A Prediction About Young Workers, Artificial Intelligence, and Employment Through 2031
The Missing First Rung
Artificial intelligence is weakening the productive function of entry-level work before employers have found a replacement for its training function.
The labor market for young college graduates is sending us an important signal.
The overall American unemployment rate remains relatively low. Yet unemployment among recent college graduates reached approximately 5.7% in early 2026, while 41.5% of employed recent graduates were underemployed—working in occupations that typically do not require a college degree.1
RBC describes the same contradiction from another direction. The economy continues to grow, but white-collar employment is declining, hiring is concentrated in relatively few sectors, and businesses are investing heavily in artificial intelligence and capital-intensive infrastructure that does not produce large numbers of traditional entry-level jobs. RBC explicitly identifies a growing skills-matching problem for new graduates.2
The first conclusion is obvious: young people are having trouble finding appropriate work.
The more important conclusion is that the traditional first rung of the professional ladder is beginning to disappear.
For decades, companies hired young graduates to perform the lower-level cognitive work of the organization. They gathered information, prepared first drafts, updated spreadsheets, summarized meetings, created presentations, performed routine analysis, and learned by assisting more experienced workers.
That work had two functions.
It produced something the company needed, and it trained the young employee.
Artificial intelligence is weakening the first function before employers have found a replacement for the second.
A competent professional equipped with AI can now perform much of the work previously assigned to several junior employees. Remote work has simultaneously made observation, mentoring, and informal apprenticeship more difficult. New York Fed researchers estimate that remote work may explain 64% of the post-pandemic increase in unemployment among young college graduates. They also find that this deterioration began before widespread generative-AI adoption, meaning AI did not create the original problem—but may now accelerate it.3
My prediction is that this does not end with an entire generation remaining unemployed.
Instead, between 2026 and 2031, the labor market will reorganize itself around a new kind of young worker: the graduate whose synthetic subconscious allows one person to maintain several productive relationships, operate at the level of a small organization, and learn without relying entirely on a traditional employer.
The transition will be disorderly.
The economy appears reasonably healthy while the entrance narrows.
The market separates people who describe potential from people who demonstrate capacity.
Entry-level jobs begin decomposing into maintained capabilities.
Employers discover the delayed cost of eliminating junior apprenticeship.
The personal stack becomes employable capital—and ownership becomes contested.
The workforce separates by productive identity: institution-dependent operators and Stack People.
2026: The Entry-Level Bottleneck
During the remainder of 2026, the contradiction becomes increasingly visible.
Companies will continue saying that they cannot find qualified workers. Recent graduates will continue saying that they cannot find jobs. Both will be telling the truth.
Employers do not lack applicants. They lack applicants who can create enough immediate value to justify the cost of hiring, training, supervising, and integrating an inexperienced person.
The conventional graduate presents a degree, a résumé, and a promise:
Hire me, train me, and eventually I will become productive.
The employer increasingly compares that proposition with another:
Give the existing employee an AI system and see whether the department can operate without adding headcount.
In many cases, the employer will choose the second option.
This will not initially produce dramatic mass unemployment. It will produce something quieter: fewer openings, longer searches, more temporary positions, more graduates accepting work below their educational level, and more people remaining in jobs they would ordinarily have left.
The economy will appear reasonably healthy while the entrance narrows.
Young workers will initially interpret the problem personally. They will assume their résumé is inadequate, their interview skills are weak, or they chose the wrong major.
Some of those explanations will be true. But the deeper problem will be structural. The number of people seeking the first professional opportunity will exceed the number of organizations willing to provide apprenticeship.
By the end of 2026, a small group of graduates will begin responding differently. Instead of asking only, “Who will hire me?” they will ask:
“What economic output can I maintain?”
That question begins the transition.
2027: The Graduate Market Divides
In 2027, the market for young workers will split into two groups.
The first group will remain applicants.
They will use AI to improve résumés, write cover letters, prepare for interviews, and submit more applications. Because everyone has access to similar tools, application volume will surge. Employers will receive hundreds or thousands of polished, nearly indistinguishable applications.
AI-generated résumés will make the résumé less valuable.
AI-generated cover letters will make the cover letter almost meaningless.
The second group will use AI not merely to apply for work, but to perform work.
They will arrive with functioning systems, completed projects, recurring publications, small clients, automated research processes, operating dashboards, industry briefings, customer-service workflows, or documented evidence that they can maintain an outcome.
