AI Exposure Linked to Slower Wage Growth Across Hundreds of Occupations

If the Apollo findings hold as more data arrives, AI's near-term labor effect may reshape compensation policy before it reshapes headcount.

A study of hundreds of U.S. occupations found that workers in roles with higher exposure to artificial intelligence experienced weaker wage growth compared to workers in lower-exposure roles, according to research published by Apollo Global Management Chief Economist Torsten Slok and reported by Bloomberg on August 22, 2026. The employment effect, by contrast, remained relatively small across the occupations studied. The research did not specify a precise publication date for the underlying dataset, and the full methodology has not been released publicly as of this writing.

Slok's findings arrive as federal policymakers and congressional committees are actively examining AI's effect on the U.S. labor market. The Senate Committee on Health, Education, Labor, and Pensions and the House Committee on Education and the Workforce have both held hearings in the current Congress on workforce automation, though neither committee has advanced binding legislation tied specifically to AI wage effects as of the congressional record through August 2026.

The Apollo study also identified a potential offsetting factor: Slok noted that AI is contributing to record business formation rates, which could generate new employment over time. The U.S. Census Bureau's Business Formation Statistics, which track weekly business applications, would be the public record most likely to confirm or refine that claim. The Census data for the second quarter of 2026 had not been fully published as of this story's publication date.

The wage compression dynamic, if confirmed at scale, would carry direct implications for federal tax revenue projections and Social Security funding calculations, both of which are sensitive to aggregate wage levels. The Congressional Budget Office has not yet released a formal economic outlook that incorporates AI-driven wage suppression as a modeled variable, according to CBO publications available through August 2026.

What remains unknown is whether the wage effect Slok identified is concentrated in specific sectors such as finance, legal services, or administrative support, or distributed broadly. The full occupational breakdown of the study has not been made public. A formal release of the underlying dataset by Apollo, or a parallel analysis by the Bureau of Labor Statistics using its Occupational Employment and Wage Statistics program, would be the records that could resolve that question.