Skills reset

AI is undermining the training ground

Skills reset

Jobs. Or the lack of them. That’s the dominant narrative in SA right now. From the desperation of those on or below the breadline, to the quieter anxiety of graduates and white-collar workers, the reality is jobs are hard to come by.

In my own household, that reality is no longer academic. My 24-year-old son has a finance and economics degree and is trying to find his first proper role. He has done what young people are told to do: study hard, craft a CV, apply widely and stay resilient. Yet the door into professional work feels narrower than it should. Is that the economy? Is it artificial intelligence (AI)? Or is something more subtle happening at the bottom of the jobs ladder?

The public debate is all over the place. Anthropic boss Dario Amodei warned that AI could wipe out half of all entry-level white-collar jobs within five years. Nvidia’s Jensen Huang dismissed such fears as ‘complete nonsense’. And then Standard Chartered’s Bill Winters, while outlining plans to cut 7 800 back-office roles by 2030, told investors that AI is ‘replacing in some cases lower-value human capital with the financial capital and the investment capital that we’re putting in’. He tried to walk that back, but there was no unsaying it.

As usual, the truth is more nuanced. At the macro level, the jobs apocalypse has not arrived, or at least not yet. But at the bottom of the professional ladder, where graduates are trying to get a foothold, something is definitely shifting.

PwC’s 2026 AI Jobs Barometer found that entry-level roles are being automated away. The most AI-exposed are now seven times more likely to require traditionally senior skills: judgement, leadership, strategic thinking.

A recruiter I know concurred. Admin work, data entry, customer service, junior financial analysis, basic graphic design and basic coding are the most vulnerable. But, as these jobs disappear, others appear. Cybersecurity, ethical hacking, sustainability managers, drone operators, electric vehicle technicians and professional Twitch streamers (I kid you not) barely featured in career guidance a decade ago. Demand is as high as ever for teachers and healthcare workers. And AI can’t cook, unblock your toilet, care for the elderly or rewire your house.

Here is what AI is exceptionally good at: tasks. Summarise this. Draft that. Run this model. Give it a discrete piece of work, and it will outperform any graduate you’ve just hired.

What it cannot do is judgement. We all know that a job is not just a list of tasks; instead, it involves tasks held together by something harder to name: context, instinct, relationships, the ability to hear what your client hasn’t said. That remains stubbornly human.

But here’s the thing. Entry-level roles have always been task-heavy by design. The junior does the modelling, the research, the first draft, not because it is glamorous, but because that is how we all learnt. You do the work, you get it wrong, someone corrects you, and slowly, painfully, you develop judgement. AI is not just automating those tasks. It is dismantling the learning ground, removing the mechanism by which a young person becomes someone who knows what they are doing. Those replacing the training ground with an AI agent may enjoy a short-term productivity gain. But eventually, they will wonder who is left to run things.

In SA, the skills mismatch is being compounded by a broken vocational training system that is failing the country. It sucks up more than R20 billion a year in public money, with little to show for it. The gap between what our training system produces and what the industry needs was already wide; AI is turning it into a chasm.

In the absence of a credible vocational route, students pile into universities, emerging with vanilla BCom or social science degrees the market does not want, while our economy runs short of the skills it urgently needs. The irony is that a qualified artisan is today more insulated from AI disruption than a marketing graduate.

The advice for graduates is blunt: upskill and keep upskilling. Employers are no longer focused only on the degree you hold. They want to know what you can do. The reckoning is long overdue for the education system. This was a crisis long before ChatGPT, but AI has shortened the window to fix it. We urgently need a credible, properly funded technical education system that takes artisans as seriously as engineers and trades as seriously as theory.

As for our labour laws that make hiring people so costly and complicated that investing in AI agents is becoming the safer option… that’s a column for another day.

By Sasha Planting