The Vanishing Entry Level

Executive Summary

Artificial intelligence (“AI”) is raising productivity by absorbing routine work, but much of that work has long been where inexperienced employees develop judgment, context, and professional capability. Employers that have eliminated entry-level positions without replacing the development those roles provided now face an emerging consequence of having traded future talent for short-term savings. While a smaller senior-level team can deliver today’s work at lower cost, those savings carry a compounding expense: a weakening pipeline of the specialists, managers, and leaders who will be critical to their organizations’ viability tomorrow.

This paper introduces the Two-Ledger Test, a framework for evaluating the workforce implications of AI deployment. The first ledger records the immediate operating gains AI delivers; the second records its effect on long-term workforce capability. The paper then presents Registered Apprenticeship as the preeminent model for growing the talent whose development AI displaces. Registered Apprenticeship deliberately builds the structured practice, feedback, context, and career progression that entry-level work has traditionally provided. Putting this model into practice requires expertise in occupational analysis, competency mapping, workforce development, and apprenticeship implementation.

For leaders, the implication is direct: before automating entry-level work, apply the Two-Ledger Test to identify the developmental value at risk, and use Registered Apprenticeship to rebuild the pathways through which inexperienced employees become tomorrow’s senior experts. The employers that successfully navigate this historic, AI-driven labor transformation will be those that keep both ledgers in balance. Rather than trading future talent for today’s cost savings, they will remain committed to cultivating the workforce their future depends on.

Entry-Level Work Does More Than Produce Output

Every entry-level role has an obvious purpose. A junior developer resolves tickets. An analyst cleans data. A service representative handles routine inquiries. An operations associate keeps a process moving.

Perhaps less apparent, these roles also turn inexperienced employees into capable ones. In other words, entry-level work is not merely a source of lower-cost output. It is part of the organization’s talent-production system.

A junior developer gradually learns how systems interact and where a seemingly small change can create operational risk. An analyst learns which patterns matter and which are misleading. A service representative learns when a routine problem is actually an exception that requires judgment.

Experienced workers are built through real assignments, feedback, and greater responsibility over time. Entry-level employment has long been one of the primary ways organizations create specialists, managers, and institutional knowledge holders.

When junior work disappears, the savings may be immediate, but the capability gap appears later, when the organization needs the experienced people it never developed.

Entry Routes are Narrowing

The evidence now warrants attention from employers.

The Stanford Digital Economy Lab, drawing on ADP payroll data covering millions of U.S. workers, found that employment among workers ages 22 to 25 in highly AI-exposed occupations stood about 19 percent below where it would have been if it had kept pace with employment among similarly aged workers in less-exposed occupations. The divergence has widened as employment outcomes for young workers in the most AI-exposed occupations have weakened relative to those in less-exposed work.[1]

The change is appearing primarily in hiring, not layoffs. Employers are not necessarily removing large numbers of junior employees already on the payroll. They are opening fewer doors for people trying to enter these occupations.

Firm-level research shows a similar pattern. Using resume data covering roughly 62 million U.S. workers at more than 285,000 firms, researchers found that at companies actively integrating generative AI, junior employment declined relative to non-adopting firms while senior employment trends remained largely unchanged. The decline was concentrated in occupations most exposed to generative AI and was driven primarily by slower hiring rather than increased separations or promotions.[2]

The Stanford researchers note that the pattern has persisted and widened. It is strongest in occupations where AI is used to automate human tasks and where work depends heavily on codified knowledge, meaning information that is formal, standardized, and documented. Where AI is used more to complement workers, employment is flat or rising, particularly among more experienced workers.[1]

The result is a more top-heavy workforce. Employers continue to rely on experienced talent while reducing the opportunities through which people acquire that experience.

