Locked out — How AI automation is shrinking the entry-level market for graduates

More than 200,000 South African students will receive their qualifications between now and June. The entry-level job market awaiting them is smaller, more competitive and steadily being restructured by artificial intelligence (AI).

South Africa’s official unemployment rate stood at 31.4% in Q4 2025, according to Stats SA’s latest Quarterly Labour Force Survey — the lowest reading since Q3 2020, but still among the highest in the world for a middle-income economy. Graduate unemployment is 10.3%, the lowest rate across all education levels. But that number tells only part of the story. The formal-sector roles that graduates have historically used as their entry point into professional life are contracting, and AI is an accelerating factor.

A changing entry-level landscape

South Africa’s AI market grew 31% year on year between 2023 and 2024, according to the Harambee Youth Employment Accelerator. However, that growth is not producing a new wave of junior roles to absorb graduates; it is raising the bar for the roles that already exist.

The positions that once provided graduates with their first foothold – administrative coordinators, data capture clerks, junior analysts, entry-level customer service agents – are precisely those most exposed to automation. AI tools that perform these tasks cost a fraction of a salary, require no leave and run continuously. What remains in many of these roles is work of a different nature: supervising AI outputs, interrogating automated processes, applying judgement where the machine cannot.

The result is that AI literacy is becoming a baseline hiring requirement. Most graduates entering the market this year are not yet fully prepared for workplace AI literacy, facing a gap between traditional education and the rapid adoption of AI in industry.

A digital divide that operates before the interview

South Africa’s graduate employment challenge is not only about the number of available jobs, it is about who can access the ones that remain. AI literacy is built through exposure to tools, platforms, live data environments and current software, and that exposure is not equally distributed.

A graduate from a well-resourced institution with reliable internet access and an updated curriculum competes on different terms than one from an under-resourced institution with intermittent connectivity and course content that has not kept pace with industry. Both hold degrees, but in a market increasingly screened by AI tools, not all degrees are being read as equal.

The gap in practical readiness is widening with each graduating cohort, and employers, whether through automated screening or direct assessment, are registering the difference.

What needs to change

The direction of travel is clear: AI will continue reshaping the economy regardless of policy choices. What is not predetermined is whether that reshaping happens in a way that includes or excludes the majority of young South Africans entering the workforce. That outcome depends on decisions that employers, institutions, and the government are not yet making at the required speed.

For universities and technical and vocational education and training colleges, the gap between what is being taught and what the market now requires is widening with each graduating cohort. Among unemployed youth, nearly six in 10 have no previous work experience – a figure that points directly to the failure of integrated workplace learning to keep pace with employer demand. AI literacy and applied workplace exposure need to be built into qualifications as standard, not treated as specialist additions for students who seek them out.

Funding models are also overdue for reform. The current system ties money to course completion rather than employment outcomes, which means it funds activity without accountability for results.

A shift towards outcome-based funding, where results, not inputs, determine what gets resourced, combined with broader recognition of micro-credentialing, short-course training and work-integrated learning within the levy system, would release funding that the existing framework is too rigid to deploy.

The more uncomfortable truth is that employers will continue to choose the most productive and cost-effective option available to them, and increasingly, that option is automation. Expecting them to absorb underprepared graduates out of goodwill is not a strategy.

The real pressure falls on higher education institutions to produce graduates who arrive ready to operate in an AI-integrated environment not after two years of on-the-job adjustment, but from day one. Until curricula catch up with that reality, graduates will continue to lose ground to the tools that are replacing the roles they were trained for.

The class of 2025

At 10.3%, graduate unemployment remains well below the 37.6% recorded for those without matric – the degree still carries weight. But in a market where AI is absorbing more of the routine work that defined the first two years of most professional careers, a qualification is no longer sufficient on its own to secure a first role. Graduates need the applied skills the market is demanding now, not in the next curriculum review cycle.

South Africa cannot afford to lose another cohort to structural unemployment. Whether the answer comes from employers hiring differently, institutions teaching differently, government funding differently or a combination of all three, the shift needed is one that positions graduates as capable collaborators in an AI-driven workplace, not casualties of it. DM