Recently, I wrote about the higher unemployment rate for young workers, focusing on 20- to 24-year-olds, and also how younger men are disproportionately impacted. This week, I am still focusing on the younger workforce, but looking specifically at how artificial intelligence, or AI, is impacting this age cohort.

An excellent Substack post by Derek Thompson brought this to my attention.

The debate began with the observation that opportunities for recent college graduates had “deteriorated noticeably,” according to the New York Federal Reserve. Some plausible explanations include skittish business hiring practices due to higher interest rates and tariff uncertainty. But many economists pointed out that entry-level white-collar work is exactly the type of task generative AI can perform, such as simple research, data entry, memo writing, Excel spreadsheets and PowerPoint presentations.

Media coverage quickly escalated, with headlines warning that the “AI job apocalypse” might already be here, and tech leaders like Anthropic’s Dario Amodei predicting a “bloodbath” that could eliminate half of all entry-level white-collar jobs within five years.

But soon, a counter-narrative emerged. Analysts using aggregate government data found little evidence that AI was displacing young workers on a large scale. The Economic Innovation Group and commentators like John Burn-Murdoch at the Financial Times argued that hiring in many exposed sectors was actually stabilizing or recovering.

For a minute, the debate seemed settled that AI is not the culprit. Then, new evidence from Stanford economists reignited the conversation — and its results ring true to me. Using ADP payroll data covering millions of workers, the study found that young workers in AI-exposed roles, such as software developers and customer service reps, saw employment fall by 13% since ChatGPT’s release.

In contrast, older workers and those in less exposed, more physical jobs — such as home-health aides — saw steady or even rising employment.

Increases in health care jobs, in particular, make sense to me given the aging of our population. So, as is usually the case in economics, the devil is in the details. In aggregate, overall employment may not be affected as much, but in certain higher-skill occupations, the hit could be hard, especially for younger workers.

College grads who majored in the liberal arts often relied on these entry-level positions to get their foot in the door with an employer. The replacement of those jobs with AI presents yet another barrier.

It may be anecdotal, but I often think of my own experience as someone who runs a small nonprofit. I have honestly been surprised at how much our team of four (cautiously) uses AI to data mine, summarize data and shorten my verbose reports.

Our team is a good example of the need for a senior economist (ahem, me) who knows how to validate data with both intuition and expertise, which AI cannot do. This is likely why the Stanford study showed senior, professional/technical workers have not been affected by AI (while the opposite is true for younger workers).

In terms of young people and their love of AI, I tell them to include sources in all AI inquiries in order to cross-check things. But even with that added layer of authentication, AI saves a lot of time — as well as the need to quickly hire a junior analyst. Apparently, this is what many business leaders are figuring out.

Most experts state that higher education needs to teach how to use AI and the fundamentals of coding. I wholeheartedly agree that those who move with the waves of innovation always fare better. And if you are still skeptical, look at how AI infrastructure spending has been fueling the stock market.

The changes are happening in real time.