AI is making workers more productive. Their paychecks aren't keeping up

A junior associate at a law firm works faster now. With AI handling document review and legal research that once took days, she bills more hours than she did two years ago. But her salary hasn't changed. Maybe her bonus was even cut.

That dynamic — productivity rising, compensation stagnant or falling — sits at the center of a growing debate about who actually benefits when AI makes workers more efficient. The answer increasingly appears to be: not the workers.

The bargaining power problem

The gap between productivity and compensation is not new. According to the Economic Policy Institute's Productivity-Pay Tracker, productivity has grown roughly eight times faster than typical worker pay since 1979 — a divergence driven by a steady erosion of worker bargaining power over four decades.

But AI may be accelerating the dynamic for a specific reason. When an employer can substitute AI for a worker's labor, that worker's ability to negotiate weakens. Columbia Business School professor Daniel Keum recently described the mechanism to Quartz: "A junior associate at a law firm before could demand 20% of billable hours. Now you bill more, but you take 10% — because if you demand anything more, there's AI."

What researchers are finding

The research on AI and compensation is early but consistent. A 2023 working paper by Erik Brynjolfsson, Danielle Li, and Lindsey Raymond of MIT and Stanford, published by the National Bureau of Economic Research, studied AI deployment at a Fortune 500 company and found that AI tools increased worker productivity by 14%. The gains were largest for less experienced workers. Whether those productivity gains translated into higher pay remained outside the study's scope, pointing to a gap in the research that economists say needs urgent attention.

Aaron Terrazas, former chief economist at Glassdoor, has described three ways companies quietly reduce compensation without lowering base salaries: Benefits packages shrink, bonuses get cut, and job responsibilities expand without corresponding pay increases. He calls the last of these "shrinkflation" — borrowing the consumer-pricing term for when a product gets smaller without the price changing.

The structural explanation

Behind the individual cases is a longer-term shift. According to Bureau of Labor Statistics data tracked by the Federal Reserve, labor's share of nonfarm business income has been declining for decades, falling from about 64% to roughly 56% in recent years.

The concern among economists is that AI represents not just another chapter in that decline but a structural inflection point. Previous automation targeted physical tasks, leaving intact the cognitive premium that white-collar workers earned. AI targets that premium directly.

Whether workers ultimately capture a share of the productivity gains AI enables remains an open question. The historical pattern suggests they eventually do. The present moment suggests the "eventually" may be doing a lot of work.