#THINKFUTURE
Prelude to the series: Building the Future on a Cracked Foundation
AI may transform some of the work traditionally given to beginners. But what happens to the pathway through which beginners become experts?
Imagine a young graduate arriving for their first day of work.
They have completed their qualification, but they have never worked inside an organisation. They understand the principles of their profession, but not yet how those principles are applied when information is incomplete, deadlines are tight and different stakeholders want different outcomes.
Their first assignments are unlikely to be strategically important. They may conduct basic research, prepare a first draft, check information, update a spreadsheet, analyse a straightforward problem or assemble a presentation for someone more experienced.
The work may appear routine. Some of it may even appear inefficient. The graduate takes longer than an experienced colleague, asks questions, misses important details and produces work that requires correction.
Yet something consequential is happening.
The graduate is learning how the organisation works. They are developing professional standards, contextual understanding and the ability to recognise when an apparently correct answer does not make sense. They are observing how experienced people interpret uncertainty, challenge assumptions and exercise judgement. Through practice, feedback and gradually increasing responsibility, the beginner starts becoming capable.
Now imagine that the graduate never arrives.
The research is completed by an AI system. The first draft is generated in seconds. The spreadsheet is updated automatically. The routine analysis is accelerated by software, and the presentation is assembled using existing organisational information.
The work is completed faster and potentially at lower cost. The productivity case appears compelling.
But what happened to the person who was supposed to learn through doing it?
The experience paradox
Young people have long faced a familiar contradiction. They need experience to secure employment, but they need employment to acquire experience.
In South Africa, this is not a theoretical concern. In the first quarter of 2026, the official unemployment rate among economically active people aged 15 to 34 was 45.8%. The challenge extends beyond those officially classified as unemployed. Approximately 45.6% of the broader population aged 15 to 34 were not in employment, education or training.
These two measures have different denominators and should not be treated as interchangeable. The official unemployment rate applies to young people participating in the labour force, while the not in employment, education or training rate applies to the wider youth population. Together, however, they demonstrate the scale of young people’s exclusion from pathways into work and development. Statistics South Africa, Quarterly Labour Force Survey, Quarter 1 of 2026
The barrier is not only a shortage of qualifications. It is also a shortage of opportunities to acquire experience. Among the 4.8 million unemployed young people aged 15 to 34 in the first quarter of 2025, 58.7% reported having no previous work experience. Nearly six in ten unemployed young people were still waiting for their first opportunity to enter the labour market. Statistics South Africa, South Africa’s Youth in the Labour Market
Separate Statistics South Africa analysis of labour-market transitions during 2024 found that 9.8% of people with previous work experience moved into employment, compared with 2.6% of those without experience. This is a strong association, although it does not prove that experience alone caused the difference. Education, age, location, economic conditions and other factors may also influence employment prospects. Statistics South Africa, Who Is More Likely to Transition into Employment?
The figures nevertheless reveal a difficult cycle. Employers prefer candidates who have already demonstrated that they can perform in a workplace. Young people cannot demonstrate that ability until an employer gives them a first opportunity.
Artificial intelligence did not create South Africa’s youth unemployment crisis. The crisis long predates generative AI and reflects deeper structural problems involving economic growth, education, inequality, geography and access to opportunity.
AI could, however, add another layer to the experience paradox.
The work most accessible to AI may also be the work through which people begin
Many activities that generative AI can perform or accelerate are also activities traditionally allocated to graduates and junior employees. These include summarising information, preparing initial drafts, conducting routine research, producing standard documentation, performing basic analysis and completing administrative tasks.
This does not mean that every entry-level job will disappear. Jobs consist of multiple tasks, and exposing some tasks to AI does not automatically eliminate an entire occupation.
The International Labour Organisation estimates that one in four workers globally are employed in occupations with some degree of generative AI exposure. Only 3.3% of global employment falls within its highest exposure category. The ILO concludes that job transformation is more likely than the complete replacement of occupations, while clerical occupations remain the most exposed. Exposure therefore indicates the potential for tasks to change. It does not mean that every exposed job will disappear. International Labour Organization, Generative AI and Jobs: A Refined Global Index of Occupational Exposure
The distinction between a job and its component tasks is important. An organisation may continue employing junior people while using AI to change what they do. It may also reduce entry-level recruitment because fewer people are required to produce the same volume of work. The eventual outcome will depend on the occupation, industry, economy and choices made by individual organisations.
The OECD’s 2026 Employment Outlook identifies several ways in which generative AI could affect young people. Large language models may be able to perform simpler cognitive tasks that are common in entry-level roles, while more sophisticated expert work may be more difficult to automate. Even where entry-level work is not more intensively automated, a broader reduction in labour demand may affect young people disproportionately because they are trying to enter the labour market rather than retain established positions.
The OECD also cautions against assuming that generative AI is already the primary cause of deteriorating employment prospects for young people. Its assessment indicates that the role of recent large language model advances appears limited so far, while weaker economic conditions and longer-term changes in technology and skills requirements provide more established explanations. AI should therefore be treated as an emerging pressure on entry-level pathways, not as the proven cause of current youth unemployment. OECD Employment Outlook 2026
Early evidence from the United States adds to the concern. Researchers associated with the Stanford Digital Economy Lab, using payroll data, found declining employment among workers aged 22 to 25 in highly AI-exposed occupations. Employment among more experienced workers in the same occupations remained comparatively stable or continued to grow.
