In the 1980s, there was New Wave music, then there was Devo. Five men in hazmat suits with flower pots on their heads who played and moved in an unsettlingly mechanical manner, Devo stood out in a sea of bad haircuts. Musically, the band was best known for the hit Whip It, and – a few decades later – Don’t Roof Rack Me, Bro about Presidential candidate Mitt Romney’s questionable decision to bring the family dog on a road trip by strapping him to the roof.
Devo, the subject of a new documentary on Netflix, was intentionally controversial, saying things like “the band never said they didn’t like religion or didn’t care for Christianity, they just said they would rather have cancer,” and “we’re basically a musical laxative for a constipated society.” But it wasn’t entirely a joke. When asked by an interviewer “what would make you all happy,” Devo responded with as much sincerity as they could muster: “We want people movers like they have at Disneyland.”
Growing up in the 50s, band members were promised people movers and flying cars. What they got in the 60s was Vietnam and Kent State – Devo’s co-founders were at Kent State in 1970 when the National Guard shot and killed four student protesters – followed by wanton abandonment of those ideals in the 70s and 80s. So before it was a band, Devo (short for de-evolution) was a homespun philosophy: contrary to appearances, humans are actually moving backwards. Hence Devo’s first album: Are we not men? We are Devo!
No one really knows what the future holds. Not 1950s futurists like Walt Disney. Not Devo which – to promote their 1981 album New Traditionalists – wore plastic wigs modeled on JFK’s hairstyle, but which were widely confused for Ronald Reagan’s bad haircut.
The same is true of AI prognosticators. AI predictions are all over the place. While the AI 2027 scenario from the AI Futures Project predicts our extermination in five years, two Princeton researchers argue there’s no cause for alarm because AI is a normal technology subject to a range of time-tested safety mechanisms.
Debates over extermination of entry-level jobs are no less heated. For every Dario Amodei (Anthropic’s CEO) who believes AI could replace half of all entry-level jobs, there’s a Matt Garman (AWS’s CEO) who says that’s the “dumbest thing” he’s ever heard. So let’s summarize the arguments on each side.
On one hand:
Even before AI, as Wharton’s Peter Cappelli neatly recapped, “everybody want[ed] to hire somebody with three years’ experience, and nobody want[ed] to give them three years’ experience.”
AI does a better job at the routine and repetitive tasks common in entry-level jobs.
So hiring managers will redefine entry-level roles to focus on higher value client work, product work, and project work – work that can’t be done (or at least not well) without relevant prior work experience. And on the treadmill to and through college, few new grads obtain relevant prior work experience.
There’s already a blizzard of anecdotal evidence on social media from dispirited 20-somethings like a 2023 computer science grad who applied for 5,762 tech jobs, got 13 interviews, but no offers. You’ll also find plenty of comments better suited for a cheap T-shirt e.g., I just graduated with a computer science degree and my only interview was with Chipotle.
And on the other hand:
New technology has always increased demand for workers in new occupations.
Entry-level workers are the least expensive, bring fresh ideas and new thinking, and are the most AI-native and best situated to take advantage of AI at work; in short, AI is about to make entry-level workers more productive.
AI allows workers with less education and training to perform more complex tasks.
What doesn’t make the list of optimistic arguments? Matt Garman’s point that employers would be crazy to reduce entry-level roles because “without junior employees gaining experience, companies will have no skilled workforce in the future.” Rarely are companies accused of prioritizing the long term over short-term results like reducing expenses to hit a quarterly or annual target. Also, while AI might replace some entry-level roles, it will create new ones. Of course, but will the net effect be profoundly negative and transform an already bad situation – 52% underemployment, 12% unemployment for grads in their 20s – into an existential crisis of career launch and economic mobility?
You’ve probably already used AI for a task and thought: “that’s something I don’t need to hire a researcher or analyst to do.” This appears to be the crux of the matter. In arguing the optimistic case in Washington Monthly, Progressive Policy Institute’s Bruno Manno reframes the last point in the optimistic case – i.e., AI allows workers with less education and training to perform more complex tasks – as “AI raises the bar for demonstrated expertise while lowering barriers to acquiring it.” But expertise is a portmanteau for both skills and experience – a bag begging to be unpacked. Sure, AI raises the bar for demonstrated skills while lowering barriers to acquire those skills. But while AI raises the bar for demonstrated experience, it does zilch to help aspiring career launchers acquire experience. Manno posits hypothetical AI simulations, but I don’t know any hiring manager who equates a simulation with real work experience.
I’m here with news that the optimistic case has been on a monthlong losing streak. And if the trend continues, apprenticeships – which by definition do not require prior work experience – are about to become critical national infrastructure.
On August 26, three Stanford economists released the paper Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence. In reviewing comprehensive labor market data from ADP, America’s largest payroll software platform, they found that, as of July, in sectors with AI exposure, the number of entry-level hires has already declined by 13% relative to expected. In AI-exposed sectors like software development and customer service, there was no impact for more experienced workers; only entry level positions have been displaced. Remarkably, the number of employed software developers ages 22-25 is already down nearly 20%. Equally, there was no impact for entry-level jobs with little AI exposure to date. And since entry-level employment in AI-exposed occupations didn’t move in lockstep prior to ChatGPT’s release three years ago, there can be no explanation other than AI.
