Is Generative AI Hollowing Out the Entry Level? New Data Suggests a 'Seniority-Biased' Shift
A Harvard paper offers concerns about the bottom rungs of corporate ladders
A recent working paper offers a sobering, data-driven look at the initial labor market impacts of generative AI, suggesting that its effects are not evenly distributed across the corporate ladder. The research, posted on August 31, 2025, analyzes a massive dataset of U.S. résumés and job postings and finds that since early 2023, firms actively adopting AI have significantly reduced their hiring of junior workers while continuing to expand their senior ranks. This challenges the narrative of AI as a universal productivity tool and points toward a more disruptive, "seniority-biased" form of technological change.
The Harvard paper, titled "Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data," directly addresses many growing suspicions surrounding AI's effect on early-career professionals. While some studies suggest AI complements less-experienced workers, making them more productive, anecdotal evidence has raised alarms about the automation of entry-level tasks.
The authors frame the stakes clearly, stating, "If AI disproportionately targets entry-level tasks, the bottom rungs of these ladders may be eroding" (Lichtinger, p. 2). Their research provides some of the first large-scale empirical evidence that this erosion may already be underway.
The Problem: Eroding the "Bottom Rungs" of the Career Ladder
The paper extends the classic economic literature on skill-biased technological change, which focused on how technology favored educated or skilled workers over less-skilled ones. The authors argue for a new dimension: seniority.
Even within high-skill, white-collar professions, junior employees are often responsible for the "intellectually mundane tasks... routine yet cognitively demanding activities such as debugging code or reviewing legal documents, which are likely to be especially exposed to recent advances in AI" (p. 2).
The concern is that if these entry points vanish, it could have severe long-term consequences for career progression and economic inequality. The authors articulate this risk:
The stakes go beyond short-term job losses. A large share of college graduates' lifetime wage growth comes from within-firm advancement, starting in low-paid entry roles (Deming, 2023). Early-career earnings also have long-lasting effects on inequality... If AI disproportionately affects junior positions, it could have lasting consequences for the college wage premium, upward mobility, and income disparities. (p. 2)
This perspective is crucial for industries that rely on an apprenticeship model, including law, consulting, and technology. If the roles that serve as training grounds for future experts are automated away, firms may face a critical pipeline problem for senior talent in the years to come.
Tracking Real-World AI Adoption and Its Consequences
To move beyond speculation, the researchers developed a novel method to identify which firms were actively implementing generative AI and then tracked their employment patterns over time.
Identifying "AI Integrator" Roles
The study's core innovation is its method for measuring AI adoption. Instead of relying on broad, occupation-level "exposure" scores, the authors pinpointed firms making tangible moves to implement the technology. They did this by "detecting job postings that explicitly recruit 'AI integrator' roles" (p. 3). This two-step process involved first flagging job postings with AI-related keywords (like "LangChain," "RAG," or "prompt engineering") and then using a large language model to confirm that the role was genuinely focused on implementing or operating AI systems within the company.
In today’s colloquialism, the researchers found the "receipts"—the job ads proving a firm was putting money and manpower into integrating AI.
This method identified 10,599 "adopter" firms, or about 3.7% of their sample, providing a clear basis for comparison against non-adopters.
A Widening Employment Gap
Using their statistical models, the paper compares workforce changes at adopting firms versus non-adopting firms. A difference-in-differences (DiD) analysis revealed a stark divergence beginning in the first quarter of 2023. While junior employment trends were parallel for both groups before 2023, "junior headcount at adopting firms fell sharply relative to controls declining by 7.7 percent after six quarters" (p. 4). In contrast, senior employment at these same firms continued to grow, showing no negative impact from AI adoption.
A more rigorous "triple-difference" model, which controls for firm-specific shocks, confirmed the result. It found that the gap between junior and senior employment within adopting firms widened significantly right as generative AI became widespread. The evidence suggests that AI adoption isn't just affecting overall headcount, but is actively reshaping the seniority composition of the workforce.
Examples: A Closer Look at the Mechanisms
The paper goes beyond the top-line finding to explore how these changes are happening and who is most affected.
It's a Hiring Problem, Not a Firing Problem
One of the most significant findings is that the decline in junior positions is not due to layoffs. Instead, it "is driven primarily by a sharp slowdown in hiring after 2023Q1" (p. 4). AI-adopting firms hired, on average, 3.7 fewer junior workers per quarter compared to non-adopters. This suggests a strategic shift: as AI tools become available to handle routine tasks, firms are simply not opening as many entry-level positions. Interestingly, the data also shows a slight
increase in promotions for existing junior staff in adopting firms, suggesting that "while fewer juniors are being recruited, those who remain may face enhanced opportunities for internal advancement" (p. 20).
