[The End of Education] Entry-Level Jobs Vanish as the ‘Superhuman Labor’ Premium Widens, Accelerating Labor-Market Polarization

AI absorption of entry-level tasks narrows career pathways for young workers
Experienced professionals combine industry expertise with verification skills to drive productivity
Headcount-based personnel systems recede as individual performance assessments proliferate

Generative artificial intelligence (AI) is reshaping the labor market unevenly across different levels of experience. As AI absorbs work traditionally assigned to early-career employees, including research and basic coding, young people are losing opportunities to acquire practical experience. Seasoned professionals, meanwhile, are emerging as “superhuman labor,” using AI proficiently to boost productivity and handle workloads previously divided among multiple employees. As lean teams begin matching the output of large organizations, companies are placing greater emphasis on individual processing speed and output quality than on headcount. AI is also quantifying disparities in employee capabilities, transforming the criteria used to determine hiring, placement, promotion and compensation.

Hiring Freeze Starts at the Entry Level

According to the World Economic Forum’s (WEF) “AI and the Future of Entry-Level Work,” released on August 12, barriers to youth employment have risen sharply across advanced economies including the United States, the United Kingdom and Sweden. Companies adopting AI are reducing graduate and internship recruitment before resorting to large-scale layoffs of existing employees. AI has taken over foundational work through which new hires traditionally gained experience, including research, drafting, basic coding and administrative tasks. Experienced workers are consequently becoming more productive, while young people are losing access to the labor market itself. An analysis by Stanford University researchers found that employment among Americans aged 22 to 25 fell 16% in occupations with high levels of AI adoption after ChatGPT was launched in November 2022.

Concerns are mounting that AI could exacerbate youth unemployment in developing economies by reducing labor demand in manufacturing and services before sufficient jobs have been created. Countries dependent on business process outsourcing (BPO) industries such as call centers and data entry face particularly acute risks. The International Monetary Fund (IMF) warned last year that 36% of the workforce in the Philippines’ BPO industry was exposed to AI, leaving routine service-sector jobs at the greatest risk of elimination. Connie Bayudan-Dacuycuy, a senior research fellow at the Philippine Institute for Development Studies (PIDS), said, “AI can alleviate shortages in aging workforces across advanced economies, but it may intensify labor surpluses in developing countries with large youth populations. By reducing the number of jobs available, it could push young people into low-productivity domestic employment or migrant labor overseas.”

$28 Billion in Labor Income Erased

The decline in entry-level recruitment is particularly concerning because it also severs the pathways through which young workers build careers. First jobs have traditionally served as training grounds where employees learn basic professional skills, organizational practices and methods of dealing with clients. Experience accumulated through repeated report writing, data analysis and basic coding provided the foundation for advancement into middle management and skilled positions. As AI absorbs this process, companies are cutting recruitment of new employees who require training and favoring experienced candidates capable of delivering immediate results. Young people cannot find jobs through which to build experience, while companies refuse to hire them because they lack experience, creating a self-perpetuating cycle. The longer entry into the labor market is delayed, the greater the likelihood that initial disparities in income, promotion prospects and job security will accumulate throughout an individual’s working life.

Anxiety surrounding the proliferation of AI is also affecting the wages and job security of existing workers. Apollo Global Management analyzed 321 occupational categories using Anthropic’s AI usage data and statistics from the US Bureau of Labor Statistics (BLS). The analysis found that real wage growth in occupations with high AI exposure had lagged that of non-exposed occupations by 6.7 percentage points since 2023. Among low-income occupations, the gap widened to 10.7 percentage points. Although employment levels showed no statistically significant change, the failure to translate productivity gains into wages is estimated to have reduced the labor income of approximately 5.8 million US workers by at least $28 billion annually.

Table 1. The “Digital Luddite Movement” Targeting Corporate AI Adoption

AI Anxiety Fuels “Digital Luddism”

Fear of wage pressure and job losses is escalating into a “digital Luddite movement” aimed at deliberately disrupting corporate AI strategies. A survey of 2,400 knowledge workers in the United States and Europe by enterprise AI company Writer and workplace research organization Workplace Intelligence found that 29% of respondents had intentionally obstructed their employer’s adoption of AI. Among Generation Z—those born between 1997 and 2012—the figure reached 44%. Younger employees proficient in AI also accounted for a disproportionately large share of resistance to its adoption, challenging conventional assumptions about generational receptiveness to technology.

Resistance has taken the form of confidential-data disclosure, circumvention of internal policies and deliberate performance sabotage. Some employees entered internal information into external AI tools that had not been approved by their companies or refused to use mandatory in-house programs. There were also documented cases of employees submitting low-quality output to make AI systems appear less capable. Employees deliberately undermined the productivity gains anticipated by their employers, diminishing the effectiveness of AI investment. Among respondents who acknowledged engaging in obstruction, 30% cited fears that AI could replace their jobs. Industrial-era workers smashed textile machinery; their AI-era counterparts are resisting by degrading the quality of data and output.

