Since the Chat GPT’s late 2022 release, countless headlines have predicted an apocalyptic future for workers:
“We asked Chat GPT what jobs he thought would replace – and this data is not good news for professionals or reporters.”
“Goldman Sex predicts that artificial intelligence will lose 300 million jobs or lack them”. Forbes, March 31, 2023
CBS News, CBS News in April 5, 2023, “How many US workers here say Chattagpat can change”.
Fox Business, Millions of people can be excluded from work in April 5, 2023, “Chat GPTA is a list of jobs lists because it can work better than humans.”
“Chat GPT gets his jobs. Now they run dogs and fix the air conditioners” – Washington Post, June 2, 2023
“This AI company wants to get your job” – New York Times, June 11, 2025.
And still, in the subsequent three years, we still have to see such mass Labor market According to a October 1 report by the Yale Budget Lab, alarmists have been predicted.
The report states that people have not changed between jobs, new roles have not been revealed on a scale and workers have not been automatically automatically from their positions. For now, AI has not come for your job.
This does not mean that people are not upset. A 13-18 survey by Reuters/IPSO found that 71 % of respondents are concerned AI “will permanently remove many people from work.” Fear is clearly real.
Nevertheless, the budget lab says it is not surprising that AI has not yet disrupted the job market. History tells us that the new technology usually takes decades to eliminate the workplace.
The report states that “computers have not become normal in offices for a decade of release for the public, and it took them more time to change the office workflow.” Even if the new AI technologies will maximize the labor market, or more, dramatically, it is appropriate to expect that it takes more than 33 months to implement the massive effects. “
Over time, the AI is expected to be anxious to the labor market, but some professionals – and experience levels – especially suffer from this risk.
The job safety is uneven from the field
While AI-motivation Job’s relocation And Creation This is unclear, when the change happens, it is likely to be in some fields compared to others. Most of this depends on how automated your role.
On September 8, a study by the University of Pennsylvania School of Business ranks the jobs that are the highest and least automation by Generative AI.
Most exposure professions (about 50 % and more):
Office and administrative help: 75.5 %.
Business and financial work: 68.4 %.
Computer and mathematics: 62.6 %.
Sale and related profession: 60.1 %.
Moderate exposure profession (30 % -49.9 %):
Administrative profession: 49.9 %.
Arts, design, entertainment, sports and media: 45.8 %.
Architecture and engineering: 40.7 %.
Life, physical and social science: 31 %.
Low exposure (20 % -29.9 %):
Academic instructions and library: 29.5 %.
Community and social service: 27.5 %.
Healthcare practitioners and technical: 23.1 %.
Protective service: 20.7 %.
Transport and Material Dynamic: 20 %.
The lowest exposure (less than 20 %):
Food preparation and service: 18.1 %.
Personal care and service: 17.5 %.
Healthcare support: 15.5 %.
Installation and maintenance and repair: 13.1 %.
Farming, fishing and forests: 9.7 %.
Construction and extraction: 8.9 %.
Cleaning and restoration of building and foundations: 2.6 %.
A Chat GPT-based model was used in a report by San Bernie Sanders (D-VT.
Fast food and counter workers (89 %).
Customer Service Representative (83 %).
Labor and freight, stock, and material movement (81 %).
Secretaries and administrative assistants, except legal, medical and executive (80 %).
Stickers and Order fillers (76 %).
Book capping, accounting, and auditing clerk (76 %).
Office clerks, general (66 %).
Teaching Assistant, Pre -School, Early, Middle, and Secondary School, Except Special Education (65 %).
Accountant and Auditor (64 %).
Retail salesperson (62 %).
Jeanators and cleaners, except maids and house capping cleaner (61 %).
Software developers (54 %)
Waiter and waitress (53 %).
Young workers face the most AI barriers
The evidence is growing that the effects of AI on his job are being targeted by the under -age workers.
Stanford University’s August 2025 studies suggest that General Z workers (who are 22 to 25 years old), who are in the highest profession of AI, have suffered a 13 % reduction in employment since 2022. High exhibition sectors include software developers, software engineers and customer service, call centers and auxiliary roles.
Stanford’s report also found that employment drops are the highest in the characters where AI replaced them rather than expanding them. Early carrier workers perform job functions, which have the most capacity of automation.
In other words, young workers are most exposed to these characters because according to Stanford’s study, they have more “learning” and less employment experience. AI can more easily replace more codified aspects of its work-rules, phased processes and formal education teachings. But it is struggling with the type of work that requires “text knowledge” that is with old workers, such as calling for a decision, being intuitive and shortcut.
There is more evidence that AI seems to be prejudicing a seniority on a job. On September 8, a study by Harvard University reviewed a large -scale datastate of US experiences and job postings to find out if AI is interrupting work for junior employees than AI senior. Researchers found that in the early 2023, Junior Series served in AI -using firms. This decline was mainly due to companies serving early career workers. This study has found that in the fields where young workers are most affected are spit and retail trade.
Initial career workers have taken notice of services for changes and are desperate for their future prospects. According to a Deutsche Bank’s September 23 report, about a quarter of the ages of 18 to 34 in the United States and the whole of Europe are concerned that AI will exclude them from work in the next two years.
We still do not know how AI is changing the job
The Budget Lab report states that the problem with measuring the effects of AI at the workplace is that we do not yet have fully accurate data about the exposure.
Large language models (LLMs) such as interstropic and open AI data are useful, but incomplete. They do not calculate all the tasks, professions, AI tools, real -world barriers or obstacles to adoption. All that is to say is, draw conclusions about the labor effects of AI with salt grain.
For now, what we know is that AI is permanently going to many workplaces. Another report from Stanford University has found that the adoption of LLM at work has increased from 30.1 percent in December 2024 to 45.6 percent from June and July 2025.
–