Software Development Shows the Pattern Clearly

Coding is among the tasks most exposed to large language models. Federal Reserve Board researchers found that aggregate employment of coders decelerated sharply after the introduction of ChatGPT. They showed the slowdown could not be explained by coders being concentrated in industries that were already weakening, which points to an occupation-specific shock. Coder employment continued to grow, but much more slowly than before 2022.[3]

Hiring data shows how strongly the software market favors experience.

Software development had the lowest entry-level share and the highest senior-level share among the sectors analyzed by Indeed Hiring Lab.[4] Over time, this hiring pattern may create a workforce increasingly dependent on experienced engineers whose replacements are not being developed at the same rate.

Together, the employment and hiring data show what a thinning entry pathway looks like. Employers still need software engineers, but they increasingly favor people who already possess the judgment needed to direct AI tools, evaluate their output, and manage operational risk. If fewer workers receive opportunities to develop those capabilities, today’s preference for experienced talent will become tomorrow’s shortage of it.

Customer Service Shows the Pattern Beyond Technology

Customer service provides a very different example. It is a large, cross-industry occupation that typically requires a high school diploma, no related work experience, and short-term on-the-job training. The Bureau of Labor Statistics counted about 2,666,000 customer service representatives in 2025. BLS projects a 5 percent decline from 2025 to 2035, even as about 289,500 openings are projected each year on average. Nearly all are from the need to replace workers who transfer to other occupations or exit the labor force.[5]

Customer service is also highly exposed to generative AI. In an Anthropic analysis of approximately 800 U.S. occupations, customer service representatives ranked second on a measure of observed AI exposure, behind only computer programmers. The measure reflects how extensively occupational tasks appear in automated, work-related AI use.[6] The U.S. Bureau of Labor Statistics identifies this Anthropic analysis as one of its observed-evidence sources for occupational AI exposure in its 2025–35 employment projections.[7]

That exposure matters for the developmental pathway. A peer-reviewed study of customer-support agents at a Fortune 500 firm found that AI assistance increased productivity by 15 percent on average, with the largest gains among less-skilled and less-experienced workers. Agents with two months of tenure who used the tool performed as well as, or better than, agents with more than six months of tenure who did not.[8]

Customer service illustrates both the promise and the risk. AI assistance can help less-experienced employees reach acceptable performance sooner, but faster AI-assisted performance does not necessarily demonstrate independent judgment. Employers must still determine whether workers are learning to diagnose exceptions, evaluate recommendations, and act when the system is incomplete or wrong.

AI may accelerate performance without accelerating judgment.

The Seniorization Paradox

AI is also changing what employers expect from the entry-level positions that remain.

PwC’s 2026 Global AI Jobs Barometer revealed that highly AI-exposed U.S. entry-level roles were seven times more likely than the least-exposed roles to require capabilities traditionally associated with experienced workers, including strategic decision-making, mentorship, stakeholder management, creativity, and face-to-face interaction.[9]

The jobs that made this shift performed much better in the labor market. Between 2019 and 2025, AI-exposed entry-level roles that added a substantial number of traditionally senior skills (what PwC calls “seniorized” roles) grew 35 percent, while comparable roles that did not make that shift declined 10 percent.[9]

Employers increasingly expect entry-level workers to arrive with senior-level capabilities while simultaneously automating the experiences that traditionally produced those capabilities.

Employers are raising the capability threshold for entry-level work while automating many of the tasks through which those capabilities were traditionally developed.

Expecting these capabilities earlier does not eliminate the need to develop them. It shifts more of the burden onto workers while increasing the risk that employers will later face shortages of qualified, experienced talent.

What a Redesigned Entry Pathway Must Provide

The practical question is not whether employers should preserve every junior task. It is whether redesigned work still gives less-experienced employees the conditions required to build independent capability.

Four functions are especially important:

• Practice: repeated opportunities to apply knowledge in real working conditions.
• Feedback: correction and guidance from managers, mentors, customers, and quality processes.
• Context: exposure to exceptions, relationships, operational constraints, and institutional knowledge.
• Evidence of progression: credible proof that a worker is ready for greater responsibility.

