The Real Danger Is Not That Artificial Intelligence Takes Every Job, but That It Makes Opportunity Even More Unequal
Two Young People Wake Up to the Same Technology
Imagine two 22-year-olds beginning their day.
The first opens a laptop in a comfortable apartment near a major American city. She has a university degree, reliable high-speed internet and access to the latest artificial-intelligence tools. Before breakfast, AI helps her research a prospective employer, analyze a spreadsheet, improve a presentation and prepare for an interview. She does not regard artificial intelligence as a competitor. She regards it as leverage.
The second young person also understands that AI is transforming work. But his experience is very different. He is looking for his first serious job, and some of the entry-level work through which previous generations learned a profession is becoming easier to automate. Employers want experience he has not yet had an opportunity to acquire, and another online course does not solve the problem of getting somebody to open the first door.
Both young people belong to Generation Z. Both are living through the same technological revolution. But they may be heading toward two very different futures. That is what worries me about artificial intelligence.
The greatest danger may not be that AI takes everybody’s job. The evidence does not currently support such a sweeping conclusion. The more immediate danger is that AI becomes an extraordinary accelerator for people who already possess education, connections, technology and opportunity while becoming another barrier for those trying to obtain them. If that happens, we will not simply have an AI revolution. We could have an AI class divide.
AI Is Not Coming for Everyone Equally
The public conversation about artificial intelligence frequently swings between extremes. On one side, AI will supposedly create extraordinary prosperity and liberate human beings from repetitive work. On the other, machines are coming for everybody’s jobs. Reality is more complicated.
The International Labour Organization estimates that one in four workers worldwide is employed in an occupation with some exposure to generative AI. But its research also concludes that transformation, rather than complete replacement, is the more likely outcome for most of those jobs because human input remains necessary.
That distinction matters enormously. Exposure is not unemployment, and the ability of AI to perform some tasks within an occupation does not mean the entire occupation disappears. But averages can also hide who is absorbing the disruption.
Researchers at Stanford’s Digital Economy Lab examined payroll data covering millions of American workers through June 2026. They found no evidence of widespread economy-wide job displacement associated with AI. Yet among workers ages 22 to 25 in highly AI-exposed occupations, employment was about 19 percent below where it would have been had it kept pace with employment among similarly aged workers in less-exposed occupations. The researchers found that the difference appears primarily through reduced hiring rather than mass firing. That should get our attention.
The First Rung Is Where Careers Begin
Young workers have always entered the labor market at a disadvantage. They possess less experience because they have been alive for fewer years and working for even fewer. We then created the peculiar institution known as the entry-level job requiring experience. Artificial intelligence could make that contradiction worse.
Think about how professional careers traditionally developed. A young accountant began with relatively straightforward assignments. A junior programmer handled simpler coding. A young journalist covered routine stories. A beginning analyst conducted basic research. A new employee prepared the first draft rather than making the final decision. Those tasks did more than produce work. They produced workers.
The young employee observed experienced colleagues, made manageable mistakes, received corrections and gradually developed judgment. Routine work was part of the apprenticeship, even when nobody called it an apprenticeship.
Now consider the tasks generative AI performs increasingly well: summarizing documents, organizing information, producing initial drafts, writing basic computer code, answering routine inquiries and conducting preliminary analysis. From an employer’s perspective, automation can be perfectly rational. Why pay someone to spend six hours completing a task that technology can help complete in twenty minutes?
But society has another question to answer. If we automate some of the work through which beginners acquire experience, how do beginners become experienced? We cannot remove the first rung of the ladder and then criticize young people because they cannot climb.
The Young People Who Learn to Command AI
There is another side to this story, and it is extraordinarily promising. For a capable young person, AI can function almost like an intellectual power tool. Someone starting a small business can conduct market research that once required consultants. A programmer can work faster. A writer can organize research. A student can receive individualized explanations. An entrepreneur can translate materials, analyze information and test ideas without hiring a large staff.
