Is AI going to disrupt the labor market?  Researchers think it has already started to do so.

Harvard economists David Deming and Lawrence H. Summers along with Kennedy School predoctoral fellow Christopher Ong have presented a new paper that looked over 100 years of “occupational churn” for a study of technological disruption.  “Occupational churn” refers to each profession’s share in the U.S. labor market which Deming and Summers have always been interested in gauging.  With the help of Ong, they applied the metric to 124 years of U.S. Census data, initially sharing their findings in a volume published last fall by the Aspen Economic Strategy Group.

So are robots going to take over human jobs?  This sentiment has always been present, and for good cause, whenever breakthrough technologies such as keyboards, electricity, and computer-based manufacturing emerge.  The 1950s, ’60s, and ’70 demonstrated volatility that surprised but eventually made sense to Summers while Deming characterized the 2000s and 2010s with having “automation anxiety.”  The study however revealed that the labor market has enjoyed stability and low churn between 1990 and 2017 when the pace of disruption slowed.

Copyright: Kindel Media

But for 2019 onwards, it appears that there’s a major change set to happen with four labor market trends they believe pointing towards AI as the new breakthrough technology.

1. High-paying jobs are on the rise

The first trend sees job polarization being replaced by general skill upgrading.  Job polarization refers to increased employment opportunities in the high and low-skill occupations while middle-skill jobs go through a relative decline.  Extending across multiple decades and various economic states, this phenomenon has been observed to be influenced by technological shifts such manufacturing automation and the widespread adoption of office software.

However between 2016 and 2022, the report noted that low and middle-skill jobs have both declined while high-skilled, high-paying jobs have slightly increased.  The report adds that data collected through 2024 share similar results, denoting the end of polarization as of 2016 and the start of a trend towards skill upgrading.

2. Low-paying service work employment is flat or in decline

Job polarization during the 2000s was seen as a result of middle-skill production jobs being replaced by low-paid service work.  Low-skill jobs enjoyed robust growth during the 2000s but slowed down in the early 2010s and was flat throughout the rest of the decade, falling rapidly in 2020 when the COVID-19 pandemic happened.  While low-paying occupations have partly recovered in 2024, most service sector employment is back to the same level they started before their rapid growth back in the 2000s.

The decline in low-paying service work can’t just be pinned on the emergence of AI alone, however, as other factors to consider include the aforementioned pandemic disruption, increasing wages, and a tighter labor market.  

3. STEM occupations are on the rise

After having a decline in the 2000s, STEM (science, technology, engineering, and math) jobs are now enjoying rapid employment growth from 6.5 percent in 2010 to nearly 10 percent in 2024.  This growth also extends to business and management occupations such as science and engineering managers, management analysts, and other business operations specialists.

Firms have also increased investments in AI-related technologies to match the rising number of technical talents they’re hiring and developing.  Mostly driven by the need for more computing power, software and information processing investments are now above 4 percent once again, the same level they were prior to the dot-com bubble burst and the 2001 recession, while research and development spending as a share of GDP have now reached a record high of 2.9 percent.

4. Retail sales jobs are in decline

Even before the pandemic, retail sales occupation has been on the decline.  Retail sales dropped by 850,000 jobs between 2013 and 2023, which translates to a decrease from 7.5 to 5.7 percent share of employment and converts into a 25 percent reduction of share in the job market in just a decade.

This is being seen as an effect of online retail or e-commerce’s early adoption of predictive AI models around the mid-2010s to generate personal recommendations based on customers’ browsing and buying histories along with predicting local product needs for stocking warehouses.  Online retail has more than doubled its share of all retail sales at 15.6 percent from 7 percent in 2015.

Labor productivity growth in retail trade went up in the same period that retail sales occupation declined, mimicking what happened with manufacturing production jobs 50 years prior.  Online retail’s demand for light delivery service truck drivers for their last-mile package delivery and “stockers and order fillers” in their large warehouses resulted in employment growth in these occupations.

Copyright: u_fg0tkeqgiy

Conclusion

Is AI going to replace you at work?  Looking at these four trends, the answer is going to depend on what you do for a living.

Summers acquiesced how “highly empowering” AI can be that it might lead to “certain types of activities won’t be done by people anymore.”  Data exists to corroborate that automation claims jobs with Deming citing early 20th-century telephone operators in a Substack post.  The study notes sales and administrative-support occupations possibly experiencing future declines in employment as AI innovates and improves on tasks relating to these jobs- personalized pricing algorithms and product recommendations, inventory management, oral and written transcription, and automated scheduling, just to name a few.

As AI is being utilized more and more to boost productivity, there are some jobs and tasks it might not be as effective as human knowledge workers.  The technology would exist but human output would be much more valuable that rather than replace knowledge workers, companies start increasing expectations from their human workforce.  The study closes by recommending investing too in STEM education if this is the case, along with training and reskilling workers to help with their adaptability and effectivity with these new technologies.

Additional Reference Article: Is AI already shaking up labor market? – The Harvard Gazette, Christy DeSmith (February 14, 2025)

Featured Image Copyright: Frank_Rietsch

Top Image Copyright: ThisIsEngineering

Read more