The distinction will no longer be between the student who used AI and the student who did not. Nearly everyone will use AI.
The distinction will be between the person who uses AI to describe potential and the person who uses AI to demonstrate capacity.
Employers will begin asking candidates to show their work rather than describe themselves. Portfolios, simulations, paid trials, project histories, and live problem-solving sessions will become more important than generic interviews.
Applicant
Uses AI to describe potential through polished application materials.
Operator
Uses AI to demonstrate capacity through systems, projects, and maintained outcomes.
A new inequality will appear.
Students from affluent families will have more time to build portfolios, accept unpaid projects, and experiment. Students who need immediate income will take whatever work is available. Without deliberate intervention by schools, the AI economy could initially deepen rather than reduce class differences.
At the same time, hidden overemployment will attract public attention. Some workers will quietly hold two or three incompatible full-time remote jobs, using AI to create the appearance of full participation in each.
Employers will respond with stricter moonlighting policies, activity monitoring, mandatory office attendance, and contractual restrictions.
This backlash will temporarily confuse two very different practices:
- collecting several salaries while making incompatible promises; and
- transparently providing defined outputs to several organizations.
The first is deception.
The second is the beginning of a new labor market.
2028: The Job Begins to Break Apart
By 2028, companies will begin decomposing some entry-level jobs into smaller units of economic value.
A small company may discover that it does not need a full-time communications coordinator. It needs a weekly newsletter, three social posts, customer-response support, and a monthly performance report.
It may not need a full-time research analyst. It needs a Monday briefing, competitor monitoring, and immediate notification when an important condition changes.
It may not need a junior project manager. It needs meetings documented, assignments tracked, delays identified, and exceptions escalated.
These are not necessarily jobs.
They are maintained capabilities.
The young worker equipped with a synthetic subconscious will be able to provide several of them simultaneously. One person may maintain communications for one organization, research for another, customer follow-up for a third, and reporting for a fourth.
This worker will not simply be freelancing in the traditional sense. Traditional freelancers repeatedly sell isolated pieces of labor.
The new worker will operate persistent systems.
The AI will retain procedures, prepare routine outputs, watch for changes, and bring unusual conditions to human attention. The person will supply judgment, accountability, relationship management, inspection, and decision-making.
This is when underemployment begins converting into overemployment.
A graduate may still hold an ordinary job that uses only part of their capability. But instead of waiting for that employer to discover the unused portion, the graduate will direct it toward another legitimate economic relationship.
The most capable young workers will begin looking less like employees and more like very small firms.
2029: Employers Discover the Apprenticeship Crisis
By 2029, companies will encounter the delayed consequence of reducing junior hiring.
They will begin running short of experienced middle-level workers.
A company cannot eliminate the first five years of a career and still expect an abundant supply of workers with five years of experience.
The organizations that stopped hiring junior employees in 2025, 2026, and 2027 will discover that AI preserved current output but did not automatically create the next generation of managers, professionals, relationship holders, and institutional leaders.
This will produce an apprenticeship crisis.
Employers will need to rebuild the first rung, but they will not recreate the old entry-level job. The old model was too expensive and too dependent on assigning young people routine work that AI can now perform.
Instead, companies will create AI-native apprenticeships.
A young person may supervise an AI system from the beginning rather than spending two years performing the system’s routine tasks manually. The apprenticeship will emphasize exception handling, judgment, customer interaction, ethical decisions, systems thinking, and responsibility.
AI will also become part of the training infrastructure.
The synthetic subconscious will explain unfamiliar documents, simulate customer situations, preserve institutional knowledge, evaluate work against company standards, and provide constant low-cost feedback.
Human mentors will remain important, but they will intervene at higher-value moments.
The apprentice will no longer learn primarily by repeating simple tasks thousands of times. The apprentice will learn by examining why the system fails, where predictions diverge from Actual, and when human judgment changes the outcome.
This transition will be difficult because routine work historically trained intuition. Removing the routine task may also remove the experiences through which expertise developed.
The best apprenticeship systems will therefore use AI not only to complete work, but to expose the reasoning, exceptions, tradeoffs, and consequences embedded in the work.
2030: The Personal Stack Becomes Employable Capital
By 2030, the labor market will begin recognizing that a capable young person brings more than skills to an organization.
The person brings a personal stack.
