AI chatbots have moved quickly into everyday office life, but a new study suggests the labor market has not shifted nearly as fast. Based on survey data from Danish employees and linked government records, the research finds broad use of tools like ChatGPT, modest time savings, and little evidence so far of meaningful changes in wages, working hours, or retention.
What the study measured
The research comes from the National Bureau of Economic Research, a US nonprofit. It uses survey data from about 25,000 Danish employees across 7,000 workplaces in 2023 and 2024.
The study looked at eleven occupations considered likely to be affected by generative AI. The list included accounting, customer service, financial consulting, IT support, journalism, law, marketing, software development, and teaching.
A key strength of the work is that employee survey responses were linked to government records on wages, working hours, and employment status. That allowed the researchers to compare what workers said about AI chatbot use with labor market outcomes recorded outside the survey itself.
The study compared wages and working hours before and after ChatGPT's introduction in November 2022. It then looked at whether employees who used AI chatbots differed meaningfully from those who did not.
Chatbot use is widespread, but gains are modest
The study found that AI chatbots are already normal tools for many workers. In the survey, 64% of respondents said they had previously used these tools at work.
Employer support appears to matter. In companies that actively support AI adoption, usage rose to 83%. About 38% of companies use their own chatbots, often adapted for their needs, and 30% of employees have taken part in AI-related training.
Those workplace efforts also seem to reduce some gaps in adoption. The difference between male and female users fell from 12 to 5 percentage points when companies promoted AI use.
Still, the measured productivity effect was limited. On average, employees saved 2.8% of their working time with AI chatbots. That is a real change in daily work, but far from the sweeping labor market disruption many observers expected from generative AI.
Wages and hours have not clearly changed
The central finding is straightforward: the study did not detect a meaningful difference in wages or working hours between workers who used AI chatbots and those who did not. Heavy use and targeted employer support did not change that overall result.
The statistical analysis rules out average wage effects larger than one percent. At the company level, the study also found no detectable impact on total wages, working hours, or employee retention.
One explanation offered in the source is that productivity gains from AI are not usually passed on to employees. Only 3 to 7 percent of the time saved appears as higher wages.
That distinction matters. A tool can help a worker finish some tasks faster without immediately changing the worker's pay, schedule, or job status. The study suggests that, so far, AI chatbot adoption has been more visible in work routines than in traditional labor market indicators.
New tasks are appearing around AI
The findings do not imply that nothing is changing. Many workers said AI chatbots help them be more creative or improve work quality. The study also found signs that new responsibilities are forming around the technology.
Seventeen percent of users reported taking on additional responsibilities, often connected to integrating or reviewing AI systems. Even among non-users, about five percent said they had taken on new tasks, such as reviewing AI-generated content.
These findings point to a practical shift: AI chatbots may create extra coordination, checking, and integration work even when they save time elsewhere. That can make the overall labor market effect harder to see through wages and hours alone.
For employers, the study suggests that adoption, training, and internal chatbot tools can increase usage. But increased usage does not automatically translate into higher pay, shorter working hours, or measurable company-level retention effects.
Why the results may still be early
The study includes important measurement limits. To estimate time savings, it relied on employees' own estimates of how much time they saved on days when they used AI chatbots at work. Those self-reported savings were combined with reported usage frequency to calculate average time saved as a share of total working hours.
The study did not track time spent on specific tasks. That means it could not directly measure how long a defined task took before and after chatbot use. It also did not examine whether workers might underreport or misreport time savings because of concerns about consequences at work.
The authors also point to the idea of a "Productivity J-Curve." Under that view, the larger economic benefits of a new technology may appear only after companies adjust their processes to use it fully.
For now, the clearest message is measured caution. AI chatbots like ChatGPT are widely used, especially where companies encourage adoption. They save some time and are changing some job tasks. But in this study, the big labor market effects on wages, employment, working hours, and retention have not yet shown up.