AI Will Reallocate Labor to Jobs That Require New Skills
By Khia Kurtenbach and Leo Bruk
Throughout history, new technologies have emerged, replacing the tasks that humans previously completed. But without fail, the labor force adapts with humans complementing emergent technologies with new tasks.
New research out of both the US and Europe shows that this is exactly what is playing out across the labor force as artificial intelligence (AI) proliferates through the economy. So far, AI is having a greater impact on workers’ tasks than on overall employment levels.
Looking ahead, AI technologies are likely to affect more and more workers, including high-skill workers. Because AI can replace humans across non-routine tasks, AI may be more disruptive and may impact many more workers compared to past technological inventions.
To cushion the disruptive impacts of AI, workers will need to upskill to jobs that are not easily substitutable by AI. Proactive policy that aims to design and invest in such programs will be key. The businesses that promote on-the-job training and upskilling, or tailor their talent acquisition strategies to skills that complement AI technologies, will be able to take the greatest advantage of AI’s productivity-enhancing benefits.
New research out of both the US and Europe shows that this is exactly what is playing out across the labor force as artificial intelligence (AI) proliferates through the economy. So far, AI is having a greater impact on workers’ tasks than on overall employment levels.
Looking ahead, AI technologies are likely to affect more and more workers, including high-skill workers. Because AI can replace humans across non-routine tasks, AI may be more disruptive and may impact many more workers compared to past technological inventions.
To cushion the disruptive impacts of AI, workers will need to upskill to jobs that are not easily substitutable by AI. Proactive policy that aims to design and invest in such programs will be key. The businesses that promote on-the-job training and upskilling, or tailor their talent acquisition strategies to skills that complement AI technologies, will be able to take the greatest advantage of AI’s productivity-enhancing benefits.
Studies show AI is having a meaningful impact on the labor market, but worker tasks are impacted more than overall employment.
In the wake of recent advances in AI, evidence for how it is disrupting the labor force is relatively scarce. But labor economists are racing to study data and produce research that gives us the best understanding of the disruptions so far. Two recent studies – one that used a large-scale business survey by the U.S. Census Bureau and another that used patent data and administrative social security data in Germany – looked at the impact of AI both on overall employment and on worker tasks. Both studies showed that so far, AI is not destroying many jobs, even in AI-exposed industries. But both studies also showed that AI is disrupting the labor market in other ways – reshaping the tasks workers complete and resulting in more worker mobility between firms.
Starting with the U.S. Census Bureau study – From December 2023 through February 2024, researchers collected data from over 160,000 businesses on AI use. The researchers also asked firms about how AI was used, how it would impact employment, and about the tasks that AI is replacing.
Of businesses using AI, some 87% reported no net change in their employment in the last six months attributable to AI use. And firms that experienced a decline in employment were a small fraction of firms – just 3%. Even looking ahead, the vast majority of firms – 87% – expected no net change in their employment levels due to AI.
Why might AI-use have a limited impact on net employment? Well, although AI can replace some workers’ tasks, it can also generate new tasks. It may also be that AI is displacing tasks but not workers – the study shed light on this question as well, with most firms reporting using AI to replace either a small or moderate number of tasks. It’s not surprising that workers weren’t fully being replaced, as AI was not eliminating a large number of tasks.
Firms Using AI: Reported Net Employment Impact and Number of Tasks Replaced by AI
| TASKS REPLACED BY AI | ||
|---|---|---|
| Last 6 Months | Next 6 Months (Expected) | |
| Small number | 85% | 79% |
| Medium number | 13% | 18% |
| Large number | 2% | 3% |
| NET CHANGE IN EMPLOYMENT | ||
|---|---|---|
| Last 6 Months | Next 6 Months (expected) | |
| Increase | 3% | 7% |
| Decrease | 3% | 6% |
| No Change | 95% | 87% |
Now turning to the study on European labor markets – Germany’s IZA Institute of Labor Economics looked at administrative social security data, industry-level technology measures, and individual survey data to assess changes in worker tasks, roles, and impacts on the labor market depending on exposure to AI or robotics technologies. This study arrived at similar conclusions to the U.S. Census Bureau study.
AI has had a measurable, but not extreme, influence on the market. Figure 1 shows the polarity in how AI and robots have impacted employment. For “one standard deviation” of AI exposure in a given industry, employment declined by 5% (based on cumulative worker days of employment). In contrast, researchers found no displacement effects from robots. In other words, while there is evidence that AI automation resulted in some job losses, the overall impact on employment has been marginal thus far. Notably, researchers can say with 95% confidence that AI had a negative impact on workers without a college degree, which was not true for robots.
AI has had a measurable, but not extreme, influence on the market. Figure 1 shows the polarity in how AI and robots have impacted employment. For “one standard deviation” of AI exposure in a given industry, employment declined by 5% (based on cumulative worker days of employment). In contrast, researchers found no displacement effects from robots. In other words, while there is evidence that AI automation resulted in some job losses, the overall impact on employment has been marginal thus far. Notably, researchers can say with 95% confidence that AI had a negative impact on workers without a college degree, which was not true for robots.
The study found evidence of AI’s imprint on the labor market beyond pure displacement. Notably, there was significant worker reallocation in AI-exposed industries. For example, some workers moved to new companies within the same industry, while others moved to different industries that were less exposed to AI.
Figure 2 summarizes the researchers’ results on worker reallocation with respect to AI and robot industry exposure. Researchers analyze whether a subject (employee of a company) moves to another company, occupation, or industry because of this change. For example, the first bar denotes a one standard deviation increase in AI exposure rates correlates with a 1.4% increase in the probability of moving to a different firm. The most probable outcome as AI spreads around the market is that of workers moving to a lower AI exposed industry.
Figure 2 summarizes the researchers’ results on worker reallocation with respect to AI and robot industry exposure. Researchers analyze whether a subject (employee of a company) moves to another company, occupation, or industry because of this change. For example, the first bar denotes a one standard deviation increase in AI exposure rates correlates with a 1.4% increase in the probability of moving to a different firm. The most probable outcome as AI spreads around the market is that of workers moving to a lower AI exposed industry.
The IZA study also examined changes in specific tasks due to AI exposure. Unsurprisingly, the most exposed tasks are those that are routine and repeatable and do not require physical labor. An AI model would have no problem, say, filling out a spreadsheet, but it would struggle moving boxes. However, it may come as a surprise that gathering information has the lowest exposure among tasks given AI’s web search capabilities. On average, the tasks least impacted are non-routine and physical, while the jobs bearing the brunt of this technological change are routine digital tasks.
Business and Policymaker Takeaways
Even if AI does not altogether replace workers, the disruptions and shifts coming to the labor market will be significant. Studies already show that AI technologies will change the tasks workers complete as these innovations shift the assignment of workers from routine to non-routine tasks. These changes will increase efficiency and result in more productive use of human capital. But at the same time, society will need policies (such as re-training programs) to cushion the disruptions and help labor develop a competitive advantage, or else workers will be left behind. Businesses will also need to invest in developing new human capital through upskilling or retooling their talent acquisition in order to take full advantage of these new efficiencies.