Not to be outdone, on August 31 two Harvard economists shared their take with a less catchy title: Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data. This study used data from LinkedIn and Revelio Labs covering 62 million workers and nearly 250 million job postings. Rather than trying to pinpoint the impact of AI by sector or occupation, the Harvard researchers identified companies signaling an intent to take advantage of AI – i.e., by recruiting for new “AI integrator” roles – and then compared hiring patterns. Between January 2023 and July 2024, entry-level headcount at AI adopters declined by 7.7% with no comparable decline for more senior roles. The only semi-divergent finding is where the declines are occurring fastest: not in coding, but wholesale and retail trade where entry-level hiring fell by 40%. But the conclusion is the same: “generative AI constitutes a form of seniority-biased technological change, with adverse consequences for junior relative to senior employment within the firm.”
Finally, a number of new surveys confirm this dire direction. One survey of 800+ hiring managers found 70-80% saying that AI can do the work of a new hire and recent graduates would be laid off as a result. Per LinkedIn, 63% of executives agree AI will take on “some of the mundane, manual tasks that entry-level employees at their organizations currently focus on.” Meanwhile, a poll of nearly 1,200 new grads in the U.S. and UK found 59% saying “it has been very challenging to find a full-time, entry-level job this year” with 79% believing AI is reducing the number of entry-level jobs in their desired field. Not surprisingly, 60%+ of the class of 2026 is pessimistic about landing a good first job.
As the Harvard economists conclude: “Generative AI appears to shift work away from entry-level tasks, narrowing the ‘bottom rungs’ of internal career ladders.” Stanford’s team endeavors to explain why: “AI replaces codified knowledge, the ‘book-learning’ that forms the core of formal education. AI may be less capable of replacing tacit knowledge, the idiosyncratic tips and tricks that accumulate with experience.” All of which has started to dislocate would-be career launchers with more book learning than tacit knowledge.
Perhaps we needn’t worry because OpenAI CEO Sam Altman says in a decade college graduates will all be working IN SPACE at some “some completely new, exciting, super well-paid” job. But even OpenAI partner Microsoft is worrying a bit.
What’s clear is that AI is on its way to being widely adopted for entry-level roles. While the Stanford paper attempted to identify entry-level positions where AI is augmenting rather than displacing, we’re still in the first inning of agentic AI. As use cases proliferate, identifying true entry-level (little or no prior experience) roles where AI isn’t displacing is likely to become a needle-in-a-haystack exercise.
It’s also clear that AI’s impact on entry-level hiring won’t be equitable. If relevant work experience becomes essential, those with the connections to obtain it will be in the pole position. Counterintuitively, wealth and status may matter even more than they do in today’s tuition-based system of career launch. And it may matter even more what school you attend. In determining which universities are most susceptible to declining entry-level hiring, the Harvard study finds a U-shaped pattern. While Harvard and its confrères are barely affected, graduates of tier 2 and tier 3 institutions are being hit hard. One possible explanation is that graduates of these institutions have historically been hired into good jobs by companies likely to adopt AI, but AI is allowing these firms to hire even more selectively. Meanwhile, graduates of tier 4 and 5 schools haven’t typically launched into these kinds of businesses.
On the skills side, AI skills are likely to tower in importance over any others. Students are seeing absurd salaries being thrown at entry-level data engineers and data analysts with AI capabilities. AI companies have been busy rolling out tools, training, and certifications to meet demand. As millions of companies will need help setting up and managing AI agents – and making sure they don’t go haywire – there’s little risk of underemployment for grads with these skills.
Finally, while I’m seeing a growing number of commentaries about how AI will reduce the importance of technical education and training, reaching the heartwarming conclusion that a liberal arts renaissance is on the horizon, I fear this is wishful thinking. The main thrust of AI’s impact on education and workforce will be to reduce the importance of classroom-based education and increase the importance of work experience, or work-based learning at a minimum. In a nutshell, AI is in the process of devaluing degrees and making work experience more valuable.
This is why Bruno Manno concludes his Washington Monthly piece with an apprenticeship call-to-arms: “By the end of an apprenticeship, the ‘entry-level’ paradox is resolved since the apprentice is no longer inexperienced. Instead, the apprentice has the skills and track record of a productive worker. They make experience part of the first job rather than a barrier to getting it.”
In 20 years, it’s likely that classroom-based education will play a lesser role from high school on up. Traditional teaching models will shrink by tens of billions of dollars — public and private provision alike — to be replaced by work-based learning models like apprenticeships, co-ops, internships, and projects with an employer-in-the-loop, but orchestrated by intermediaries and schools that seize the opportunity.
My favorite Devo song is their cover of the Rolling Stones’ Satisfaction. It’s all electronic, herky jerky, and mesmerizing. They performed it on SNL and came across as five young men from Ohio who were really quite dissatisfied.
Growing up in the 1950s, Devo was promised people movers and flying cars but ended up with Vietnam and Reagan. Two generations later, young Americans are being promised AGI and agents to do their scut work but will probably end up underemployed or unemployed. Which will lead them to view this so-called progress as devolution. So our choice is clear: make major changes or prepare for social unrest + Devo cover bands.
We should also prepare for more unrest in higher education. On one hand, the Trump Administration’s omni-front attack on higher education couldn’t come at a worse time. On the other hand, most colleges and universities are about to have a big problem that will make their little problems disappear.