A recent white paper by Apollo Global Management, an asset manager, reveals that AI's impact on the U.S. labor market is primarily through wage compression rather than job displacement. The study, which tracked wage and employment data for 321 occupations from 2015 to 2025, found no detectable effects on overall employment. Instead, jobs with high exposure to AI experienced an average 6.7% decline in real wage growth after 2023, coinciding with the rise of tools like ChatGPT.
This wage compression was most pronounced among the lowest earners. Service workers saw an average 24.3% decline in earnings growth since 2023, while workers in the bottom 25% of earners experienced a 10.7% decline in real wages over the same period. Conversely, top earners showed no significant statistical effect, suggesting they are better positioned to absorb or even benefit from AI adoption. This indicates that firms are capturing AI productivity gains by holding compensation growth flat rather than reducing their workforce.
The research estimates that 5.8 million U.S. workers, representing about 3.7% of the labor force, are currently in high-exposure occupations feeling these effects. This number is considered a conservative lower bound, as AI adoption is still in its early stages and expected to deepen. The findings highlight a shift in how AI is impacting the workforce, moving the discussion from job cuts to issues of pay equity and income inequality, particularly for the most economically vulnerable workers.
