Is AI Stealing Jobs?
The Answer Is More Complicated Than You Might Think
“It’s kind of like a nepo baby intern.” Jeremy Merrill, data reporter at the Washington Post, grins at his own joke. It’s an early July morning in Morningside Heights, where Merrill is teaching a class on AI. Outside the windows of the classroom, security guards pace the lawn outside of Columbia University’s Butler Library. Up, down. Up.
Inside the classroom, Merrill paces: “The second one is [it’s] magic. Magic is efficacious, it works. But you don’t know how it works. And you can’t ever get an explanation. You don’t have the opportunity of going to Hogwarts and asking Professor McGonagall.”
Merrill is hawking on the joys—and limitations—of AI. The public release of OpenAI’s ChatGPT in November 2022 has caused seismic societal shifts: LLMs are now being sued for everything from enabling teen suicide to intellectual capital theft. Everyday people are are using artificial intelligence to cheat on college papers, generate and profit from explicit OnlyFans content, and to read daily news. To the untrained eye, AI’s excellence and efficiency seemingly knows no limits. Many find themselves asking: is AI taking over most of our jobs too?
The answer seems to be a little complicated: yes and no.
When I analyzed thirty professions deemed most likely to be impacted by AI in the next decade (I mostly took from this chart by the U.S. Bureau of Labor Statistics), I was taken aback to find that most jobs in decline have been steadily decreasing—or had even plateaued—in the last decade. My other surprise? Computer science grads entering the software and development industry, which otherwise has had a very healthy job share growth of 27 percent between 2014 and 2024, are reporting the highest unemployment rates seen by the industry in decades. Why?
One answer might be found within David Marguerite, of the Luxembourg Institute of Socio-Economic Research. He says that the general decline we see in these white-collar professions (see Figure A) may be reflecting the impacts of artificial intelligence for some time now. What laypeople have come to consider “AI” is actually very narrow: “By ‘AI’ I mean everything that can fall into machine learning, natural language processing, large language models, all these big families of AI algorithms. It’s not generative AI per se,” he says.
If this was true, I couldn’t understand why all the jobs I had plotted didn’t experience negative growth. I asked Marguerite, and he pointed me to a concept he calls automation versus augmentation AI. In the former, AI directly replaces the task normally carried out by a human, says Marguerite. This is the sort of AI involvement that most of us have come to fear and relish—self-driving cabs and chatbots that calculate complex spreadsheets in seconds. But in the latter, augmentation AI does not replace work but complements it. Marguerite is even optimistic. He believes that through augmentation, AI tools like those that help diagnose complex diseases and teach lesson plans to neurodivergent children will create a whole new tier of specialized jobs—for humans—in the labor force.
Marguerite’s distinction helped explain why certain professions—like software developers and financial advisors—still grew in the last decade while postsecondary teachers and budget analyst jobs plummeted.
I found another clue in the categorization of the professions themselves:
Most professions in decline are relatively small—data entry
clerks, illustrators
The one exception? Food preparation workers and servers. Last year, they made up approximately 8.8
percent of the U.S. labor force--or nearly 14 million workers.
In the last decade, food-serving jobs grew 17 points
slower than the national average—a gap that translated
into 420,000 missing jobs last year.
7.
18.
11.
1.
2.
16.
27.
13.
12.
28.
10.
21.
3.
19.
5.
22.
4.
15.
25.
24.
6.
23.
9.
8.
8.
14.
26.
17.
29.
29.
Between 2014 and 2024, the average growth rate for
all occupations was 14 percent. The bubbles in the
shaded area represent professions that grew more
slowly than this national average.
-20
-15
-10
-5
0
5
10
15
20
25
30
35
40
The May 2024 U.S. Bureau’s Occupational Employment and Wage Statistics release does not include data for Colorado and its areas.
