Conventional wisdom holds that artificial intelligence will reduce demand for lawyers by automating tasks that once required human labor. This Essay challenges that assumption by applying the Jevons paradox to legal practice. In 1865, William Stanley Jevons observed that efficiency gains in resource use increase rather than decrease total consumption. This Essay argues that AI will expand legal work, create new categories of labor, and restructure rather than diminish the legal profession.
This Essay identifies four mechanisms through which the paradox manifests in legal practice. (1) Efficiency gains lower the effective cost of legal services, enabling clients who previously could not afford lawyers to enter the market. (2) Reduced costs surface vast latent demand: Ninety-two percent of civil legal problems faced by low-income Americans currently receive inadequate or no legal help, representing untapped market potential that AI could activate. (3) Quality standards evolve to expect the thoroughness that technology enables, converting efficiency gains into quality improvements rather than cost reductions. (4) AI creates new legal tasks (prompt engineering, quality verification, ethical oversight) even as it automates existing ones. Historical precedents from computerized legal research, word processing, and e-discovery demonstrate that legal technology consistently expands, rather than contracts, the profession.
This Essay examines implications for access to justice, showing how AI efficiency gains could serve the fifty million Americans currently excluded from legal assistance while simultaneously increasing aggregate legal work. The analysis explores how legal education must transform to prepare lawyers for practice where AI handles certain cognitive tasks while distinctly human capabilities become more valuable. Finally, this Essay identifies a regulatory complexity feedback mechanism: As AI reduces compliance costs, political economy dynamics may generate additional regulation that recreates the work AI displaced.
