Skip to research
Opening / Research

Analysis / Pay & conditions

Higher productivity is only the beginning of the worker’s story

AI can improve output in a defined workplace. Whether workers receive better pay, shorter hours or greater control is a separate question that reporting must answer.

By Opening Research · Published September 9, 2026
Evidence checked through September 9, 2026 · Editorial and AI disclosure

Output is one outcome

A customer-support study published in the Quarterly Journal of Economics finds that access to an AI assistant increased productivity, with particularly large gains for less experienced workers. The setting provides useful evidence about how a tool changes performance at work. It does not establish how gains are distributed across the economy.

A productivity metric alone cannot tell a worker whether staffing will fall, targets will rise, hours will shorten or pay will improve. Those outcomes need their own measurements. The connection between technical improvement and worker welfare is a question for evidence, not an automatic consequence.

Sources: Erik Brynjolfsson, Danielle Li and Lindsey R. Raymond, Quarterly Journal of Economics 140(2), 889–942

Contracts can specify the terms

The Writers Guild of America’s guidance on its 2023 agreement describes rules governing AI-generated material, disclosure and whether writers can be required to use AI. These provisions demonstrate that collective bargaining can address how a technology enters a workplace. They describe a particular agreement, not protections for every worker or a guarantee about subsequent employment.

The reporting task is to follow implementation: what employers disclose, which work is covered, how disputes are resolved and what workers can contest. A contract announcement is the beginning of an accountability story.

Sources: Writers Guild of America

Measure who receives the gain

A serious workplace study should follow output alongside pay, paid hours, unpaid work, staffing, training and worker discretion. It should distinguish an employer’s forecast from observed results, and a workforce average from the experience of a vulnerable group.

This is a worker-first editorial choice: the consequences for the people doing the work belong in the main result. It does not predetermine whether a particular tool helps or harms. A system that reduces drudgery deserves that finding; a system that increases pace while shifting risk to workers deserves an equally clear account.

What accountability requires

Company claims should specify the baseline, period, sample and outcome. Worker interviews can reveal unmeasured burdens and help test whether official metrics describe everyday work. Interviews should be attributed or protected under a stated source agreement, and they should not be converted into unsupported prevalence estimates.

Opening’s conclusion is that productivity should remain part of the record while distribution becomes a standing question. The public interest lies in understanding both what technology makes possible and who gets to decide what follows.

Scope and limits

Analysis of published research and historical contract guidance. No original worker interviews were conducted for this article, and no current legal protection is asserted beyond the cited historical agreement.

Sources

  1. Erik Brynjolfsson, Danielle Li and Lindsey R. Raymond, Quarterly Journal of Economics 140(2), 889–942. Generative AI at Work. 2025 published article.
  2. Writers Guild of America. Artificial Intelligence: 2023 MBA provisions. Historical contract guidance; accessed September 9, 2026.

Cite this article

Opening Research. “Higher productivity is only the beginning of the worker’s story” September 9, 2026. https://opening.works/research/productivity-worker-gains. Include your access date. Cite the original study when using its estimates.

Suggest a correction or send evidence. Reuse terms.