The declines were concentrated in occupations where observed AI use was more likely to automate work than augment the person performing it. This is important emerging evidence, but it does not establish that AI alone caused the employment changes. Nor should findings from the United States be generalised uncritically to South Africa or every profession. Stanford Digital Economy Lab, Canaries in the Coal Mine?
The future of entry-level employment is therefore not predetermined. There is not yet conclusive evidence that AI is eliminating entry-level jobs across economies. There is, however, enough evidence of changing tasks and emerging labour-market pressures to make the organisational question increasingly difficult to ignore.
A job is not the same as a developmental pathway
The usual debate asks whether AI will replace particular jobs. That question may be too narrow.
An entry-level job has traditionally served at least two purposes. It produces work for the organisation, and it creates an opportunity for a person to develop.
These purposes have often been combined. Junior employees contribute by performing relatively structured activities while learning from the work, their mistakes and the people around them. As their knowledge and judgement develop, they take responsibility for increasingly difficult problems.
AI may begin to separate these two purposes.
An organisation may no longer need a junior employee to produce a first draft, summarise a document or conduct a basic analysis. It may still need a developmental pathway through which a junior employee learns to evaluate the draft, question the summary and recognise when the analysis is misleading.
The output can be automated. The learning cannot simply be assumed.
This is where an apparently sensible productivity decision could create a longer-term organisational problem. If fewer young people enter a profession, fewer people may accumulate the experience required for its intermediate and senior roles. Organisations could continue relying on their existing experts without noticing that the pipeline behind them is narrowing.
The consequences may not appear in the next quarterly report. They could become visible years later, when experienced specialists retire, leadership positions become more difficult to fill, and organisations discover that too few people were developed to replace them.
This is a strategic risk rather than an established universal outcome. It is nevertheless a risk that leaders should examine before the consequences become visible.
The problem is therefore not only whether young people can find employment.
It is whether organisations can continue developing experienced people after removing some of the work through which experience was previously acquired.
When productivity and development become separate decisions
For many years, organisations did not always need to distinguish between completing junior work and developing junior people. Both occurred through the same assignments.
AI changes that assumption.
A task may produce several forms of value at the same time:
immediate productive value through the output it creates
developmental value through the knowledge and experience it builds
organisational value through the future capability pipeline it supports
A conventional automation decision may measure the first form of value while overlooking the other two.
If an AI system can complete a task faster, the immediate productive value may be preserved or improved. The developmental value may not be. Unless the organisation intentionally replaces the learning opportunity, it may remove more than a unit of work.
This does not mean that routine tasks should be protected from automation merely because beginners once performed them. Some junior work is repetitive without being meaningfully developmental. Some AI-enabled work may create better learning opportunities by giving beginners faster feedback, broader exposure and more time to focus on interpretation.
The more important distinction is between automating a task and redesigning a developmental pathway.
These are not the same decision.
Who is responsible for building the new first step?
It would be easy to treat this as a problem for young people to solve.
They must become more adaptable. They must learn to work with AI. They must develop digital skills, build portfolios, gain practical exposure and find new ways to demonstrate their value.
All of this may be necessary. None of it resolves the organisational question.
Educational institutions will need to reconsider how learners prepare for AI-enabled work. Governments will need to address the economic and social consequences of constrained employment pathways. Young people will need opportunities to work with AI rather than merely compete against it.
Organisations also have a role.
Entry-level employment has functioned as part of the system through which organisations develop future capability. If technology changes that system, organisations cannot simply assume that universities, individuals or the wider labour market will continue producing experienced people for them.
This does not require organisations to preserve obsolete positions or employ people to perform work that no longer creates value. It requires leaders to recognise that removing a task and replacing a developmental experience are different decisions.
Organisations may need to redesign early-career pathways deliberately. Supervised AI-enabled work, apprenticeships, simulations, rotations, observation, structured feedback and progressively more complex assignments could all become more important.
These approaches should not yet be treated as proven replacements for genuine workplace experience. Their effectiveness will depend on how they are designed, the quality of supervision and whether young people receive real exposure to uncertainty, responsibility and consequences.
The central leadership question is more fundamental.
When organisations evaluate the work AI can perform, are they measuring only the output being replaced, or are they also examining the capability that was being developed through producing it?
A warning from the first rung
Young people may be among the first to encounter this problem, but they will not be the only people affected.
When organisations reduce entry-level opportunities without redesigning how people develop, they may weaken the pathways through which future specialists, managers and leaders emerge. A decision that improves productivity today could reduce the capability available tomorrow.
This is where the question becomes larger than employment.
What if organisations are investing in the future while weakening the human foundation on which that future depends?
The new THINK FUTURE series, Building the Future on a Cracked Foundation, will examine a possibility that receives far less attention than AI productivity, cost reduction or technological adoption:
What if the greatest organisational risk of AI is not simply that it changes jobs, but that it changes how people become capable?
The series will explore what happens when organisations automate developmental work, narrow pathways into professions, disrupt the formation of expertise and continue planning for future capability as though the old development system still exists.
It begins with a deceptively simple leadership question:
What if we are solving the wrong problem?
Before leaders ask how much work AI can perform, perhaps they need to ask something more difficult:
If organisations remove the first opportunities through which people learn to work, what else might they be removing without realising it?