That stack will contain methods, agents, research procedures, quality controls, communication patterns, industry knowledge, project memory, trusted collaborators, and accumulated ways of solving problems.
Two graduates with the same degree will therefore arrive with radically different productive capacities.
One will arrive with access to general-purpose applications.
The other will arrive with a developed operating environment that has already produced measurable results.
Hiring will begin to resemble the acquisition of a small productive system.
Employers will ask:
- What can your stack currently maintain?
- How has it performed in real situations?
- What information does it retain?
- How do you verify its outputs?
- How does it protect confidential data?
- Can it integrate with our systems?
- Which parts belong to you, and which become company property?
- What happens when you leave?
This will create conflict over ownership.
Companies will want control over the systems used to perform company work. Stack People will resist surrendering the accumulated capability that makes them valuable.
Employment agreements will evolve to distinguish company data and proprietary processes from the worker’s general methods, personal agents, acquired expertise, and reusable infrastructure.
The same issue will appear in universities.
Students will be less willing to leave their synthetic subconscious outside the classroom. A student’s AI will know the student’s history, weaknesses, vocabulary, research interests, conceptual models, and preferred ways of learning.
Forcing that student to begin every course with an empty screen will look increasingly irrational.
Education will begin shifting from testing what a student can produce alone toward testing what the student-plus-stack can responsibly accomplish.
2031: The Young Workforce Separates Into Two Classes
By 2031, the labor market will not be divided simply between people who use AI and people who do not.
AI will be nearly universal.
The important division will be between people who rent isolated AI capabilities and people who own an accumulating operating environment.
The first group will remain institution-dependent.
They will enter an employer’s applications, follow its procedures, use its approved AI interface, and perform the tasks assigned to them. When they leave the organization, much of their accumulated digital capability will remain behind.
They will still have valuable careers. Large organizations will continue needing people who can operate effectively within governed systems.
But their productive identity will remain attached to the institution.
The second group will be Stack People.
They will bring portable capability into each economic relationship. They will connect to company systems through approved interfaces while retaining their own methods, memory, judgment, and operating infrastructure.
They may have one primary employer, several clients, a teaching role, a small digital product, and participation in temporary project teams—all supported by the same synthetic subconscious.
Their economic lives will not fit neatly into the word “job.”
They will maintain portfolios of responsibility.
Some will earn more than their traditionally employed peers. Others will value resilience more than maximum income. Losing one client will not mean losing their entire livelihood. Knowledge gained in one setting will improve performance in another, subject to strict confidentiality boundaries.
Institution-dependent
The productive identity remains attached to the organization’s governed system.
Stack Person
Portable capability enters each economic relationship while responsibility stays human.
The traditional graduate asks an institution for a place inside its system.
The Stack Person arrives with a system and asks where it can create value.
What Happens to the College Degree?
The degree will not disappear by 2031.
It will lose its ability to function alone.
A degree will continue signaling discipline, exposure to a field, and successful completion of a recognized program. Regulated professions will continue requiring formal credentials.
But for many forms of knowledge work, the degree will become the beginning of the evidence rather than the conclusion.
The graduate will increasingly need three things:
A credential, a demonstrated capability, and a trusted history of completed work.
Universities that understand this will stop treating AI merely as a cheating problem. They will help students build governed synthetic subconscious systems, complete real projects, develop judgment, protect confidential information, and leave school with productive infrastructure.
Universities that do not adapt will continue graduating applicants into a market that increasingly wants operators.
The Larger Prediction
My prediction is not that AI eliminates employment for young people.
It is that AI destroys the old bargain through which young people entered professional life before a new bargain is ready.
The old bargain was:
AI reduces the amount of lower-level work available. Remote work weakens informal apprenticeship. Employers become reluctant to pay an inexperienced person to learn.
The new bargain will be:
That is a much harder entrance requirement.
It is also a much larger opportunity.
The successful young worker of 2031 will not necessarily work more hours than the successful young worker of 2026. The worker will maintain more productive relationships because predictable cognitive work has been transferred to a synthetic subconscious.
This person will be overemployed not because every hour has been sold twice, but because one developed human intelligence can now support more than one economic role.
The danger is that millions of young people will wait for the old labor market to return.
It will not return in its previous form.
The opportunity is that young people can become economically larger than the positions initially offered to them.
Between now and 2031, the first rung of the career ladder will not simply vanish.
It will move.
It will no longer be found only inside an employer.
Young people will have to build part of it themselves.