New Data Confirms That AI Is Already Taking Human Jobs, Roles

In March, Shopify’s CEO told his managers he was implementing a new rule: Before asking for more head count, they had to prove that AI couldn’t do the job as well as a human would. A few weeks later, Duolingo’s CEO announced a similar decree and went even further — saying the company would gradually phase out contractors and replace them with AI. The announcements matched what I’ve been hearing in my own conversations with employers: Because of AI, they are hiring less than before.

When I first started reporting on ChatGPT’s impact on the labor market, I thought it would take many years for AI to meaningfully reshape the job landscape. But in recent months, I’ve found myself wondering if the AI revolution has already arrived. To answer that question, I asked Revelio Labs, an analytics provider that aggregates huge reams of workforce data from across the internet, to see if it could tell which jobs are already being replaced by AI. Not in some hypothetical future, but right now — today.

Zanele Munyikwa, an economist at Revelio Labs, started by looking at the job descriptions in online postings and identifying the listed responsibilities that AI can already perform or augment. She found that over the past three years, the share of AI-doable tasks in online job postings has declined by 19%. After further analysis, she reached a startling conclusion: The vast majority of the drop took place because companies are hiring fewer people in roles that AI can do.

Next, Munyikwa segmented all the occupations into three buckets: those with a lot of AI-doable tasks (high-exposure roles), those with relatively few AI-doable tasks (low-exposure roles), and those in between. Since OpenAI released ChatGPT in 2022, she found, there has been a decline in job openings across the board. But the hiring downturn has been steeper for high-exposure roles (31%) than for low-exposure roles (25%). In short, jobs that AI can perform are disappearing from job boards faster than those that AI can’t handle.

Which jobs have the most exposure to AI? Those that handle a lot of tech functions: database administrators, IT specialists, information security, and data engineers. The jobs with the lowest exposure to AI, by contrast, are in-person roles like restaurant managers, foremen, and mechanics.

This isn’t the first analysis to show the early impact of AI on the labor market. In 2023, a group of researchers at Washington University and New York University homed in on a set of professionals who are particularly vulnerable: freelancers in writing-related occupations. After the introduction of ChatGPT, the number of jobs in those fields dropped by 2% on the freelancing platform Upwork — and monthly earnings declined by 5.2%. “In the short term,” the researchers wrote, “generative AI reduces overall demand for knowledge workers of all types.”

At Revelio Labs, Munyikwa is careful about expanding on the implications of her own findings. It’s unclear, she says, if AI in its current iteration is actually capable of doing all the white-collar work that employers think it can. It could be that CEOs at companies like Shopify and Duolingo will wake up one day and discover that hiring less for AI-exposed roles was a bad move. Will it affect the quality of the work or the creativity of employees — and, ultimately, the bottom line? The answer will determine how enduring the AI hiring standstill will prove to be in the years ahead.

Some companies already appear to be doing an about-face on their AI optimism. Last year, the fintech company Klarna boasted that its investment in artificial intelligence had enabled it to put a freeze on human hiring. An AI assistant, it reported, was doing “the equivalent work of 700 full-time agents.” But in recent months, Klarna has changed its tune. It has started hiring human agents again, acknowledging that its AI-driven cost-cutting push led to “lower quality.”

“It’s so critical that you are clear to your customer that there will always be a human,” CEO Sebastian Siemiatkowski told Bloomberg. “Really investing in the quality of the human support is the way of the future for us.”

Will there be more chastened Siemiatkowskis in the months and years ahead? I’m not betting on it. All across tech, chief executives share an almost religious fervor to have fewer employees around — employees who complain and get demotivated and need breaks in all the ways AI doesn’t. At the same time, the AI tools at our disposal are getting better and better every month, enabling companies to shed employees. As long as that’s the case, I’m not sure white-collar occupations face an optimistic future.

Even Siemiatkowski still says he expects to reduce his workforce by another 500 through attrition in the coming year. And when Klarna’s technology improves enough, he predicts, he’ll be able to downsize at an even faster pace. Asked when that point will come, he replied: “I think it’s very likely within 12 months.”

Aki Ito is a chief correspondent at Business Insider.