Executive Summary: AI Is Not Taking Jobs. It Is Disassembling the Employment Unit.
LFP-12.01 — The AI Disruption#
The question dominating public discourse about AI and employment is the wrong one. How many jobs will AI eliminate? The more consequential question for health coverage is: what is AI doing to the structure of employment relationships? Job elimination creates a quantitative problem. Employment restructuring creates a structural problem. Workers remain employed but in arrangements that fall outside the ESI framework. They earn too much for Medicaid and most ACA subsidies. Their employment relationships are too fragmented for any single employer to sponsor group coverage. They fall between coverage categories rather than into any of them.
The mechanism is task disaggregation. A full-time job is a bundle of tasks that together justify a full-time employment relationship with a single employer. AI does not destroy the tasks. It changes the economics of performing them by reducing the labor input required per unit of output. A marketing department that previously justified four full-time positions now justifies one senior professional with AI-powered content generation, campaign management, and analytics. The work is still happening. The employment relationship that bundled it into four benefit-eligible positions has dissolved. Brynjolfsson, Li, and Raymond documented this dynamic in a 2023 NBER study of 5,179 customer support agents: access to a generative AI assistant increased productivity by 14 percent on average and by 34 percent for less experienced workers, precisely the workers filling mid-level team roles.
Generative AI reaches into non-routine cognitive work, the category that labor economists had identified as durable against prior automation waves: research synthesis, content production, basic legal analysis, financial modeling, project coordination. These tasks require reduced human labor input per unit of output, and the employment units that bundled them dissolve.
The KFF 2024 Employer Health Benefits Survey found that among firms with three to nine workers, only 46 percent offered health benefits, compared to 93 percent of firms with 50 or more employees. ESI coverage as measured by the Current Population Survey held at approximately 53.8 percent of the total US population in 2024, statistically unchanged from 2023, through a period of low unemployment and strong wage growth. The stagnation persists not because workers cannot find jobs but because the jobs they hold are structured in ways that do not produce ESI eligibility. AI is pushing more of the workforce into the left tail of the firm-size distribution where offer rates are lowest and where group coverage viability is most constrained.
The coverage consequence is structural. The product and regulatory responses designed for workers who lose jobs do not apply to workers whose employment relationships were disaggregated before they produced coverage eligibility.