In the early phases of artificial intelligence integration, these technologies often serve as powerful enablers that significantly enhance the efficiency and output of human labor. Workers discover that with AI-driven tools, they can accomplish tasks more swiftly, with improved accuracy and creativity, which in turn temporarily enhances productivity and tends to elevate wages. This period of accelerated output, fueled by the collaboration between human expertise and machine intelligence, typically precedes the stage when automation begins to displace certain professions altogether, thereby altering the structure of the labor market.
However, according to Ioana Marinescu, an associate professor at the University of Pennsylvania’s School of Social Policy & Practice and coauthor of a recent Brookings Institution study on what she calls “intelligence saturation,” the benefits of this wage expansion may already be nearing their apex. In her conversation with *Business Insider*, Marinescu revealed that her research suggests the economy has surpassed 14% automation of intelligence-based activities — a stage she believes is closer to the threshold of diminishing wage growth than to a scenario of negligible automation. This level of automated cognitive work, she warns, may indicate that the cycle of rising pay fueled by new digital tools is approaching its natural limit.
Drawing on the analytical framework from her Brookings model, Marinescu and her coauthor Konrad Kording postulate that once approximately 37% of cognitive or intelligence-oriented tasks are handled by automated systems, wage growth across the economy could begin to reverse. Nonetheless, she clarifies that the exact inflection point — the moment when pay actually starts to decline — depends heavily on several parameters such as the speed of technological diffusion, the adaptability of labor markets, and the interplay between different sectors of work.
Marinescu highlights a series of warning signs that could signal the economy is nearing this crucial threshold. Among the earliest indicators is a visible deceleration in wage growth, which would be accompanied by a structural transition in employment from intelligence-intensive occupations — such as those involving analytical, linguistic, or creative reasoning — to roles that are more physical or manual in nature. Her research indicates that automation of intelligence tasks has been consistently advancing since the 1980s, transforming the landscape of work over successive decades. She cites data demonstrating that the share of routine cognitive jobs diminished from 49% in the late 1970s and 1980s to about 35% by 2018, a 14-percentage-point decline that she interprets as a measurable proxy for the degree of automation within knowledge-based professions.
This ongoing transformation places today’s workforce on what Marinescu and Kording term the “AI pay curve.” The curve illustrates a developmental cycle in which employee wages initially climb as workers harness intelligent technologies to become more effective; subsequently, those gains taper off, and eventually, as AI assumes a greater share of cognitive functions, wages begin to fall. According to Marinescu, for wages to actually decrease, a necessary precondition would be a reduction in the proportion of workers employed in intelligence-driven jobs relative to those in physical occupations. This erosion might unfold if, for instance, layoffs in technology-oriented sectors are not offset by equivalent expansions in other high-skill, intelligence-based roles, while employment in manual or service-oriented work remains static.
If such a structural adjustment takes root — for example, if software engineers, translators, or digital marketing writers are gradually supplanted by increasingly competent AI systems and no new categories of cognitively demanding work emerge to absorb displaced talent — the overall share of employment in the intelligence economy will shrink, triggering conditions conducive to a broad wage downturn.
Marinescu’s model further proposes that the initial waves of automation will target the most easily replicable forms of intellectual work — those functions where algorithms can quickly surpass human performance in pattern detection, linguistic translation, or straightforward written communication. In her analysis, these jobs, like basic translation or routine marketing copywriting, are most at risk in the near term. Should these positions disappear without the simultaneous creation of new AI-complementary cognitive roles, the collective proportion of intelligence-based employment will diminish, thereby setting the stage for stagnant or falling wages.
The tempo of this transformation depends, according to Marinescu, on how intertwined the physical and intelligence segments of the economy become. The more substitutable these sectors are — that is, the more easily AI systems can replace diverse forms of human input — the more rapid and severe the decline in compensation tends to be. She notes that workers concentrated in physical occupations will experience diminishing benefits from further automation if artificial intelligence can readily replicate their counterparts in the cognitive field. Similarly, within the intelligence economy itself, the degree of substitutability matters: when different cognitive tasks become less interchangeable, wages can continue to rise for longer, but once the decline begins, it occurs more sharply.
Looking ahead, Marinescu identifies one primary indicator that economists and policymakers should monitor closely: the relative balance of employment between the physical and intelligence-based sectors. In essence, this means tracking whether the workforce is migrating toward or away from occupations that artificial intelligence can easily perform. If current trends reveal an exodus from desk-based, cognitively driven positions toward embodied or manual forms of labor that historically grow more slowly, it may signal that the crest of the AI pay curve — the point where wage growth peaks before reversing — is already near. The outcome of this shift will determine not only the trajectory of future wages but also the broader adaptability of human labor in an increasingly automated world.
Sourse: https://www.businessinsider.com/ai-pay-boost-could-soon-hit-peak-research-2025-11