This could democratize capabilities that were once available primarily to large organizations and wealthy individuals. The World Economic Forum’s employer research shows that AI and big data are among the fastest-growing skill areas, but something else in the findings is equally important. Employers continue to place enormous value on analytical thinking, resilience, creative thinking, technological literacy, leadership and lifelong learning.
That suggests that the future may not belong simply to people who know how to use AI. It may belong disproportionately to people who bring judgment, education, experience and creativity to AI. That distinction is crucial because technology does not arrive in an economically neutral society.
The Young People AI Could Leave Behind
Now imagine the same technological revolution from the other side of the economic divide. What does “learn AI” mean to a young person without a reliable computer? What does an AI revolution mean in a school struggling to provide basic technology? What does it mean for a rural student with unreliable connectivity, or for someone whose education never developed the analytical skills necessary to distinguish a convincing AI answer from a correct one?
UNICEF’s recent research with young people in five African countries makes this problem concrete. Children themselves identified affordability, internet access and device availability as barriers that could prevent them from benefiting equally from AI, particularly in rural and low-income communities. The problem is not confined to Africa.
A teenager in a poor American neighborhood and a teenager at an elite suburban school may technically have access to the same publicly available AI system. That does not mean they have equal access to what AI can make possible. One may have advanced coursework, mentors, a modern laptop, internships and parents with professional networks. The other may possess a smartphone and enormous talent but little of the institutional infrastructure that converts talent into opportunity. Giving both people access to a chatbot does not erase those differences.
America Could Develop an AI Class Divide
This is where the American conversation needs to become more sophisticated. We often talk about technological access as though possession of the technology produces equality. The history of the internet should have taught us otherwise.
Access matters, but so does the quality of access. Skills matter. Education matters. Professional networks matter. Social capital matters. The ability to take an unpaid internship matters. Knowing someone who can explain how an industry works matters. AI could magnify these differences.
Imagine two recent graduates doing the same job. One understands the field deeply and uses AI to eliminate routine work, investigate alternatives and increase productivity. The other has been trained primarily to perform the routine tasks AI now handles. The first worker becomes more valuable. The second becomes more vulnerable. Now extend that difference across millions of people.
We could gradually create a labor market divided between people whose knowledge allows AI to amplify their productivity and people whose tasks allow AI to substitute for some of their labor.
Stanford’s findings do not prove that this future is inevitable. The researchers explicitly find no broad employment collapse. But their evidence of a widening employment gap among young workers in AI-exposed occupations is exactly the kind of early warning policymakers, universities and employers should take seriously. Waiting until the consequences become unmistakable would be a peculiar definition of preparedness.
The Global Divide Could Become Even Larger
While researching my book, Stolen Youth: How Kenya’s Young Are Reclaiming Their Future, I became increasingly interested in a contradiction affecting Generation Z across national boundaries. Young people are better connected to the world than any previous generation, yet their actual opportunities remain profoundly shaped by where they were born and the institutions surrounding them. AI could intensify that contradiction.
A talented young person in Nairobi can potentially access the same extraordinary technological capabilities as someone in Philadelphia. That is revolutionary. Geography no longer determines access to knowledge as completely as it once did. But geography has not disappeared.
Electricity matters. Broadband matters. Education matters. Language matters. Capital matters. Employers matter. Political stability matters. The ability of an economy to create productive work matters. The ILO estimates that overall GenAI exposure is considerably greater in high-income countries than in low-income countries. That might initially appear to mean poorer countries are safer from disruption, but it also reflects differences in occupational structure and technological adoption.
Being less exposed to technological disruption is not necessarily an advantage if it also means being less able to capture technological productivity. That is the global danger. Wealthier societies and wealthier people could receive more of AI’s benefits while vulnerable workers and poorer societies encounter disruption without receiving an equivalent dividend.
Education Cannot Simply Tell Students to Learn AI
Whenever technology changes employment, our favorite solution is reskilling.
