World Frictions | 11/3/2026 | Tsuyoshi Hadano

If Opportunity Is Open to Everyone, Why Do the Same People Keep Getting It?

AI expands access to knowledge, yet learning, proving skills and hiring remain unequal. Evidence from the US, Europe, Japan and emerging economies.

Knowledge has never been easier to reach. We can search for answers, translate unfamiliar material, learn a new skill online, and use AI to attempt work that once seemed to require a specialist. It is tempting to say that opportunity has finally become available to everyone.

Yet the view from the hiring desk is different. A promising new occupation may still require a degree and previous experience. Some workers have the time and resources to retrain; others do not. Attractive jobs cluster in particular companies, cities and countries.

Making the tools of achievement more accessible is not the same as making the opportunities to achieve equally available.

The question is not whether AI is good or bad for equality. It is whether the institutions that decide who gets hired, trained and trusted have changed as quickly as the tools themselves.

A wider entrance to knowledge, but a narrower gate to employment?

Consider the contrast in LinkedIn’s August 2026 research. In the United States, the typical advertised compensation in an AI occupation was approximately $177,000, against $80,000 for a non-AI occupation. Yet women made up only 26% of US hires into AI jobs in 2025, compared with 50% in non-AI hiring. Among people already working in AI occupations, 91% held at least a bachelor’s degree.

Source: LinkedIn, August 18, 2026

These figures have limits. Twenty-six percent is a share of hires, not a female applicant’s chance of success. Ninety-one percent describes the education of workers in AI occupations, not the probability that an applicant without a degree will be rejected. Advertised compensation is not the realized earnings of all workers. Nor does the dataset, by itself, establish that an individual employer discriminated against any applicant.

Still, the contrast raises a powerful question: when a valuable new career path appears, who can actually enter it?

AI may help a self-taught person build a portfolio, learn programming or analyze business data. But if the employer filters candidates on credentials before reviewing that work, the applicant may never reach the stage where ability can be demonstrated. A degree holder may also struggle to find the experience demanded for an entry-level role. Cheaper tools cannot force organizations to change how they assess risk.

Removing a degree requirement is not enough

In 2026, the OECD examined the growing interest in skills-first hiring. Rather than treating degrees and previous job titles as the main proxies for ability, employers can define the actual competencies a role requires and give applicants ways to demonstrate them.

But the OECD adds an important warning: simply deleting degree requirements does not automatically produce opportunities for underrepresented groups. Structured assessments, competency-based interviews, apprenticeships and alternative entry pathways must make the change real.

Source: OECD, A Skills-First Labour Market, 2026

An employer may understandably want someone who can contribute immediately. Yet if every employer demands three years of experience, where does the first year come from?

This is a much older contradiction than AI. The technology makes it more visible. It is becoming easier to acquire some skills independently, but a self-directed learning project is not necessarily accepted as work experience. The capacity to learn and the chance to have that learning recognized are different things.

The people asked to adapt may not be those who capture the new opportunities

The International Labour Organization’s March 2026 analysis, drawing on occupational data from 84 countries, shows that the impacts of generative AI are not gender-neutral. Twenty-nine percent of female-dominated occupations were in categories exposed to generative AI, compared with 16% of male-dominated occupations. Differences in occupational composition, including concentration in administrative and clerical work, help explain the gap. Meanwhile, women remain underrepresented in AI-related occupations.

Source: ILO, March 5, 2026

Exposure must not be confused with actual job loss. In many roles, AI is more likely to change tasks, working conditions and required skills than to eliminate the occupation entirely.

The broader contradiction is that the people whose work is most likely to change do not necessarily have equal access to the training, networks and growing roles that may follow. Someone working long hours, caring for a family member, or combining multiple jobs cannot necessarily attend a course at the same time as someone with more flexibility. What looks like a difference in willingness to learn may partly reflect a difference in the ability to make time for learning.

Even businesses in the same market do not start from the same position

Eurostat reports that in 2025, 20% of EU enterprises with at least ten employees used AI technologies. The share was approximately 55% among large companies and 19% among small and medium-sized businesses. Across member states, adoption ranged from 42% in Denmark to about 5% in Romania.

Source: Eurostat, Digitalisation in Europe 2026

These are rates of enterprise adoption, not measures of individual workers’ talent. But workplace conditions matter. A larger business may have a technical team, a training budget and time to redesign workflows. A small clinic, retailer or service company may be concentrating on today’s customers and cash flow. A free tool does not erase those differences.

This also affects the future labor market. A worker who can experiment in a supported environment may develop experience that others are never given the chance to obtain. That is a plausible mechanism, not a causal finding established by the Eurostat numbers.

Japan: a skills shortage, or a shortage of paths to develop skills?

Japan presents another tension. Businesses say they need more people able to drive digital transformation, yet hiring and development practices may not have changed enough to create them.

