Some workers are afraid of losing their job to AI. Some top CEOs seem to agree.
Fiverr CEO Micha Kaufman told his employees in early 2025 that “AI is coming for your jobs.” Shopify CEO Tobi Lutke notified employees they needed to justify why AI couldn’t do a job before hiring anyone else. And Amazon CEO Andy Jassy, in a memo, said outright that AI would lead the e-commerce giant to “reduce our total workforce.”
Plenty of other business leaders have followed suit, issuing memos and directives riffing on the same theme: AI is increasingly able to do the work of humans, and employees need to acknowledge that, either by mastering AI to become massively more productive or upskilling to do work that AI can’t.
But does the data back this up? Much like the story of AI itself, it’s complicated.
AI’s Current Impact on Jobs
It’s complicated for several reasons.
First, ChatGPT took the world by storm in November 2022. That timing coincided with the beginning of a multi-year wind-down in headcount and hiring that companies began to engage in after exiting the pandemic. That’s why it’s hard to quantify the extent to which AI may have been to blame for recent headcount reductions. For instance, Microsoft recently laid off 9,000 people, but did not explicitly point to AI as the reason. The broader tech sector has seen job losses in the hundreds of thousands over the last two years, but many experts attribute this to previous overhiring.
Second, while CEOs are talking more about AI’s impact on employment, few are eager at the moment to directly attribute staff reductions to AI. Klarna CEO Sebastian Siemiatkowski was one of the most vocal to stick his neck out, saying the company cut staff by 40% thanks to AI. But a year after those comments, he admitted the company actually hired a substantial number of people back because AI couldn’t do everything expected of it.
Still, recent research holds some clues as to which jobs are being affected and shows early signals of how that is playing out in labor markets for these jobs.
Ali Zarifhonarvar, an economics researcher at Indiana University, has studied the impact of generative AI on a wide range of occupations. In one study, he employed a text-mining approach to analyze the International Standard Classification of Occupations (ISCO) database, which provides detailed descriptions of job roles and responsibilities. The approach revealed intricate details about job tasks and skills that are vulnerable to AI.
According to the findings, the category of “professionals” is most exposed to AI, with 95 distinct occupations susceptible to the “full impact” of AI disruption, meaning the majority of their core tasks could be automated. In this category, the study highlights jobs most under threat. These include media jobs, which consist of roles in advertising, technical writing, journalism, and content creation. They also include legal industry jobs, such as paralegals and legal assistants. And they encompass finance-related roles, which include market research analysts and personal financial advisors.
The second most-affected group is “technicians and associate professionals,” where 60 distinct occupations are estimated will be exposed to the full impact of AI. The top-affected professions include software developers, computer programmers, and data scientists.
The jobs most affected by AI share some commonalities, said Zarifhonarvar. They are especially vulnerable because they involve tasks such as data processing, writing, analyzing information, and providing advice, all of which AI can now do better and faster than humans.
“This marks a shift from previous waves of automation that mainly affected routine tasks,” he said.
That trend is similar to what was found by other research. A study released by Microsoft in July 2025 analyzed 200,000 real-world interactions with the company’s AI tool, Copilot. Jobs with the highest exposure to AI were interpreters and translators, sales representatives, writers and authors, customer service reps, and reporters, the study found.
That exposure is beginning to lead to a real impact on some of those jobs, according to research from Ozge Demirci, an assistant professor at the U.K.’s Imperial College London Business School. Her work shows notable declines in online freelance demand in writing and software work following the introduction of generative AI tools. Demand for writing work on online freelance platforms has dropped by about 30%. Demand for software work has dropped by 20%.
“These categories are particularly susceptible because AI can now generate credible text and code quickly and cheaply, often meeting basic client needs without human input,” she said.
Much of the data, however, is predictive of the future, not descriptive of the present.
Few Safe Harbors
Looking ahead, several experts painted a grim picture. The World Economic Forum’s 2025 Future of Jobs Report estimates 92 million current jobs will be destroyed due to AI between now and 2030. Dario Amodei, co-founder and CEO of Anthropic, a major AI lab, predicts up to 50% of white-collar work could disappear in the next one-to-five years.
“There will likely be job displacement,” Zarifhonarvar said. While the economy may grow in the future as AI creates new opportunities and jobs, there will also be significant job losses during the transition.
So which jobs are actually safe from AI moving forward?
Manual jobs that require human physical activity are safest from AI disruption, according to Zarifhonarvar’s data. Categories least affected include craft and related trades workers, elementary occupations (those involving physical labor), and plant and machine operators and assemblers.
These workers are relatively protected, since their work relies on physical skills, hands-on tasks, and human interaction, said Zarifhonarvar. However, they may still face risks eventually from labor-saving technologies like robotics.
Demirci agreed, noting that jobs relying heavily on in-person presence, manual skills, or nuanced human interaction are much less likely to be displaced any time soon. These include healthcare roles, on-site service jobs, or work requiring complex client communication.
This is also corroborated by Microsoft’s research. Its study found physical jobs virtually untouched by AI disruption, with roles like roofers, nursing assistants, construction workers, and plant operators being least affected by AI.
Additionally, it is likely that AI will increase the capabilities of top managers and benefit high-paying leadership jobs, said Gino Gancia, an economics professor at Italy’s University of Milano-Bicocca.
“AI will benefit jobs that involve high-stakes, high-responsibility decisions,” Gancia said. “These decisions will not be delegated to machines, at least in the near future.”
But even with some jobs being safe from AI, the technology could create longer-term structural problems in the labor market that affect everyone.
“Inequality will increase, especially at the top,” said Gancia. “Society must prepare for this.”
Reskilling and upskilling existing workers are the key issue, says Demirci. Workers increasingly are expected to integrate AI into their work, and those who fail to adapt risk falling way behind. The danger is a widening gap between those who can upskill quickly and those who cannot.
Logan Kugler is a technology writer specializing in artificial intelligence based in Tampa, FL. He has been a regular contributor to CACM for 15 years and has written for nearly 100 major publications.
