AI Chatbots Yet To Prove Their Worth

New study finds little effect on pay or work hours.

A new study has found that AI chatbots are having a surprisingly minor impact on employee earnings and work hours.

The working paper, released this week by the US-based National Bureau of Economic Research (NBER), analyzed data from 25,000 workers across 7,000 Danish workplaces and found "no significant impact on earnings or recorded hours in any occupation" linked to AI chatbot use.

Economists Anders Humlum of the University of Chicago Booth School of Business and Emilie Vestergaard, Ph.D. candidate at the University of Copenhagen, conducted the study.

They focused on professions commonly considered vulnerable to AI disruption, such as accountants, software developers, journalists, customer support specialists, financial advisors, legal and marketing professionals, office clerks, teachers and HR specialists, but found that neither wages nor hours shifted noticeably with AI adoption.

On average, workers who used AI tools reported saving about 3 percent of their time. However, only a fraction of that productivity gain, between 3 and 7 percent, translated into increased pay.

AI is seen as both a revolutionary force and a disruptive threat.

Since OpenAI introduced ChatGPT just over two years ago, companies around the world have embraced gen-AI technology, reshaping job descriptions and fueling wild swings in tech sector valuations.

High-profile firms like Duolingo and Shopify have made headlines by reducing human hiring in favor of AI integration. But Humlum and Vestergaard's research paints a more measured picture of AI's effect on everyday work.

"While adoption has been rapid, with firms now heavily invested in unlocking the technological potential, the economic impacts remain small," they wrote.

The study's methodology, uniquely enabled by Denmark's detailed employment records, allowed researchers to cross-reference AI usage surveys with real-world data on hours worked and wages earned.

The country's labor market shares key characteristics with the U.S., including relatively high AI adoption and similar employment dynamics, but boasts more precise tracking of employment outcomes.

No job category studied showed a statistically significant shift in pay or hours. Time savings from AI averaged 3 percent, often reallocated to other tasks, not leisure.

More than 80 percent of study participants reported redirecting saved time to other responsibilities, including tasks newly created by AI use, such as reviewing AI-generated output or adapting processes to prevent misuse.

Moreover, only a small portion of the productivity gain was returned to employees in the form of pay.

Workplaces that offered explicit encouragement and training around AI use saw slightly higher gains.

Companies investing heavily in AI technologies have promised massive productivity improvements, and some early-stage studies focused on specific tasks, such as writing marketing copy or code, have shown substantial time savings.

But Humlum cautions that these studies often focus on AI's best-case scenarios.

"In a broader occupational survey, where AI can still be helpful, we see much smaller savings," he said.

The study joins a growing body of evidence suggesting that AI's real-world impact has yet to catch up with its hype.

A recent IBM survey of 2,000 CEOs found that only one in four AI projects met their expected return on investment. The primary adoption driver was fear of falling behind competitors - not clear business value.

This article originally appeared on our sister site Computing.