Table: Iryna Humenyuk • Source: U.S. Bureau of Labor Statistics • Created with RAWGraphs
1. Food preparation and serving-related occupations | 2. Business and financial-operations occupations | 3. Drivers/sales workers and truck drivers | 4. Software and
web developers | 5. Postsecondary teachers | 6. Secondary school teachers | 7. Legal occupations | 8. Lawyers | 9. Database and network administrators and
architects | 10. Executive secretaries and executive administrative assistants | 11. Paralegals and legal assistants | 12. Civil engineers | 13. Designers | 14. Claims
adjustors, examiners, and investigators | 15. Personal financial advisors | 16. Graphic designers | 17. Writers and editors | 18. Electrical engineers | 19. Computer
hardware engineers | 20. Aerospace engineering | 21. Data entry and information-processing workers | 22. Electrical and electronic engineering technologists and
technicians | 23. Aerospace engineering and operations technologists and technicians | 24. Credit analysts | 25. Illustrators | 26. News analysts, reporters and
journalists | 27. Budget analysts | 28. Insurance claims and policy-writing clerks | 29. Architects, except naval
1.
Most professions in decline are relatively
small—data entry clerks, illustrators
The one exception? Food preparation workers and servers. Last year, they
made up approximately 8.8 percent of the U.S. labor force--or nearly 14
million workers.
In the last decade, food-serving jobs grew 17 points
slower than the national average—a gap that translated into
420,000 missing jobs last year.
7.
18.
11.
1.
2.
16.
27.
13.
12.
28.
10.
21.
3.
19.
5.
22.
4.
15.
25.
24.
6.
23.
9.
8.
14.
26.
17.
29.
Between 2014 and 2024, the average
growth rate for all occupations was 14 percent.
The bubbles in the shaded area represent
professions that grew more slowly than this
national average.
-10
0
10
20
30
40
-20
The May 2024 U.S. Bureau’s Occupational Employment and Wage Statistics release does not include
data for Colorado and its areas.
Table: Iryna Humenyuk • Source: U.S. Bureau of Labor Statistics • Created with RAWGraphs
1. Food preparation and serving-related occupations | 2. Business and financial-operations occupations | 3. Drivers/
sales workers and truck drivers | 4. Software and web developers | 5. Postsecondary teachers | 6. Secondary school
teachers | 7. Legal occupations | 8. Lawyers | 9. Database and network administrators and architects | 10. Executive
secretaries and executive administrative assistants | 11. Paralegals and legal assistants | 12. Civil engineers |
13. Designers | 14. Claims adjustors, examiners, and investigators | 15. Personal financial advisors | 16. Graphic
designers | 17. Writers and editors | 18. Electrical engineers | 19. Computer hardware engineers | 20. Aerospace
engineering | 21. Data entry and information-processing workers | 22. Electrical and electronic engineering
technologists and technicians | 23. Aerospace engineering and operations technologists and technicians | 24. Credit
analysts | 25. Illustrators | 26. News analysts, reporters and journalists | 27. Budget analysts | 28. Insurance claims
and policy-writing clerks | 29. Architects, except naval
Most professions in decline are
relatively small—data entry clerks,
illustrators
The one exception? Food preparation workers and
servers. Last year, they made up approximately 8.8
percent of the U.S. labor force--or nearly 14 million
workers. .
In the last decade, food-serving jobs grew 17 points
slower than the national average—a gap that translated into
420,000 missing jobs last year.
7.
7.
18.
11.
1.
2.
16.
27.
13.
12.
28.
10.
21.
3.
19.
22.
5.
4.
15.
25.
24.
6.
23.
9.
8.
14.
26.
17.
29.
The bubbles in the shaded area
represent professions that grew
more slowly than the national
average.
-20
-10
0
10
20
30
40
The May 2024 U.S. Bureau’s Occupational Employment and Wage Statistics
release does not include data for Colorado and its areas.
