New research examining Denmark’s labor market reveals that generative AI tools like ChatGPT have had virtually no measurable effect on wages or employment so far, challenging predictions of immediate workforce disruption.
Economists Anders Humlum and Emilie Vestergaard from the University of Chicago and University of Copenhagen analyzed data from 25,000 workers across 7,000 Danish workplaces during 2023-2024.
Their working paper, “Large Language Models, Small Labor Market Effects,” specifically targeted 11 occupations commonly considered vulnerable to AI automation, including software developers, accountants, and customer service representatives.
“We were surprised by how quickly these tools have been adopted,” Humlum told reporters. “Most workers in exposed occupations now use AI chatbots regularly, often with employer encouragement. But when we examined the economic outcomes, these tools haven’t moved the needle in any meaningful way.”
AI Adoption Modest Early Impact on Employment and Wages
The study’s statistical analysis confidently ruled out average employment or wage effects larger than 1 percent across all studied occupations, despite widespread AI adoption.
This finding comes amid ongoing debate about AI’s potential to transform workplaces. While many tech leaders and economists have warned about possible job displacement, this early data suggests a more nuanced reality where implementation challenges may be slowing economic impact.
The research revealed that corporate investment significantly boosted AI tool adoption, with 64 to 90 percent of users reporting time savings. However, these efficiency gains proved more modest than anticipated in practice, averaging just 2.8 percent of work hours approximately one hour per week.
Interestingly, AI chatbots actually created new tasks for 8.4 percent of workers, including those who didn’t directly use the tools. Teachers now spend additional time detecting
AI-generated homework, while other professionals devote hours to reviewing AI outputs or crafting effective prompts.
“There’s a learning curve with these technologies,” explained workplace technology analyst Maria Fernandez, who wasn’t involved in the study. “Organizations are still figuring out how to integrate AI effectively into existing workflows. The tools might save time on specific tasks but introduce new complexities elsewhere.”
AI Productivity vs. Worker Earnings: A Disconnect?
The researchers also discovered that only 3 to 7 percent of productivity gains translated into higher earnings for workers, raising questions about who ultimately benefits from any efficiency improvements.
This study contrasts with a February randomized controlled trial that found generative AI increased worker productivity by 15 percent. Humlum suggested this discrepancy likely stems from laboratory experiments focusing on ideal AI-suited tasks, whereas most real-world jobs involve complex responsibilities that AI cannot fully automate.
While providing valuable early insights, the Danish study has limitations. The 2023-2024 timeframe captures only the initial deployment phase of generative AI, potentially missing delayed effects or impacts from more sophisticated future implementations. Additionally, Denmark’s labor market has unique characteristics that might not reflect experiences in other countries.
Some experts believe certain sectors may already be feeling localized impacts not captured in the broader data. Freelance creative professionals, for instance, have reported increasing competition and downward price pressure since AI image and text generators became widely available.
“This research provides a helpful baseline, but we shouldn’t mistake it for the final verdict on AI’s economic impact,” noted labor economist Thomas Chen. “The technology is evolving rapidly, and its integration into workplaces remains in early stages. The true transformation may still be coming.”
As organizations continue experimenting with generative AI applications beyond simple chatbots, researchers will need ongoing studies to track potential shifts in employment patterns, wage distributions, and job quality across different sectors and regions.
For now, the evidence suggests that predictions of immediate workforce disruption from Artificial Intelligence may have been premature, with meaningful economic impacts requiring more time and deeper integration than initially anticipated.
