The Entry-Level Jobs AI is Actually Killing
A Stanford study confirms what new grads already fear: AI isn't replacing all work—it's specifically targeting the positions that used to onboard talent. Here's what the data shows and what you can do about it.
The Squeeze Is Real—And It's Only Young Workers Feeling It
<cite index="1-1">In the first large-scale analysis of real payroll data since ChatGPT went viral, researchers at Stanford and the NBER found that entry-level workers in AI-exposed jobs have seen a 13% drop in employment since late 2022.</cite> But here's the kicker that makes the Stanford findings especially sharp: <cite index="1-13,1-15">Jobs in fields like software engineering and customer service are still growing, but not for people aged 22 to 25—older workers in the same jobs have seen employment rise by 6 to 9 percent.</cite>
This isn't a sign of economic collapse. <cite index="1-18,1-19">The drop is laser-focused on junior workers in AI-exposed roles. It is not because the whole economy is slowing down—in fact, employment overall is still rising.</cite> The research team, <cite index="1-12">led by Erik Brynjolfsson, analyzed millions of payroll records from ADP, the largest payroll processor in the U.S.</cite>
Where It Hits Hardest
<cite index="2-10">Some of the occupations with the biggest genAI hit included software development and customer service.</cite> And the decline is accelerating. <cite index="5-4">Employment for 22-to-25-year-olds in the most AI-exposed jobs is shrinking 3.8% a year—and accelerating—while workers aged 35-40 in the same fields are growing.</cite>
The specificity matters. <cite index="2-18">By July 2025, employment for software developers aged 22 to 25 declined by nearly 20% from its peak in 2022.</cite> <cite index="6-1">For 22-25 year olds, jobs where "automative" users of AI are especially prevalent show heavily declining entry level employment rates.</cite> Meanwhile, <cite index="6-4,6-5">for jobs where AI is merely "augmentative," the entry level jobs picture is much more muddled.</cite>
This distinction matters because it shows the problem isn't universal. <cite index="1-16">Young workers in fields less impacted by AI, like healthcare, are doing fine.</cite>
Why Young Workers First?
The mechanism is brutal in its logic. AI is built to automate routine, codifiable work—exactly what entry-level jobs are designed to teach. <cite index="9-1">31% of employers say AI has raised experience requirements for their entry-level roles.</cite> Worse, <cite index="9-12,9-2">38% of employers have already moved basic data entry and processing away from entry-level workers and onto AI.</cite>
What replaces this training ground? <cite index="17-6,17-7,17-8">AI didn't kill the junior job—it made it senior. AI is absorbing the grunt work, the data pulls, the first drafts, the basic modeling. What's left for the human, even at entry level, is the hard part: judgment, strategy, knowing what matters.</cite>
The result: <cite index="9-8,9-9">Only 22% of employers provide formal, mandatory AI training for all employees, and another 23% offer it only to specific departments, leaving most workers with fragmented optional resources or nothing at all.</cite> Young workers are asked to think like seniors before they've had the chance to learn the basics.
The Skills Mismatch Is Becoming the Real Problem
<cite index="9-1">31% of employers have raised experience requirements for entry-level jobs.</cite> But where do new grads get that experience if entry-level jobs themselves are disappearing?
<cite index="5-2">Firms that rebuild a junior on-ramp around AI oversight rather than eliminating junior roles will have a talent pipeline the rest of the market is quietly losing.</cite> Some organizations are starting to figure this out. <cite index="18-6">Even with widespread use of AI, one major enterprise hired 25,000 fresh graduates in 2025 and expects to exceed that number in 2026.</cite>
But that's the exception. <cite index="12-7,12-8">Nearly 4.6 million students who wanted internships could not secure one, which further limits opportunities for new graduates to gain experience and is impacting their access to employment.</cite>
What New Entrants Actually Need to Do
The Stanford study doesn't prescribe a fix, but the data suggests what works. If companies are automating the grunt work and demanding judgment-level thinking from day-one hires, young professionals need to lean into what AI can't do: critical evaluation, communication, problem-solving, and business acumen.
Here's the practical shift: <cite index="23-1,23-2">Learn to use mainstream tools like ChatGPT, Gemini, Claude, and Copilot as everyday assistants. In software roles, practice AI-assisted coding, test generation, and using AI to learn new languages quickly.</cite>
But more important: <cite index="21-10,21-11,21-12,21-13">Hiring managers are increasingly prioritizing problem-solving skills (can you understand a business problem and design an effective solution?), AI literacy (can you use AI tools productively without becoming overly dependent?), code review capabilities (can you identify mistakes in AI-generated output?), and communication skills.</cite>
What This Means for Hiring and Your Career
The Stanford data confirms a painful truth: AI isn't waiting for the policy debates. It's already reshaping the labor market in real time, and the pain is concentrated exactly where it hurts most—workers trying to break into their first real role.
If you're building a team, <cite index="5-2">watch whether companies respond by redesigning entry-level roles around AI oversight rather than eliminating them—the firms that rebuild a junior on-ramp will have a talent pipeline the rest of the market is quietly losing.</cite>
If you're starting out, stop waiting for the "entry-level" job as your parents knew it. <cite index="23-3,23-4,23-5">Artificial intelligence is already automating many repetitive junior tasks, but entry-level jobs are shrinking and being redesigned, not vanishing entirely.</cite> The jobs that remain will expect more from you faster—and that's actually your advantage if you prepare.
Right Now: Audit What You Actually Do With AI
If you're in a tech role or learning for one, stop and document: What parts of your work does AI already do well? What requires your judgment? That boundary line is where you need to get good. If you're using ChatGPT or Claude for work, you're already in practice. Push beyond "generate something." Focus on "verify, critique, refine, and explain why this version works." Tools like NeonCodex AI can help you practice working with AI-generated output at scale, testing your ability to evaluate and improve what models produce.
The career advantage isn't in competing with AI. It's in becoming the person who makes AI work.
Source: [Ars Technica](https://arstechnica.com/ai/2026/08/ai-is-hitting-entry-level-jobs-hardest-stanford-study-finds/)