AI chatbots are rapidly appearing in workplaces, fueling speculation about massive productivity gains and significant labor market disruption. However, a recent large-scale study offers a more grounded perspective, suggesting that despite widespread adoption, these tools have not yet led to major shifts in employee earnings or recorded work hours across various professions.
Based on a working paper from the National Bureau of Economic Research (NBER), the study examined AI chatbot use across 7,000 workplaces and surveyed 25,000 workers in Denmark – a country known for high tech adoption and detailed labor data. The core finding is clear: AI chatbot adoption has had “no significant impact on earnings or recorded hours in any occupation” during the study period of late 2023 and early 2024.
This finding holds true even for white-collar roles widely believed to be most susceptible to AI disruption, such as accountants, customer support specialists, financial advisors, HR professionals, IT support specialists, journalists, legal professionals, marketing professionals, office clerks, software developers, and teachers.
Why Isn’t AI Boosting Pay or Cutting Hours?
The study’s results challenge narratives of immediate, transformative economic effects. Researchers found several contributing factors explaining the modest impact:
Minimal Time Savings: While AI can save time on certain tasks, the average time saved across all occupations was quite small – roughly 3% of work hours, amounting to less than two hours per week for most.
Time Reallocation: Instead of leading to fewer hours worked or more leisure, workers primarily used the saved time for other work tasks (over 80%). This includes reinvesting time into the task they were using AI for (around 25% of workers) to refine output or conduct additional checks.
Creation of New Tasks: Significantly, AI use created new tasks for 8.4% of workers surveyed. This includes work like fact-checking AI-generated content, editing outputs, ensuring compliance with professional standards (like journalistic ethics), and even detecting AI use by others (like teachers checking student work). Any time saved by AI is often offset by these new responsibilities, leading to little net change in overall hours.
Limited Pay Reflection: Even where minor productivity gains occurred, only a small fraction (between 3% and 7%) translated into higher pay for workers.
Real-World Complexity: The messy reality of workplaces differs from controlled experiments. Much AI use is informal, lacking formal employer training or policy (#shadowAI). Without clear guidance or systems in place, employees may not feel empowered to leverage AI for maximum efficiency or to negotiate for higher pay based on AI-enhanced output. Some may even be reluctant to highlight their increased productivity, fearing they would simply be assigned more work without commensurate compensation increases.
Need for Deeper Change: Experts suggest that realizing substantial economic gains from AI requires more than just adopting the tool. It necessitates significant “organizational adjustment, a range of complementary investments, and improvements in worker skills, via training and on-the-job learning.” The study found greater productivity gains in workplaces where employers actively encouraged AI use and provided training.
Specific Sector Insights: Journalism
Journalists, often cited as being on the front lines of AI adoption, reflect the broader trend. The study found journalists to be early and active AI users, often with employer encouragement and access to training and in-house tools. They utilize AI for tasks like brainstorming, drafting outlines, summarizing research, and writing breaking news reports.
Yet, despite this high engagement, journalists also reported no significant change in earnings or hours. Their time savings were minimal (often just a few minutes per day) and primarily reinvested in other tasks or refining AI outputs. Notably, only a small percentage (12.6%) reported increased job satisfaction from using AI, significantly lower than in fields like software development (30.5%). This suggests integration is happening, but without necessarily improving the overall work experience or leading to major labor market shifts yet.
Beyond Productivity: New Tasks & Work Dynamics
Some critics argue that focusing solely on earnings and hours might miss the true impact of AI. They draw parallels to past technologies like educational computers, where early studies focused on narrow metrics like test scores and missed deeper changes in teaching methods, thinking skills, and student engagement. Similarly, AI might be changing how work is done, altering workflows, requiring new skills (like AI prompting and output review), and potentially affecting work intensity or creating new tasks even for non-users.
Concerns have also been raised about employer responsibility when AI adoption is merely “encouraged” without formal policy or training. This informal adoption, or #shadowAI, may negatively impact work intensity and the working environment if not properly managed, potentially allowing employers to avoid full responsibility for the tool’s effects on occupational health and safety or overall working conditions.
What This Means for the Future
This study adds to a growing body of evidence tempering the most extreme predictions about AI’s immediate effects on the labor market. It aligns with other recent indicators, such as companies like Klarna adjusting their rhetoric on AI-driven layoffs and survey data showing many AI projects failing to deliver promised ROI.
The research doesn’t mean AI won’t have a transformative effect, but it suggests that like past technological revolutions (e.g., the steam engine’s impact unfolded over decades), its full economic and labor market consequences may take considerable time to materialize and will likely require complex adaptations beyond simple tool adoption. For now, AI appears to be augmenting work in small ways, creating new tasks that offset potential time savings, and its benefits are not yet broadly translating into higher pay or fewer hours for the average worker.
References
- https://finance.yahoo.com/news/study-looking-ai-chatbots-7-110300796.html
- https://www.linkedin.com/posts/christinajcolclough_time-saved-by-ai-offset-by-new-work-created-activity-7324449627489861632-0ORb
- https://pressgazette.co.uk/news/study-finds-journalists-are-on-ai-frontline-and-yet-to-be-replaced-by-tech/
- https://www.linkedin.com/posts/jerrycrisci_study-looking-at-ai-chatbots-in-7000-workplaces-activity-7330247445861339137-pPVw
- https://bgr.com/tech/study-finds-chatgpt-isnt-taking-anyones-job-at-least-not-yet/