Employees can produce acceptable work with AI assistance without necessarily becoming more capable. A healthy entry pathway must do both. It must help people contribute while steadily increasing what they can understand, evaluate, and own independently.

The Two-Ledger Test

Most AI business cases focus on immediate operating results. Those measures matter, but they are incomplete when the work being redesigned is also the work through which future talent is developed.

The Two-Ledger Test asks employers to assess both outcomes.

Employers can put the Two-Ledger Test into practice by asking five questions whenever AI changes how routine, entry-level work is performed:

1. Which tasks will AI perform, assist, or leave to employees?
2. What did less-experienced workers learn by performing those tasks?
3. Where will they now practice the judgment those tasks helped develop?
4. How will supervisors determine whether workers can perform independently?
5. What evidence will show whether the redesigned pathway is producing people ready for more consequential work?

The goal is not to prevent automation. It is to identify what developmental value will disappear and decide how that value will be replaced. A redesign is incomplete when it specifies what AI will do but not how employees will build the judgment, context, and independence the organization will need later.

Before eliminating a junior task, employers should ask what workers learned by doing it. If that learning still matters, it needs a new home.

Registered Apprenticeship Redesigns the Pathway

Registered Apprenticeship offers a structured way to preserve development as entry-level work changes. It combines paid employment with mentored on-the-job learning, related instruction, progressive wages, and a nationally recognized credential validated by the U.S. Department of Labor or a State Apprenticeship Agency.[10]

Its central advantage is that development becomes an intentional part of the job rather than an informal byproduct of junior tasks. Employers define the competencies workers must demonstrate. Apprentices practice those competencies in real assignments, receive guidance, and take on greater responsibility as their proficiency grows. In practical terms, Registered Apprenticeship provides each of the four developmental functions that an effective entry pathway must preserve:

• Practice is provided through structured on-the-job learning.
• Feedback is provided through mentoring and supervision.
• Context is developed through real assignments within the employer’s operating environment.
• Evidence of progression is created through competency assessment, increasing responsibility, wage progression, and a recognized credential.

The model is well suited to AI-exposed work. Apprentices can use AI while learning to inspect outputs, recognize errors, apply organizational context, and remain accountable for the result. Experience is demonstrated through increasingly independent and consequential work, not measured by time alone.

Apprenticeship.gov reports that 93 percent of people who complete a Registered Apprenticeship retain employment, with an average annual salary of $86,000 after completion.[10]

The U.S. Department of Labor’s evaluation of the American Apprenticeship Initiative surveyed 68 employers in 2020 representing 2,854 apprentices.[11] An extensive follow up review in 2026 conducted by the U.S. Department of Labor Chief Evaluation Office reaffirmed this magnitude, concluding that the most comprehensive and rigorous estimates place U.S. employer ROI between 40 to 90 percent.[12]

In the federal evaluation, most participating employers realized positive financial returns even under a conservative valuation of apprentice productivity. Employers also frequently reported a stronger skilled-worker pipeline, greater employee loyalty, improved culture, and benefits related to retention, productivity, and future manager development.[11]

These findings reinforce the central point: when employers embed structured learning inside real work, they can strengthen both current performance and the future talent pipeline.

The Pipeline Employers Cannot Afford to Lose

The vanishing entry level is not only a problem for recent graduates. It is a production problem for employers.

Every experienced employee begins as an inexperienced one. Organizations need specialists, managers, and leaders, but that talent does not emerge fully formed. It develops through practice, feedback, context, and increasing responsibility. If AI absorbs the work through which beginners once learned, employers must deliberately create another path to experience.

Preserving that pipeline requires more than identifying which tasks AI can perform. Employers must redesign roles, define the competencies workers need, structure on-the-job learning, and create credible measures of progression. JobForward helps employers put those elements in place, drawing on nearly 26 years of experience in occupational design and in developing, registering, and launching Registered Apprenticeship and work-based learning programs.

