable time; from purchasing power and time to household welfare; and from household choices to the demand that firms subsequently face. Two feedbacks close the chain: the perceived fairness and credibility of compensation and performance assessment may change effort, quality, cooperation, and method disclosure, while household demand may alter production opportunities. A central cross-stage implication concerns bottleneck location. Accelerating a non-binding stage can leave throughput unchanged while freeing effective labor capacity, whose allocation depends on staffing, assignments, and contractual rights; accelerating a binding stage can raise accepted output and consume capacity at complementary stages. These outcomes distinguish slack production capacity from legally or practically discretionary employee time. We propose a staged empirical comparison that observes technical performance, accepted flow, labor allocation, and employee choice separately. A finite fixed-salary workload contract is one illustrative institution, not the theory’s center. The framework integrates established findings rather than claiming a new universal theorem. Its proposed contribution is an explicit, testable account of the conditions under which technical progress becomes—or fails to become—improved human life.