Wholesale and Retail See the Biggest Hit
The impact is widespread but uneven. The researchers analyzed hiring changes across several major sectors and found that while junior hiring fell across the board in adopting firms, the effect was most dramatic in an unexpected area.
The largest reduction occurred in wholesale and retail trade, where adopting firms hired roughly 40% fewer juniors per quarter than non-adopters. This pattern may reflects the greater substitutability of junior tasks in these sectors with generative AI tools, which can automate routine communication, customer service, and documentation. (p. 4)
This finding underscores that the effects of generative AI are not confined to the tech sector but are already being felt in industries where communication and documentation are core functions.
The Vulnerable Middle: A U-Shaped Effect by Education
Perhaps the most nuanced finding relates to the educational background of the junior workers affected. The analysis revealed a clear "U-shaped pattern" when looking at the prestige of the universities from which juniors graduated.
The steepest relative declines occur among graduates of strong (Tier 2) and solid (Tier 3) schools, while the declines are smaller for juniors from elite (Tier 1) and less selective (Tier 4) institutions. Interestingly, the smallest, and statistically insignificant, decline occurs for graduates of the lowest tier (Tier 5). (p. 5)
The authors suggest this reflects a trade-off between quality and cost. Elite graduates may be seen as too productive to replace, while graduates from the lowest-tier schools are inexpensive enough to retain. It is the large group in the middle—those with moderate skills and moderate salaries—who "appear most vulnerable to substitution, experiencing the steepest reductions in employment" (p. 23).
Thoughts and Implications
This research provides compelling early evidence that generative AI is acting as a seniority-biased technological change, narrowing the entry points into professional careers. While the authors are careful to note their study's limitations—AI adoption is not random, and the analysis covers a relatively short time frame—the findings present a clear warning signal.
For inventors, IP professionals, and patent attorneys, the implications are significant. The traditional model of hiring junior associates or agents to handle routine tasks like document review or prior art searches may be fundamentally changing. One might wonder what the IP contractor, service provider, and outsource industries will look like in the near future.
The paper's findings have particularly sharp implications for law firms, an environment built on a leveraged apprenticeship model. For decades, the career path in "Big Law" and corporate legal departments has started with junior associates performing tasks that the paper explicitly identifies as vulnerable: "intellectually mundane... yet cognitively demanding activities such as... reviewing legal documents" (p. 2). These activities, including due diligence, discovery, and legal research, form the bedrock of a young lawyer's training but are also prime candidates for automation by generative AI.
The study suggests that the coming disruption may be subtle but profound. Law firms are unlikely to announce mass layoffs of first-year associates. Instead, the impact will likely manifest as a "sharp slowdown in hiring" (p. 4), leading to smaller summer associate programs and incoming first-year classes.
Firms that adopt AI integration roles will find they can accomplish the same volume of work with a leaner team, reducing the need to hire large cohorts of junior talent primarily for leverage. This aligns with the finding that the adjustment happens through "slower entry rather than layoffs" (p. 24).
This trend presents a potential long-term crisis for talent development. If the foundational "bottom rungs" of the legal career ladder are eroded (p. 2), it begs the question of how future partners and senior counsel will be trained. The rote work of document review is where associates learn to spot issues, understand contractual language, and develop the instincts that define legal expertise.
Without this hands-on experience, firms will need to invent new ways to cultivate legal judgment. The paper’s observation that promotions for existing juniors increased in AI-adopting firms (p. 20) may offer a clue: firms might pivot to investing more heavily in a smaller, more elite pool of associates, fast-tracking their development into senior roles.
Furthermore, the "U-shaped pattern" of impact based on university prestige could reshape legal recruiting (p. 5). Graduates from the most elite (Tier 1) law schools and universities may remain in high demand for their perceived pedigree. At the same time, graduates from lower-tier schools might be seen as a cost-effective option for tasks that still require human oversight—e.g., humans reviewing AI-produced work.
The greatest pressure could fall on the vast middle: graduates from well-regarded but non-elite national and regional law schools. These individuals, who have historically formed the backbone of large associate classes, may find themselves the most "vulnerable to substitution" as firms use AI to replace moderately expensive junior labor (p. 23).
For the legal profession, this seniority-biased shift isn't just about technology; it's a potential restructuring of the very pathway to becoming a seasoned legal expert.
Citation
Lichtinger, Guy and Hosseini Maasoum, Seyed Mahdi, "Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data" (August 2025). Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5425555.
Disclaimer: This is provided for informational purposes only and does not constitute legal or financial advice. To the extent there are any opinions in this article, they are the author’s alone and do not represent the beliefs of his firm or clients. The strategies expressed are purely speculation based on publicly available information. The information expressed is subject to change at any time and should be checked for completeness, accuracy and current applicability. For advice, consult a suitably licensed attorney and/or patent professional.