There is some truth in that response. The World Economic Forum estimates that 39 percent of workers’ existing skill sets could be transformed or become outdated between 2025 and 2030. Its employer survey suggests that, out of every 100 workers, 59 would need training during that period. But “reskill” can become another way of transferring structural responsibility onto individuals.
What should people learn? Where should they learn it? Who pays? How does a parent working full time retrain? How does a young person know which skill will still be valuable five years from now? Universities and schools should absolutely teach AI literacy, but teaching students how to write prompts is not enough.
Young people need to understand when AI is wrong. They need to verify information, recognize manipulation, protect privacy and understand the ethical consequences of automated decisions. They need domain knowledge deep enough to evaluate what the machine produces. Ironically, the more powerful artificial intelligence becomes, the more valuable certain human capabilities may become.
Critical thinking matters when plausible misinformation can be generated instantly. Judgment matters when machines can produce multiple answers but cannot accept moral responsibility for choosing among them. Communication, empathy, creativity and leadership matter when technical capabilities become widely available. The educational response to AI therefore cannot be less human education. It may require better human education.
A Two-Tier Society Is Not Inevitable
There is nothing predetermined about the outcome. Governments can expand meaningful broadband access and ensure that young people have devices capable of participating in the digital economy. Schools can teach AI literacy alongside critical thinking rather than treating the technology as either a cheating machine or a miracle.
Universities can redesign the transition between education and employment. Employers can preserve genuine entry-level opportunities even as AI changes junior work. If automation eliminates routine training tasks, companies can create structured apprenticeships, mentorship programs and supervised AI-assisted work to replace the experience those tasks once provided. We also need to take technical and vocational education much more seriously.
The future economy will not consist entirely of people sitting behind computers instructing AI systems. It will require electricians, healthcare professionals, renewable-energy technicians, advanced manufacturing specialists, construction workers, mechanics and people maintaining the enormous physical infrastructure upon which the digital economy depends. Many of those occupations will use increasingly sophisticated technology without being easily reduced to a chatbot.
Governments should therefore connect AI policy to education policy, industrial policy and labor policy. It makes little sense to spend billions encouraging technological development while treating workforce adaptation as something individuals should somehow figure out themselves.
Employers have responsibilities as well. If businesses capture productivity gains from AI, they should participate in developing the workers needed for the new economy rather than expecting universities and taxpayers to carry the entire cost.
The Choice We Are Actually Making
I am not afraid of artificial intelligence because it is powerful. Human beings have repeatedly developed powerful technologies. Some destroyed occupations, some created industries, and most eventually became so ordinary that later generations could hardly imagine life without them.
What concerns me is introducing a powerful technology into societies already marked by enormous inequalities of wealth, education and opportunity and then pretending everyone is starting from the same line. They are not.
For one young person, AI could become the most extraordinary productivity tool of his generation. For another, it could eliminate the first opportunity he was hoping somebody would finally give him. For one student, it could provide personalized education previously available only through expensive tutoring. For another, inadequate connectivity could leave that promise largely theoretical.
For one entrepreneur, AI could reduce the cost of starting a business. For another worker, it could reduce the number of people the business needs to hire. All of these things can be true simultaneously. That is why the debate about whether AI is “good” or “bad” for Generation Z misses the larger question. Artificial intelligence could be enormously beneficial for some young people while deeply disruptive for others.
Our responsibility is to make sure the dividing line is not simply the wealth of your parents, the quality of the school you attended, the neighborhood where you grew up, the country where you were born or whether somebody happened to give you your first professional opportunity. The future of artificial intelligence is still being written. So is the future of Generation Z. Whether those two futures produce greater opportunity or a more rigid two-tier society will depend not only on what the technology can do, but on what governments, educators, employers and the rest of us decide to do with it.
Patrick Machayo is the author of Stolen Youth: How Kenya’s Young Are Reclaiming Their Future.