Meh. Generative AI in the workplace is delivering modest returns that don’t match the hype

A new study by the US-based National Bureau of Economic Research has cast doubt on the immediate transformative impact of artificial intelligence in the workplace, finding that the use of generative AI chatbots has so far delivered only modest gains in productivity and almost no increase in pay or reductions in working hours for employees. The working paper [restricted access], authored by economists Anders Humlum and Emilie Vestergaard, used detailed Danish employment data to assess the real-world effects of AI adoption across 7,000 workplaces and 25,000 workers, focusing on white-collar roles most susceptible to automation—such as accountants, IT support staff, journalists, HR professionals, and software developers.

Despite the rapid adoption of AI tools like ChatGPT—heralded as a revolution on the scale of the personal computer—the researchers found that AI users saved on average just 3 percent of their time. Only a small fraction of these gains translated into higher wages, with employees receiving just 3 percent –7 percent of productivity improvements in the form of pay rises. “AI chatbots have had no significant impact on earnings or recorded hours in any occupation,” the paper concludes.

The findings contrast sharply with the widespread narrative of AI’s imminent disruption of work. Companies such as Duolingo and Klarna have attracted headlines for replacing or side-lining human roles in favour of AI, and investment in AI-linked technologies has boomed. Yet in practice, most workers have seen little change to their day-to-day realities.

According to the study, most employees redirected saved time toward other work tasks rather than enjoying more leisure or rest. In many cases, new tasks emerged from AI use itself—such as editing chatbot-generated content or adapting workflows to manage AI integration. For example, Humlum noted having to redesign university exams to prevent cheating via AI.

Crucially, much of the generative AI use observed in the study was informal, with workers often adopting the tools without employer endorsement or clear guidelines—limiting the broader organisational benefits. “In a workplace where it’s not explicitly encouraged, there’s limited space to go to your boss and say, ‘I’d like to take on more work because AI has made me more productive,’” Humlum noted.

The potential for generative AI is still there

The Danish data offered a rare opportunity to match self-reported AI use to anonymised employment records, thanks to the country’s robust record-keeping. Denmark’s labour market, which shares characteristics with the US and UK in terms of white-collar employment and tech adoption, provided a representative setting for examining early-stage AI integration.

The study’s findings suggest that earlier research highlighting AI’s productivity boost may have been overly narrow—focused on a handful of AI-friendly tasks like writing code or marketing content—without capturing its broader effects across diverse jobs.

The muted impact also aligns with recent caution from business leaders. A 2024 IBM survey of 2,000 global CEOs found that only a quarter of AI projects deliver on their expected return on investment. Corporate fear of missing out (FOMO) was a stronger driver of adoption than proven value, the survey indicated.

Economist and Nobel laureate Daron Acemoglu has also warned against inflated expectations, estimating that AI will raise GDP by only 1.1% to 1.6% over the next decade—modest by historical standards. The real benefits, he argues, will require long-term investment in organisational change, employee training, and complementary technologies.

While the findings may disappoint AI evangelists, the researchers caution against dismissing AI’s long-term potential. Just as the steam engine took decades to reshape industrial production, AI may eventually prove more transformative—particularly if firms invest in training and integrate the technology more deeply into workflows.

“There are signs that with employer encouragement and structured implementation, AI use does produce more noticeable productivity gains,” the authors note.

For now, however, the great AI revolution in the workplace looks more like an incremental shift than a sweeping transformation—suggesting that, for most workers, the chatbot revolution is still waiting to happen.