Table: Iryna Humenyuk • Source: U.S. Bureau of Labor Statistics • Created
with RAWGraphs
-
1. Food preparation and serving-related occupations | 2. Business and financial-
operations occupations | 3. Drivers/sales workers and truck drivers | 4. Software and
web developers | 5. Postsecondary teachers | 6. Secondary school teachers | 7. Legal
occupations | 8. Lawyers | 9. Database and network administrators and architects |
10. Executive secretaries and executive administrative assistants | 11. Paralegals and
legal assistants | 12. Civil engineers | 13. Designers | 14. Claims adjustors, examiners,
and investigators | 15. Personal financial advisors | 16. Graphic designers | 17. Writers
and editors | 18. Electrical engineers | 19. Computer hardware engineers | 20. Aerospace
engineering | 21. Data entry and information-processing workers | 22. Electrical and
electronic engineering technologists and technicians | 23. Aerospace engineering and
operations technologists and technicians | 24. Credit analysts | 25. Illustrators |
26. News analysts, reporters and journalists | 27. Budget analysts | 28. Insurance claims
and policy-writing clerks | 29. Architects, except naval
Categories like “designers” and “business and financial-operations occupations” are too broad to be able to track meaningful changes in the data. It’s more useful to analyze individual professions, and not a sector as a whole. The one obvious exception to this rule? Food preparation workers (see Figure B). Last year they made up approximately 8.8 percent of the U.S. labor force—or nearly 12 million workers. Any shift that occurs in this sector will be felt throughout the entire economy because of its overwhelming size.
What else did I discover?
What was once one of America's smallest majors has grown, proportionally, by about 1,800 percent between 1970 and 2021. In comparison, healthcare degrees grew by 380 percent during this time—strong, but far less explosive than computer science. Why is this important? This means there is a far larger supply of computer science graduates relative to the job market, making entry-level positions much more competitive. This growth has been particularly remarkable since the stock market crash of 2008 (see red highlight in Figure C), and helps explain why so many recent computer science grads are more likely to be entertaining food service jobs at Chipotle instead of software jobs at Apple.
There’s also the problem of offshoring. Professor emeritus Richard Sylla, of New York University, believes that software fields are particularly susceptible: “I think there may be economies of scale in that area to take advantage of increased computing power and modelling capabilities.” To make an example of data analysts, firms and other organizations now likely outsource much of their data analysis to specialized data analytics companies rather than conducting the work in house. Sometimes, it’s cheaper—and yields higher quality yields for the purchasing company.
Traces of outsourcing, offshoring, and oversupply can be seen in the visualizations published in this May 2025 article published in The Economist, titled “Why AI hasn’t taken your job.” The tone is overwhelming balming. Young graduates are now more likely than the average worker to be jobless. Is technology to blame? “Well, no,” the author writes. “The data simply do not line up with any conceivable mechanism.” Because relative unemployment started to rise in 2009; because actual unemployment at six percent remains “relatively low”; because the share of white-collar work has risen slightly in the last year, the Economist concludes, Americans should worry less, not more.
I disagree.
The Economist is missing a critical point. The number of white-collar jobs might still be rising—but their share has nearly flatlined since 2009 (Figure D). To compare, the number of college graduates rose by 22 percent during that same period (from about 1,650,000 to over 2,000,000 jobs). The question this data point begets: do today’s college students deserve to know that the number of white-collar jobs in the U.S. are dwindling?
A second question, when looking at the unemployment rates that new college graduates have been facing since around the 2009 economic recession (Figure E): should postsecondary institutions be doing more to help recent grads land their first jobs?
I don’t have all the answers. But what is clear from the data that the U.S. job market is changing—rapidly, in some industries—even if AI isn’t one to solely credit or blame.
As David Autor, professor of economics at the Massachusetts Institute of Technology, told me over e-mail recently: “Job loss—especially if it takes out entire job categories or industries—is deeply scarring, economically and psychologically. If only five percent of workers lose their careers because of AI, that's still millions of very unhappy people.”