Photo by Vitaly Gariev on Unsplash
Sixteen percent. That is how much employment has fallen — in relative terms — for workers aged 22 to 25 in AI-exposed occupations since ChatGPT launched in late 2022. As of April 2026, according to reporting by Google News and Fortune, that figure comes from one of the most rigorous payroll datasets ever applied to this question: 4.6 million workers, roughly one in six Americans in the workforce, tracked across more than 730 job categories through a newly launched research project called the Stanford Canaries Dashboard.
The economist behind the finding, Erik Brynjolfsson of Stanford's Digital Economy Lab, has spent years documenting the trend. His summary, as of June 28, 2026: "Whatever it is, it's not going away."
The Evidence
The Canaries Dashboard — a partnership between Stanford's Digital Economy Lab and ADP Research, launched in June 2026 — was named deliberately. Early-career workers, the researchers argue, are the canary in the coal mine for AI's broader labor market effects. The dataset covers nearly four years of post-ChatGPT payroll records across 730-plus occupations through April 2026.
The numbers are stark. As of April 2026, employment for workers aged 22 to 25 in AI-exposed occupations is declining at 3.8% annually — up from a 2.8% annual drop recorded in April 2024. That acceleration compounds into the headline: a 16% relative employment decline for this age group since ChatGPT's public launch. Stanford co-author Bharat Chandar stated: "Our findings are consistent with the hypothesis that AI is having this effect, especially for entry-level workers."
At the same time, workers aged 30 and older in those same high-AI-exposure fields saw employment grow 6% to 12% between late 2022 and May 2025. The same technology, the same industries — opposite outcomes by age.
Chart: Employment change for early-career vs. experienced workers in AI-exposed occupations since ChatGPT's launch. The divergence reflects AI's tendency to automate task-level work while augmenting judgment-heavy work performed by more experienced employees.
Goldman Sachs adds macro-level precision. As of June 2026, AI is cutting a net 16,000 U.S. jobs per month — 25,000 eliminated, roughly 9,000 added through augmentation-driven roles — with Gen Z bearing the heaviest share of displacement. Goldman Sachs economist Elsie Peng's analysis identifies why: Gen Z workers tend to concentrate in "routine, white-collar, and administrative roles — data entry, customer service, legal support, billing" — the exact categories where AI performs most reliably.
A GMAC survey of more than 600 global recruiters, over half from Fortune 100 or 500 companies, puts a number on employer behavior: 33% of companies openly admit replacing entry-level jobs with AI, with the technology sector at 40%. Entry-level job postings are down 35% since early 2023. Workers aged 22 to 27 carried a 5.7% unemployment rate as of late 2025, against a 4.2% national average. Among the Class of 2026, 89% express concern that AI could replace their entry-level roles — up from 64% just one year earlier. BLS projections show each 10-percentage-point increase in AI exposure correlates with a 0.6-point reduction in projected employment growth. A Goldman Sachs analysis further found that a one-standard-deviation increase in AI substitution exposure widened the entry-level-to-experienced wage gap by 3.3 percentage points.
The Paradox Nobody Is Saying Out Loud
Entry-level jobs exist, historically, not because employers are feeling generous — they exist because every experienced worker was once inexperienced. The core task stack of most entry-level white-collar work (retrieving information, summarizing, scheduling, formatting, routing requests) is not just employment. It is a training ground. Junior workers do this work to build the contextual judgment and institutional knowledge that eventually makes them valuable mid-career contributors.
When AI automates that task stack, the consequence is not only fewer entry-level jobs. It is the dismantling of the mechanism by which junior workers accumulate the experience that transforms them into senior ones. MIT AI researcher Andrew McAfee has flagged this dynamic directly: automating Gen Z's starting roles risks destroying corporate talent pipelines — the damage showing up not in this quarter's earnings but in the skill gaps of five years from now. The market does not process that irony on a quarterly call. But it is real, and the compounding effect on personal finance — starting salary compression, delayed wealth accumulation, extended student loan repayment periods — is significant.
ADP Chief Economist Nela Richardson offered the clearest framing of the data: "Occupations where AI augments human work show employment growth; those where AI automates tasks show contraction." The complication is that augmentation tends to favor workers who already carry tacit knowledge, organizational relationships, and domain expertise — none of which arrive with a degree.
This age-based labor market split echoes a pattern that AI Trends documented in its enterprise AI adoption analysis: 88% of companies have deployed AI tools, but only 39% report actually winning with them — meaning the disruption is broad even when the productivity gains are uneven. The displacement reaches young workers regardless of whether companies are capturing returns.
Photo by Daniil Komov on Unsplash
The Counter-Evidence Worth Naming
Intellectual honesty requires naming where the data is contested. A Federal Reserve study of more than one million firms found what researchers described as "precisely-estimated null effects" — no statistically significant link between AI adoption and reduced job postings at the aggregate firm level. That finding stands in direct tension with the Stanford/ADP numbers and deserves to be on the table.
Some economists point to non-AI explanations for the early-career crunch. A 2022 U.S. tax law change required companies to amortize research and development salary costs over five years rather than deduct them immediately, tightening hiring budgets across the board. The Federal Reserve's rate hikes from near-zero in the 2020-to-2022 period added additional fiscal pressure. The United Kingdom saw a 46% decline in tech graduate roles in 2024 — on a near-identical rate-hike trajectory — raising a fair question about how much is AI and how much is macro tightening.
