Produtividade líquida no empreendedorismo individual: quanto a IA economiza depois da revisão e do retrabalho? Ensaio teórico com proposições e agenda de pesquisa

The most widespread promise of generative artificial intelligence (GenAI) for solo entrepreneurs is time savings. Experimental evidence documents substantial gains on well-defined tasks, but large-scale adoption studies find modest self-reported savings, and at least one randomized trial with experienced professionals found an increase, rather than a reduction, in completion time. This theoretical essay argues that part of this discrepancy stems from conflating gross productivity — the time saved in generating an output — with net productivity — the time and value saved after accounting for task specification, verification, correction, rework, context switching, and the expected cost of errors that escape review. For solo entrepreneurs, this distinction is especially relevant because the same person generates, reviews, corrects, and answers for errors, with no division of labor to absorb these costs. Integrating the economics of productivity and time allocation, the literature on quality costs and rework, research on automation and overreliance, human–AI interaction studies, and recent empirical evidence on GenAI at work, we propose an accounting model of net productivity, with explicit cost components and boundary conditions, and derive eight propositions. We discuss implications for solo entrepreneurs, limitations, and a study design with task-level randomization and an illustrative power calculation. The article presents no empirical data; its propositions are intended for testing.