The organizations that thrive in the age of artificial intelligence will not simply automate routine tasks. They will redesign the way work is done such that productivity and human capability advance together. The future workforce will not emerge automatically. Employers must build it deliberately.

Endnotes

[1] 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 Working Paper, August 2026 revision. Analysis uses high-frequency ADP administrative payroll data covering millions of U.S. workers through June 2026.

[2] Seyed Mahdi Hosseini Maasoum and Guy Lichtinger, Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Résumé and Job Posting Data, SSRN 5425555, August 31, 2025. Generative AI adoption is identified through text analysis that detects dedicated “AI integrator” job postings.

[3] Leland D. Crane and Paul E. Soto, AI and Coder Employment: Compiling the Evidence, Board of Governors of the Federal Reserve System, Finance and Economics Discussion Series 2026-018, March 23, 2026. DOI: 10.17016/FEDS.2026.018.

[4] Felix Aidala and Sneha Puri, The Labor Market Is Tilting Toward Seniority, Indeed Hiring Lab, July 23, 2026. Indeed classifies seniority based on responsibilities, required skills, training, education, and experience rather than titles or stated years alone.

[5] U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Customer Service Representatives, last modified August 27, 2026. Reports typical entry requirements of a high school diploma, no related work experience, and short-term on-the-job training; median annual pay of $44,770 and median hourly wage of $21.53 in May 2025; projected 141,800 fewer jobs from 2025 to 2035.

[6] Maxim Massenkoff and Peter McCrory, Labor Market Impacts of AI: A New Measure and Early Evidence, Anthropic, March 5, 2026. Observed exposure combines theoretical LLM capability with real-world Claude usage data, weighting automated and work-related uses more heavily; the measure is mapped to O*NET tasks.

[7] U.S. Bureau of Labor Statistics, Artificial Intelligence (AI) Exposure Categories, Employment Projections program, last modified August 27, 2026.

[8] Erik Brynjolfsson, Danielle Li, and Lindsey Raymond, Generative AI at Work, The Quarterly Journal of Economics, Volume 140, Issue 2, May 2025, pages 889–942. DOI: 10.1093/qje/qjae044. Study covered 5,172 customer-support agents using a conversational AI tool built on GPT-family models; the majority of agents worked from offices in the Philippines. The paper reports a 15 percent average productivity gain in issues resolved per hour, a 36 percent gain for the lowest skill quintile, and small but statistically significant declines in resolution rates and customer satisfaction for the most skilled workers.

[9] PwC, AI Reshapes Global Labour Market into Two Distinct Paths, Rewarding Human Skills: PwC 2026 Global AI Jobs Barometer, June 15, 2026. The Barometer analyzed more than one billion job advertisements across 27 countries and territories; the entry-level findings cited here are based on 2.4 million U.S. entry-level job postings. PwC defines a skill as “traditionally senior” if it appeared more than 50 times in experienced, non-entry-level, high-AI-exposure job postings in 2019 but fewer than five times in entry-level, high-AI-exposure postings in 2019.

[10] U.S. Department of Labor, Registered Apprenticeship Program and Apprenticeship.gov homepage. Figures are drawn from the national Registered Apprenticeship program.

[11] Daniel Kuehn, Siobhan Mills De La Rosa, Robert Lerman, and Kevin Hollenbeck, Do Employers Earn Positive Returns to Investments in Apprenticeship? Evidence from Registered Programs under the American Apprenticeship Initiative, prepared by Abt Associates for the U.S. Department of Labor, Chief Evaluation Office, October 2022. Summary figures also reported in Investing in Talent Development: Benefits to Employers of Registered Apprenticeships from the American Apprenticeship Initiative.

[12] Kevin Hollenbeck, The Return on Investment (ROI) to Employers from Registered Apprenticeships, U.S. Department of Labor, Chief Evaluation Office, released April 30, 2026. A literature review of U.S. and selected international employer-ROI studies.