There is also a documented "AI washing" dynamic. A Resume.org survey found 59% of companies admit to emphasizing AI's role in layoffs because, as respondents framed it, it "plays better with stakeholders than citing financial constraints." Marc Andreessen described this as a "silver-bullet excuse." Marc Benioff called it the "lazy way out."
The honest position: the Stanford/ADP correlation between AI exposure and early-career employment decline is well-documented and logically coherent. Pinning AI as the sole cause is harder. For someone doing active career planning and financial planning in mid-2026, the practical implication is the same either way.
How to Act on This
The Canaries Dashboard gives young workers something more useful than a warning — it gives them a map. The employment decline concentrates in high AI-substitution occupations: data entry, document review, scheduling, and routine customer routing. The growth concentrates in roles where AI augments rather than replaces: technical implementation, licensed judgment work, and relationship management that requires human presence and context.
A job posting where "proficiency in Excel" is the primary differentiator sits firmly in the high-substitution zone. A role requiring stakeholder navigation, domain expertise, or translating ambiguous information into a recommendation sits lower. One concrete question for any interview: "What does AI currently handle on this team, and what work genuinely requires a human judgment call?" That answer tells you more about your long-term runway than the listed salary range. If the interviewer struggles to answer, that itself is a data point.
Getting inside an organization is step one. Once there, the goal is to load your work portfolio with tasks that AI cannot cheaply replicate. The script: "I want to build skills that are hard to automate. What is the work on this team that really requires human judgment — the complex calls, the ambiguous decisions, the relationship-dependent situations? I would like to take that on." Most managers have not encountered this framing. The ones who respond well are worth working for; the ones who look blank are telling you something about the role's future.
With 43% of new graduates underemployed as of mid-2025, competition for entry-level roles is intense. The workers growing in the Stanford data are those in augmentation roles, where AI makes them faster and more accurate — not automation roles, where AI makes their position redundant. When pivoting between sectors, aim toward healthcare coordination, skilled trades, technical implementation, and any work requiring a license or physical presence. The interview framing: "I am specifically interested in roles where AI makes me more effective, not roles where AI eventually replaces me. What does that look like on your team?" Most interviewers will not have a ready answer. That gap is where the real conversation starts.
Bottom Line
In my analysis, the Stanford/ADP Canaries Dashboard is the most credible early-career AI impact dataset in circulation right now — 4.6 million workers, nearly four years of post-ChatGPT payroll records, 730-plus occupations. A 16% relative employment decline for workers aged 22 to 25 is not statistical noise, and the acceleration from a 2.8% annual drop in April 2024 to 3.8% in April 2026 suggests the trend is not self-correcting. Brynjolfsson's summary holds.
The deeper issue is structural. The career ladder is not disappearing — the bottom rung is being removed. That has compounding consequences for individual financial planning: delayed entry into well-paying work, compressed starting salaries, extended student debt repayment timelines, and deferred wealth accumulation. Treating career choice as risk management, not just passion pursuit, has always been sound advice. Right now, in June 2026, it is essential.
- As of April 2026, employment for workers aged 22–25 in AI-exposed occupations is falling at 3.8% annually — a 16% relative decline since ChatGPT launched in late 2022.
- Workers aged 30 and older in the same AI-exposed fields saw employment grow 6–12% over the same period, creating a stark age-based divergence in who the labor market rewards.
- Entry-level job postings are down 35% since early 2023; 33% of employers admit replacing entry-level positions with AI, with the technology sector at 40%.
- The practical response: pursue augmentation roles over automation-risk roles, negotiate for scope in your first 90 days, and ask explicitly where AI sits in any role you consider.
Frequently Asked Questions
How does AI specifically affect entry-level jobs for new graduates in 2026?
AI primarily targets the routine task stack that defines most entry-level white-collar jobs: data entry, scheduling, document review, customer routing, and basic formatting. Stanford's Canaries Dashboard, drawing on 4.6 million ADP payroll records through April 2026, documented a 16% relative employment decline for workers aged 22–25 in AI-exposed occupations since late 2022. The impact concentrates in white-collar roles where tasks are routine and well-defined — not in work requiring licensed judgment, physical presence, or accumulated contextual knowledge.
What entry-level jobs are actually safe from AI automation right now?
Roles requiring physical presence, licensed judgment, complex stakeholder navigation, and domain expertise built over time carry lower AI-substitution risk. Healthcare coordination, skilled trades, technical implementation, and client-facing relationship work fall into this category. ADP Chief Economist Nela Richardson noted that occupations where AI augments human work show employment growth — so the strategic goal is to find the augmentation side of any industry, not to avoid AI entirely. The distinction is between roles where AI is a tool you operate versus roles where AI is a replacement for your position.
Is AI permanently closing Gen Z's entry-level career pathways?
The Stanford data shows a persistent and accelerating decline — not a temporary correction. Employment for workers aged 22–25 in AI-exposed roles worsened from a 2.8% annual drop in April 2024 to 3.8% by April 2026. That said, MIT AI researcher Andrew McAfee has argued that companies automating entry-level work risk destroying their own long-run talent pipelines. A Federal Reserve study of over one million firms also found no significant aggregate link between AI adoption and reduced job postings, complicating the causal story. The practical implication for career and financial planning is to treat the narrowing window as real and act accordingly — rather than waiting to see how the academic debate resolves.
Disclaimer: This article is for informational and educational purposes only and does not constitute financial or career advice. Individual circumstances vary; consult a qualified professional for personalized guidance. Research based on publicly available sources current as of June 28, 2026.