The Door Was Closing at Both Ends and Nobody Told Them

A generation of graduates used AI to get through education without developing capability, and walked out into a job market where AI had already taken the entry-level roles that would have built it.

Audio summary: https://youtu.be/1aHq4TQGbpY

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/the-door-was-closing-at-both-ends?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

There is a particular kind of trap that is only visible once you are already inside it. A student discovers that AI can write a passable essay, draft a competent piece of analysis, or complete a coding assignment with minimal human input. They use it. They pass. They graduate. They arrive in a job market carrying a credential that certifies capabilities they were never required to develop, looking for entry-level roles in which they would have developed them on the job — and find that those roles have been automated away by the same technology they used to get through their degree. The door was closing at both ends. Nobody told them.

This is not a theoretical scenario. The Economist’s May 2026 analysis of graduate hiring finds that AI is not flooding the market with new entry-level roles. It is changing which skills employers want, raising the bar for the roles that remain, and leaving a graduating cohort competing for fewer positions while carrying credentials that an increasing number of employers are learning to discount.

CONTEXT AND BACKGROUND

The data on entry-level job displacement is no longer speculative. A global study by the British Standards Institution, polling 850 business leaders across Australia, China, France, Germany, Japan, the UK, and the United States, found that 39 per cent of business leaders had already reduced or cut entry-level roles due to AI, with 43 per cent expecting to do so in 2026. Anthropic CEO Dario Amodei has stated publicly that AI could absorb roughly 50% of all entry-level white-collar jobs within five years, eliminating the data entry, basic analysis, and research synthesis that constituted a new graduate’s first rung on the career ladder. Monster’s 2026 Graduate AI Readiness Report, surveying more than 1,000 recent and upcoming graduates, found that 89 per cent are concerned that AI or automation could replace entry-level roles, up from 64 per cent in 2025.

The roles under pressure are not abstract. Old Mutual Investments research identifies administrative work, data entry, customer service, junior financial analysis, and basic coding among the occupations most vulnerable to AI-driven disruption, and notes that workers in their twenties and thirties face the sharpest displacement, with fewer pathways into the workforce as companies automate routine tasks. In the technology sector specifically, a major industry warning issued by Lisa Jasper, Head of Talent Acquisition at Dariel, in May 2026 named the problem precisely: AI-assisted tools are increasingly capable of generating boilerplate code, automating routine bug fixes, and completing basic feature development — precisely the tasks that have historically served as the training ground for junior developers, the proving ground where foundational skills are built.

For South Africa, the numbers carry a weight that the global conversation does not always acknowledge. The official unemployment rate stood at 31.4% in the fourth quarter of 2025, with youth unemployment at 43.8% and the broad unemployment rate for youth aged 18 to 35 at 52.43%, with 8.8 million NEET young people remaining stable despite modest gains in formal employment. South Africa’s AI market grew 31 per cent year on year between 2023 and 2024. That growth is not producing a new wave of junior roles. It is raising the bar for the roles that already exist.

INSIGHT AND ANALYSIS

The double trap operates through two mechanisms that are rarely discussed together. The first is educational: students who use AI to bypass the cognitive work of their degrees do not simply save time. They forfeit the specific developmental processes through which critical thinking, analytical rigour, and professional capability are built. An MIT lecturer in fiction writing, Micah Nathan, described this precisely in a May 2026 Guardian essay: the struggle of translating thought into language is not an obstacle to learning but its substance. When a student delegates that struggle to a machine, they do not produce better work. They produce work that has no author, and they graduate without the cognitive architecture that the work was supposed to build.

The second mechanism is structural: the entry-level roles that would have completed that development on the job are contracting. Administrative coordinators, data capture clerks, junior analysts, entry-level customer service agents — these are the positions that have historically provided graduates with their first foothold in professional life, and they are precisely those most exposed to automation. South Africa’s AI market growth is not producing a new wave of junior roles to absorb graduates. It is raising the bar for the roles that already exist, and the graduates arriving at that bar are carrying credentials built partly on work they did not do.

I have previously written about this dynamic from the employer’s perspective — the entry-level work AI is automating was never just about output. It was about building the people who would eventually run the teams. When organisations automate the junior work without replacing it with structured developmental pathways, they do not simply save on headcount costs. They hollow out their own future leadership pipeline. The graduate who cannot find an entry-level role is the visible cost. The senior leader who never developed because the junior role was gone before they arrived is the invisible one, and it will take a decade to become apparent.

IMPLICATIONS

Harambee’s research is precise about where the exposure is concentrated. Over 40 per cent of current tasks in South Africa’s business process outsourcing and IT-enabled services sectors are susceptible to automation, and those most exposed are doing repetitive tasks in roles like customer service, data entry, and administrative support, all disproportionately junior, entry-level positions in which women and youth are highly concentrated. The greatest risk, Harambee identifies, is a mismatch between the pace of task change and the pace of skills development and transition, and this risk is unevenly distributed.

That unevenness matters. The graduate who used AI to get through university and cannot find an entry-level role is not an abstraction. In South Africa, she is likely to be young, female, from a previously disadvantaged background, and entering a labour market that has never had enough room for her. The credential she holds was supposed to be her entry point. AI has made the credential easier to obtain and the entry point harder to find simultaneously. That is not a coincidence. It is the combined product of institutional decisions about academic integrity, employer decisions about hiring and automation, and a national skills development system that has not yet caught up with either.

The responses being offered are real but insufficient at the current scale. The YES and Microsoft AI Skills Initiative is providing 50,000 young South Africans with access to free AI training and a globally recognised certification. That is a meaningful intervention. It is also 50,000 people in a country with 8.8 million NEET youth. The qualification gap and the job gap are both real, and closing one without the other produces a more educated pool of unemployed graduates rather than a solution.

For boards and executives, the strategic question is not whether to automate entry-level work. In many cases that decision has already been made or will be made on cost grounds regardless of this argument. The question is what obligation follows from it. If your organisation is reducing entry-level headcount through AI automation, you are withdrawing from a social compact that South African business has always been part of, whether it acknowledged it or not. The question is what you are putting in its place — structured graduate development programmes, AI-augmented junior roles that pair human judgment with machine capability, or simply a smaller workforce and a talent pipeline that will run dry in ten years.

CLOSING TAKEAWAY

The graduating class of 2026 is not the first generation to enter a labour market disrupted by technology, and it will not be the last. What is different about this moment is the simultaneity: the same technology that is closing entry-level roles is the one that students used to avoid developing the capabilities those roles would have built. The trap is tighter than any previous technological transition, and the consequences for a country with South Africa’s unemployment profile are more severe than the global conversation typically acknowledges.

The answer is not to ban AI in education or to preserve entry-level roles artificially. It is for employers, institutions, and government to acknowledge what is actually happening and to respond with the urgency it requires. A generation of young South Africans cannot afford to be the proof of concept for a transition that nobody planned.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net