Well, it’s finally happening. After years of tech CEOs and sci-fi movies warning us that robots are coming for our jobs, some economists at Stanford University have crunched the numbers and found the strongest evidence yet: AI is indeed starting to push people out of work.
But there’s a twist. It’s not your seasoned, coffee-stained, seen-it-all-before veteran who’s getting the pink slip. It’s the bright-eyed, bushy-tailed new graduate.
A team of researchers at Stanford dove into US payroll data from late 2022 (right when ChatGPT burst onto the scene) to mid-2025. What they found paints a surprisingly nuanced picture of our new AI-powered economy.
The big headline? In industries most vulnerable to AI—think customer service chatbots and software that can now write its own code—employment for workers aged 22 to 25 dropped by a significant 16%. Ouch.
Meanwhile, their more experienced colleagues? They’re doing just fine, thank you very much. Opportunities for seasoned workers in those same fields have either held steady or even grown a little.
So what’s going on? It turns out AI isn’t just replacing “tech jobs” wholesale. It’s replacing tasks, particularly the repetitive, entry-level grunt work that used to be the rite of passage for every new hire. Why have a junior developer spend three days writing boilerplate code when an AI can do it in three seconds? The value is shifting from doing the simple task to managing and directing the AI that does it—a skill that comes with experience.
The study’s lead author, Erik Brynjolfsson, pointed out that this isn’t just a bunch of scary anecdotes anymore. “It’s always hard to know if you’re only looking at a particular company,” he said. “So we wanted to look at it much more systematically.” And the data shows that even when accounting for other factors like pandemic recovery or recent tech layoffs, AI’s fingerprint is all over this shift.
The good news (for some) is that, so far, AI is eliminating jobs but not necessarily driving down wages for those who remain. The bad news is that the classic career ladder might be losing its bottom rungs.
So, what’s the solution? Brynjolfsson has some ideas. He suggests we might need to rethink our tax code so it doesn’t actively reward companies for automating humans out of a job. More interestingly, he and other researchers argue that AI companies should focus less on building systems that replace us and more on building systems that collaborate with us. They even propose new “centaur” benchmarks—named after the mythical half-human, half-horse creature—that measure how well a human and an AI can work together as a team.
The bottom line? The future of work might not be a straight choice between human or machine. It might be about finding the best way to be the human on the machine. For now, if you’re just starting your career, you might want to learn how to boss an AI around. It seems that’s where the jobs are headed.