Japan’s Information-technology Promotion Agency (IPA) reported that 85.1% of surveyed Japanese enterprises pursuing digital transformation described a shortage of the people needed to advance it. Comparable figures were 23.8% for the United States and 44.6% for Germany. These results describe employers’ perceived talent supply in specific survey populations, not a universal measure of each country’s workforce.

Source: IPA, Digital Talent Development in the AI Era, October 9, 2025

The IPA analysis argues for skills-based evaluation and assignments, alongside practical learning opportunities, rather than reliance on additional courses and certificates alone. Its 2026 survey release also notes that, despite AI adoption spreading particularly among larger companies, reported benefits are concentrated in efficiency and speed rather than broader value creation.

Source: IPA, July 16, 2026

This invites uncomfortable questions for managers. Do newly trained employees get an actual improvement project to lead? Can someone make a contained mistake without being permanently judged as incapable? Do job titles and departmental histories prevent existing talent from being recognized?

A company can sincerely experience a shortage of people ready for a job while also maintaining practices that limit how people become ready.

Across borders, even “free” does not mean the same thing

All of this assumes a person can access a device, a reliable connection, and learning resources in a useful language. Worldwide, those assumptions do not always hold.

The World Bank’s Digital Progress and Trends Report 2025 estimates that, as of April 2025, roughly 24% of internet users in high-income countries used ChatGPT, compared with approximately 5–6% in middle-income economies and less than 1% in low-income economies. Adoption in emerging markets is growing, but connectivity, electricity, skills and locally relevant data remain uneven.

Source: World Bank, Digital Progress and Trends Report 2025

A tool that costs nothing to download is not necessarily free to use in practice. Someone still needs a device, a connection, time to learn, and the confidence to apply it. At the same time, the Bank identifies opportunities for affordable, locally adapted AI that can broaden access to education, services and economic activity.

The evidence therefore does not justify a simple prediction that AI will inevitably widen inequality. It shows something more complicated: the same technology can open doors while the conditions necessary to walk through those doors remain unequal.

“Keep learning.” Who makes learning possible?

The World Economic Forum’s Future of Jobs Report 2025 aggregates employers’ expectations that, out of every 100 workers, 59 will need training by 2030. Employers anticipate that eleven of those workers are unlikely to receive the retraining or upskilling they need.

Source: World Economic Forum, Future of Jobs Report 2025

These are forecasts drawn from employer expectations, not observed outcomes or a confirmed count of workers already denied training. Nevertheless, they expose a fundamental question: if continual learning becomes a condition of employment, who pays for the time and opportunity to learn?

It is reasonable to ask workers to take responsibility for their development. It is not reasonable to pretend that everyone has the same hours available, the same access to mentoring or the same freedom to fail while learning.

Three different doors called “opportunity”

Based on these findings, I propose separating opportunity into three distinct stages.

First, the opportunity to learn. A person must be able to reach useful tools, receive feedback, and practice without taking an unreasonable personal risk.

Second, the opportunity to demonstrate ability. Employers need ways to assess actual skills through work samples, structured tasks and evidence of collaboration, instead of relying solely on previous titles or degrees.

Third, the opportunity to be trusted with real work. Someone’s skills cannot turn into meaningful experience unless an organization gives that person a responsibility worth learning from.

AI may help widen the first door. The second and third doors still depend heavily on the choices made by employers, managers and institutions.

This is not an argument against university education or experience. Both can carry genuine value. It is an argument for offering additional credible ways for people to show what they can do.

What organizations can change now

For recruitment, define which capabilities a role actually requires before adding blanket credential or years-of-experience requirements. Combine work samples with structured interviews that ask candidates to explain their choices. Avoid using unpaid assessments as a source of free labor, and make evaluation standards consistent.

For development, assess not only course completions but whether employees get a chance to use new skills on actual projects. Provide some learning time within paid work where practical. Create bounded responsibilities for less-experienced employees, with feedback and support.

For career mobility, ask not only what a person has done, but also what the person can do now and what responsibility could help them grow next. Internal project opportunities and cross-department assignments are possible ways to make skills visible.

And for executives, perhaps the hardest question is this: when AI saves an organization time or money, does any of that gain return to its people in the form of greater opportunity to develop?

The difference between possibility and permission

AI can make it easier to attempt work that was previously inaccessible. That is a real advance. But access to knowledge is not the same as access to a job. Building an impressive project is not the same as being invited to show it. Free lessons are not the same as time to learn.

When we claim that opportunity is open to everyone, what exactly do we mean? Access to a tool? The right to compete? Or a fair chance to be seen, assessed and entrusted with meaningful work?

Technology can open the door to possibility. People and organizations still decide who gets a place on the stage beyond it.

That may be the central friction of the AI age: not simply who can use the tools, but who is allowed the opportunity to try.