AI Proficiency Determines Market Value

The extreme productivity disparities created by AI lie at the root of this resistance. As a small cohort of AI-proficient workers takes on workloads previously handled by multiple employees, companies are prioritizing individual output over workforce size. This shift has been driven by the emergence of superhuman labor, in which a single worker combines multidisciplinary expertise with AI operating skills to perform tasks spanning several occupational categories. In an experiment involving 791 employees of Procter & Gamble (P&G), researchers at Harvard Business School (HBS) found that an individual using AI achieved performance comparable to a two-person team working without it. The long-standing assumption that larger workforces deliver greater execution capacity has consequently begun to erode.

Even when employees used the same AI, performance varied sharply according to their professional expertise and ability to verify its output. In an experiment involving 758 Boston Consulting Group (BCG) consultants, participants using AI completed 12.2% more of the 18 tasks that fell within the model’s capabilities and worked 25.1% faster. On tasks requiring complex managerial judgment, however, their probability of reaching the correct answer fell by 19 percentage points. Excessive trust in plausible but incorrect AI-generated answers reduced performance. Professional expertise—the capacity to identify a model’s limitations and determine when to stop using it—has therefore widened productivity gaps among employees.

Companies are converting these documented individual disparities into human-resources data. Amazon measures the workloads of fulfillment-center employees in real time and identifies low-productivity workers as potential candidates for dismissal. At Uber and Lyft, algorithms determine driver assignments, earnings and account suspensions. AI proficiency is also becoming a performance metric for white-collar employees. JPMorgan Chase records employees’ use of AI tools on an internal dashboard and classifies them as “power users,” “low-frequency users” or “non-users.” Meta has established the share of AI-assisted code and tool-utilization rates among developers as performance objectives, while Google has made the use of AI tools a job requirement for some developers. The volume of work completed with AI and the amount of working time saved are increasingly being accumulated as personnel data influencing promotions and compensation.

Consulting Industry’s Revenue Model Begins to Fracture

As small teams equipped with AI perform work that previously required dozens of employees, the consulting industry’s revenue formula—built on deploying large workforces as a competitive advantage—is also beginning to fracture. Consulting firms have traditionally billed clients according to the number of personnel assigned to a project and the hours they worked. Large firms mobilized hundreds of consultants and low-cost offshore personnel simultaneously, presenting scale as a core competitive strength. AI has now reduced the time required for research, analysis, coding and report writing, eroding the equation between headcount and execution capacity.

New York-based capital-markets technology consultancy 28Stone Consulting, for example, has only 230 employees and repeatedly lost bids for large projects that previously required 300 to 400 personnel. Clients recognized its expertise but chose providers with larger workforces and offshore development centers. By using AI to reduce the time required for requirements analysis, coding and testing, 28Stone has now acquired the execution capacity to compete with major consultancies such as Accenture. The benchmark for assessing project scale has shifted from personnel deployment to delivery speed and output volume. 28Stone currently executes projects for major financial institutions through a model in which industry specialists verify AI-generated work and oversee testing and on-site deployment.

As AI reduces the personnel and time required for consulting work, pricing structures premised on headcount are also changing. Completing the same work more quickly with AI reduces the billable hours that can be charged to clients. Under this formula, greater productivity translates into lower revenue. Clients have begun incorporating cost savings, revenue growth and reductions in processing time into contracts, with fees determined by the extent to which agreed targets are met. Firms receive additional compensation when they achieve their objectives and accept lower fees when performance falls short of expectations. McKinsey currently derives 25% of its global fees from these performance-linked contracts. BCG applies variable fees to three-quarters of its large AI projects. The rapid decline in project estimates following AI adoption is also delaying contract signings. Clients calculate that even late contracts can meet their original completion dates because development periods have shortened, while costs may fall to half of the initial estimate.

The transformation of the revenue model has also changed the talent sought by consulting firms. Translating clients’ demands for cost reductions and shorter processing times into functioning systems requires data scientists, engineers and machine-learning specialists. A Financial Times (FT) analysis of more than 50,000 English-language job advertisements posted by Deloitte, EY, KPMG and PwC found that the share requiring AI skills rose nearly fourfold, from less than 2% in 2022 to approximately 7% in 2025. Vacancies for audit personnel accounted for less than 3%. Paul Griggs, chief executive officer (CEO) of PwC US, said the firm’s recruitment of consultants and accountants this year remained below the level recorded two years earlier, while hiring of data scientists, engineers and machine-learning specialists had increased. BCG also recruited AI engineers, data scientists and IT architecture specialists last year and assigned approximately 4,000 employees to advanced coding and automation work.