By Michael Phillips | Riptide
Something has hardened into consensus over the past year: artificial intelligence is gutting entry-level white-collar work, and companies that keep cutting junior positions are digging a hole they won’t be able to climb out of. The story shows up in trade press, LinkedIn posts, staffing-firm press releases, and a Sept. 24 piece in the Baltimore Business Journal that leans on three sources — all from staffing and recruiting firms — to describe a talent pipeline in crisis.
It’s worth pulling the thread, because a surprising amount of the “crisis” rests on numbers that don’t say quite what they’re being used to say.
What’s real
Start with what actually checks out. The Federal Reserve Bank of New York’s own data shows something genuinely unusual: unemployment for college graduates aged 22 to 27 has been running around 5.6 to 5.7 percent through 2026, above the rate for all workers. For most of the New York Fed’s series, which runs back to 1990, a degree conferred a clear edge — recent grads were consistently less likely to be unemployed than the general workforce, even during the Great Recession. That edge started eroding around 2019 and flipped entirely by 2021; by 2026 the gap has widened to its record level. That’s not spin, and it’s not a brand-new 2026 development — it’s a multi-year reversal that’s now hit its widest point on record. Independent economists who track it, like DataTrek’s Jessica Rabe, have called it a genuine break from historical pattern.
Venture firm SignalFire’s oft-cited numbers are real too: its 2025 report found new-graduate hiring at the 15 largest tech companies had fallen more than 50 percent since 2019, with new grads making up 7 percent of Big Tech hires, down from 15 percent before the pandemic. Those figures anchor nearly every “AI is eating entry-level jobs” story published in the last year. SignalFire’s newer, June 2026 report finds it’s gotten worse, not better: entry-level hiring at the same companies is now down roughly 65 percent since 2019, and roughly 76 percent at early-stage startups.
And there’s real academic research behind the trend, not just press releases. An IZA Institute of Labor Economics discussion paper by Northeastern economist Alicia Sasser Modestino and coauthors, using near-universe vacancy data from Lightcast, finds a 14 to 15 percent relative decline in junior-versus-senior software job postings since ChatGPT’s release — a real, measurable effect, concentrated in software and largely absent in a control industry like mechanical engineering. A separate Stanford Digital Economy Lab study — “Canaries in the Coal Mine? Six Facts About the Recent Employment Effects of Artificial Intelligence,” using actual ADP payroll records, not survey responses or job postings — found, in its most recent August 2026 revision, that employment among 22-to-25-year-olds in highly AI-exposed occupations now runs about 19 percent below where it would be had it kept pace with similarly aged workers in less-exposed occupations, up from a 15 percent gap a year earlier. Experienced workers in the same occupations show no comparable gap, and the researchers find the divergence comes almost entirely through reduced hiring of young workers, not increased firing of them. It’s worth being precise about what that number means: it’s a relative shortfall against a counterfactual trend, not a literal 19 percent drop in employment. And the study’s own authors are unusually careful about their conclusion, stating plainly that they find no evidence of widespread, economy-wide job displacement — they frame their findings as early, descriptive signals, not proof of causation.
“It’s a relative shortfall against a counterfactual trend, not a literal 19 percent drop in employment.”
So no, this isn’t nothing.
What’s being oversold
But look closer at how the “crisis” framing gets built, and the load-bearing numbers get shakier fast.
Start with SignalFire’s own newest report, and read it carefully, because it complicates a different claim than the one usually pulled from it. The 2026 update confirms the entry-level number got worse, not better — that part of the narrative holds up. But the same report finds that engineering hiring overall, across all experience levels, is down only 11 percent since 2019, and software engineers now make up 55 percent of new hires at the largest tech firms, up from 46 percent in 2019. SignalFire’s own conclusion is that the feared “AI Code Apocalypse” for engineers broadly never materialized — the hiring collapse hit design, product, and marketing functions far harder than engineering. So the honest version of the SignalFire data isn’t “the crisis is overblown.” It’s narrower and stranger: engineering as a discipline is fine, even resilient, while the on-ramp into it for new graduates specifically has kept collapsing. That’s a real distinction the trade coverage tends to blur.
“Engineering as a discipline is fine, even resilient, while the on-ramp into it for new graduates specifically has kept collapsing.”
Modestino’s research, the most methodologically serious piece in the mix, doesn’t find companies replacing junior developers with AI. It finds them quietly raising the experience bar on the same job titles — a shift-share decomposition shows the effect comes from asking for more years of experience, not from eliminating junior roles or retitling them. That’s a hiring-standards story as much as an automation story, and it undercuts the simpler “AI took the job” narrative BBJ’s sources gesture at.
The Stanford finding is real, but the researchers themselves are careful not to overextend it, and other work points in a more cautious direction. The Yale Budget Lab, in a May 7, 2026 analysis using a synthetic differences-in-differences method designed to make AI-exposed and unexposed occupations properly comparable, concluded it found “no strong evidence of impacts as of yet” — direct pushback on the idea that AI’s labor effects are already visible and measurable at scale. Earlier firm-level research summarized by Brookings, much of it examining AI investment that predates the generative-AI era, found the opposite of net job loss: firms investing intensively in AI grew total employment, including in engineering and administrative roles. A newer study by Ramp and Revelio Labs, covering more than 21,000 U.S. firms, found something sharper still — high-intensity AI adopters grew entry-level headcount specifically by 12 percent over two years, while firms that adopted AI only lightly saw no significant employment change at all. That finding has its own limitation worth naming, in fairness: the firms that adopted AI intensively were already larger, more technical, and faster-growing before they adopted it, so the study shows an association, not proof that AI adoption caused the hiring growth. But it cuts hard against the idea that intensive AI adoption necessarily means fewer junior workers. None of this made it into the trade coverage BBJ’s piece sits alongside.
