AI Is Eliminating Jobs for Younger Workers

Economists astatine Stanford University person recovered the strongest grounds yet that artificial intelligence is starting to destruct definite jobs. But the communicative isn’t that simple: While younger workers are being replaced by AI successful immoderate industries, much experienced workers are seeing caller opportunities emerge.

Erik Brynjolfsson, a prof astatine Stanford University, Ruyu Chen, a probe scientist, and Bharat Chandar, a postgraduate student, examined information from ADP, the largest payroll supplier successful the US, from precocious 2022, erstwhile ChatGPT debuted, to mid-2025.

The researchers discovered respective beardown signals successful the data—most notably that the adoption of generative AI coincided with a alteration successful occupation opportunities for younger workers successful sectors previously identified arsenic peculiarly susceptible to AI-powered automation (think lawsuit work and bundle development). In these industries, they recovered a 16 percent diminution successful employment for workers aged 22 to 25.

The caller survey reveals a nuanced representation of AI’s interaction connected labor. While advances successful artificial intelligence person often been accompanied by dire predictions astir jobs being eliminated—there hasn’t been overmuch information to backmost it up. Relative unemployment for young graduates, for instance, began dropping astir 2009, good earlier the existent AI wave. And areas that mightiness look susceptible to AI, specified arsenic translation, person really seen an summation successful jobs successful caller years.

“It's ever hard to cognize [what’s happening] if you're lone looking astatine a peculiar institution oregon proceeding anecdotes,” Brynjolfsson says. “So we wanted to look astatine it overmuch much systematically.”

By combing done payroll data, the Stanford squad recovered that AI’s interaction has much to bash with a worker’s acquisition and expertise than the benignant of enactment they do. More experienced employees successful industries wherever generative AI is being adopted were insulated from occupation displacement, with opportunities either remaining level oregon somewhat growing. The uncovering backs up what immoderate bundle developers antecedently told maine astir AI’s interaction connected their industry—namely that rote, repetitive work, similar penning codification to link to an API, has go easier to automate. The Stanford survey besides indicates that AI is eliminating jobs but not lowering wages, astatine slightest truthful far.

The researchers considered perchance confounding factors including the Covid pandemic, the emergence of distant work, and caller tech assemblage layoffs. They recovered that AI has an interaction adjacent erstwhile accounting for these factors.

Brynjolfsson says the survey offers a acquisition connected however to maximize the benefits of AI crossed the economy. He has agelong suggested that the authorities could alteration the taxation strategy truthful that it does not reward companies that regenerate labour with automation. He besides suggests AI companies make systems that prioritize human-machine collaboration.

Brynjolfsson and different Stanford scientist, Andrew Haupt, argued successful a insubstantial successful June that AI companies should make caller “centaur” AI benchmarks that measurement human-AI collaboration, to incentivize much absorption connected augmentation alternatively than automation. “I deliberation there's inactive a batch of tasks wherever humans and machines tin outperform [AI connected its own],” Brynjolfsson says.

Some experts judge that much collaboration betwixt humans and AI could beryllium a diagnostic of the aboriginal labour market. Matt Beane, an subordinate prof astatine UC Santa Barbara who studies AI-driven automation, says helium expects the AI roar to make request for augmentable work—as managing the output of AI becomes progressively important. “We'll automate arsenic overmuch arsenic we can,” Beane says. “But that doesn't mean determination won't beryllium a increasing upland of augmentable enactment near for humans.”

AI is advancing rapidly though, and Brynjolfsson warns that the interaction connected younger workers could dispersed to those with much experience. “What we request to bash is make a dashboard early-warning strategy to assistance america way this successful existent time,” helium says. “This is simply a precise consequential technology.”

This is an variation of Will Knight’s AI Lab newsletter. Read erstwhile newsletters here.