Table of Contents
Despite rapid adoption, generative AI chatbots like ChatGPT are showing minimal impact on jobs and wages, challenging narratives of imminent disruption and raising questions about massive tech investments.
Key Takeaways
- A study of 25,000 Danish workers found no significant generative AI impact on earnings or hours worked in vulnerable occupations during 2023-2024.
- Many professions deemed susceptible to AI (e.g., accountants, developers, journalists) have not experienced notable labor market shifts due to chatbots.
- While AI tool adoption is high, often encouraged by employers, it hasn't translated into measurable economic benefits like higher wages or increased work hours.
- Users report minimal actual time savings, averaging just 2.8% of work hours, far less than figures seen in controlled studies.
- AI implementation creates new tasks, such as quality review and prompt engineering, partially offsetting potential productivity gains from automation.
- Only a small fraction (3-7%) of any productivity benefits achieved through AI seems to be passed on to workers as increased earnings.
- The lack of tangible economic returns challenges the justification for the enormous capital expenditure poured into developing and running AI models.
The Danish Study: AI Chatbots Show Minimal Job Market Impact
- Economists Anders Humlum and Emilie Vestergaard conducted a detailed analysis of the Danish labor market, focusing on the real-world effects of generative AI chatbots in 2023 and 2024. Their research covered 25,000 workers across 7,000 workplaces.
- The study specifically targeted 11 occupations previously identified as vulnerable to AI disruption, including:
- Accountants
- Customer support specialists
- Financial advisors
- HR professionals
- IT support specialists
- Journalists
- Legal professionals
- Marketing professionals
- Office clerks
- Software developers
- Teachers
- Despite widespread fears and predictions, the core finding was stark: "AI chatbots have had no significant impact on earnings or recorded hours in any occupation," according to the researchers. This contradicts prevalent narratives about AI displacing workers or suppressing wages.
- This lack of economic effect occurred even as adoption rates soared, suggesting that simply using the tools doesn't automatically lead to the disruptive outcomes often predicted for the labor market.
Adoption Soars, But Economic Needle Barely Moves
- The research highlights a significant disconnect between the rapid uptake of generative AI tools by workers and the absence of corresponding economic shifts. Humlum noted, "The adoption of these chatbots has been remarkably fast," with most workers in exposed fields now using them, often with employer encouragement.
- However, this widespread adoption hasn't resulted in tangible economic benefits for workers or significantly altered labor market dynamics. "When we look at the economic outcomes, it really has not moved the needle," Humlum stated.
- While company investments in AI did encourage worker adoption, the downstream effects on work quality and satisfaction were mixed, indicating that usage doesn't always equate to positive outcomes.
- The findings stand in contrast to some randomized controlled trials (RCTs) that reported productivity gains around 15%. The economists suggest this discrepancy arises because RCTs often focus on tasks with high automation potential and don't reflect the complexities and varied tasks of real-world jobs.
The Productivity Puzzle: Time Saved vs. New Tasks Created
- A key reason for the muted economic impact appears to be that AI usage generates new responsibilities that consume time, counteracting some of the efficiency gains. The study found that AI chatbots created new job tasks for 8.4% of workers, even impacting some who don't use the tools directly.
- Examples of these new tasks include educators spending time detecting AI-generated cheating, or employees dedicating hours to reviewing AI output quality and crafting effective prompts. This phenomenon echoes historical patterns where automation creates new forms of work.
- Consequently, the actual time saved by using AI tools is surprisingly low in practice. The study reported an average saving of just 2.8% of work hours per user, equating to slightly over one hour in a standard 40-hour week.
- Humlum explained the gap between experimental results and real-world data by noting that most job tasks aren't fully automatable by tools like ChatGPT, and organizations are still figuring out how to integrate these tools effectively to realize their full potential.
Investment vs. Return: Who Captures the AI Dividend?
- Even where productivity gains are achieved through AI, the study estimates that only a minor portion—between 3% and 7%—translates into higher earnings for workers. This raises crucial questions about where the value generated by AI actually accrues.
- Potential beneficiaries could be the firms themselves through increased profitability (which the study couldn't directly measure), or the gains might be diffused, simply allowing workers to complete existing tasks slightly faster without enabling them to take on significantly more valuable work.
- The minimal impact on earnings and hours challenges the massive financial investments made by tech companies in AI infrastructure. With companies like OpenAI reportedly losing money per query and others potentially scaling back AI spending due to slow business adoption, the economic justification is under scrutiny.
- Humlum emphasized that while company initiatives can improve AI tool usage (e.g., closing the gender gap in adoption), the current data shows these efforts haven't yet yielded significant economic returns. "Any story that you want to tell about these tools being very transformative, needs to contend with the fact that... they've not made a difference for economic outcomes," he concluded.
Two years post-introduction, generative AI chatbots have not significantly altered wages or employment levels, despite rapid adoption. The current real-world data suggests the transformative economic impact hyped by the tech industry remains largely unrealized for workers and the broader economy.
Sources consulted: The Register