“The study shows an association, not proof that AI adoption caused the hiring growth.”
Then there’s the number doing the most work in pieces like BBJ’s: a Robert Half finding, cited by the paper, that 41 percent of hiring managers say AI implementation reduced entry-level roles at their organization — with Robert Half’s Bill Driscoll quoted in the same section, though not directly attributed as the source of that figure. Robert Half’s own broader research, published the same year, tells a more complicated story: 54 percent of hiring managers expect a net increase in jobs from AI at their organization over the next two years, versus 23 percent who expect a decrease. Separately, the firm found roughly three in ten employers that eliminated a role because of AI later rehired for it — a reversal rate that hit 44 percent in finance. IBM’s chief human resources officer has said the company is tripling its U.S. entry-level hiring in 2026 rather than cutting it, explicitly citing the pipeline risk. That’s not a company acting on a belief that AI has already replaced its junior workforce.
Even the World Economic Forum’s widely repeated “40 percent of employers expect to reduce workforce” stat is routinely stripped of its context. The same Future of Jobs Report projects a net increase of 78 million jobs globally by 2030. It’s a reallocation number wearing a layoff number’s clothes.
“It’s a reallocation number wearing a layoff number’s clothes.”
The Amodei problem
And then there’s Dario Amodei, whose prediction that AI could eliminate half of entry-level white-collar jobs within one to five years anchors nearly every version of this story, BBJ’s included. It’s worth stating plainly what’s often left unsaid: Amodei is the CEO of Anthropic, a company with a direct financial stake in the technology whose disruptive potential he’s forecasting. He’s also the one who’s floated, unprompted, that AI companies “like us” might eventually need to be taxed to offset the disruption — an unusual thing for an executive to volunteer, and one that cuts against reading his forecast as either pure hype or pure candor. It’s a genuinely hard prediction to evaluate cleanly, made by someone forecasting the disruptive potential of the technology his company sells, who is also apparently uneasy enough about that disruption to propose taxing his own industry. That tension belongs in the story, not just the prediction.
The actual story
“Not an aggregate jobs apocalypse, but a career-ladder problem — a narrowing of the bottom rung even where overall employment holds steady or grows.”
None of this means AI isn’t changing how companies hire junior workers. The Modestino and Stanford research say it is, in a specific and measurable way: rising experience thresholds and a real, widening employment gap for the youngest workers in the most exposed occupations, running through hiring rather than firing. That’s worth taking seriously — and it points to a more precise question than the one usually asked. The debate isn’t really “is AI killing jobs” versus “is AI not killing jobs.” Stanford finds no broad economy-wide displacement, but a growing gap for young workers specifically. SignalFire finds engineering broadly resilient while its own entry-level pipeline keeps shrinking. Modestino finds employers raising experience requirements rather than deleting junior titles. Robert Half finds most employers expecting total headcount to grow while shifting the mix toward experienced hires. IBM’s own reasoning is telling here too: the company isn’t disputing that AI can do much of what its old entry-level jobs involved — it’s tripling entry-level hiring anyway, on the theory that those jobs need to be redesigned around AI rather than eliminated. Put together, that’s a consistent pattern: not an aggregate jobs apocalypse, but a career-ladder problem — a narrowing of the bottom rung even where overall employment holds steady or grows.
“The pipeline concern is legitimate. The sourcing behind the apocalyptic version of it isn’t as solid as the headlines suggest.”
But that’s a narrower and messier finding than “AI is closing the door on entry-level work,” and it’s not the story circulating in trade coverage right now. What’s circulating is a small set of contested numbers — several supplied by firms that sell staffing and AI-hiring services, one supplied by the AI industry’s own most prominent voice — that have calcified into consensus faster than the underlying research supports. The BBJ piece isn’t unusual; it’s typical. Three sources, all from companies with something to sell around the disruption they’re describing, presented as neutral labor-market analysis.
The pipeline concern is legitimate. The sourcing behind the apocalyptic version of it isn’t as solid as the headlines suggest.
Sources: Federal Reserve Bank of New York, “The Labor Market for Recent College Graduates” (2026 dashboard); SignalFire, “State of Tech Talent Report” (2025 and June 2026 editions); Samuel Westby, Alicia Sasser Modestino, and Peiran Cheng, IZA Institute of Labor Economics Discussion Paper No. 18723; Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen, “Canaries in the Coal Mine? Six Facts About the Recent Employment Effects of Artificial Intelligence,” Stanford Digital Economy Lab (August 2026 revision); Ryan Nunn, “AI Is Probably Not (Yet) the Reason for Labor Market Weakening,” The Budget Lab at Yale, May 7, 2026; Brookings Institution, “The Effects of AI on Firms and Workers,” citing Babina and Fedyk; Ramp and Revelio Labs, “A New Look at AI’s Impact on Jobs,” via ITIF’s summary; World Economic Forum, “Future of Jobs Report 2025”; Robert Half, “The AI Hiring Paradox” research brief, November 2025 hiring-manager survey, and April 2026 rehiring survey; reporting on Dario Amodei’s remarks via Axios, Fox News, and Anthropic’s public statements; Baltimore Business Journal, “AI tools threaten entry-level jobs and talent pipelines,” Sept. 24, 2026.
Keep This Reporting Free
If this work matters to you, please consider supporting it.
Your contribution helps fund independent reporting across our entire network.
Discover more from RIPTIDE
Subscribe to get the latest posts sent to your email.
