(This is the English version of my Afrikaans story in Die Papier on Friday 19 June)
While South Africa celebrates Youth Day this week, the familiar questions about young people and the future of work feel more urgent than ever. We have a generation already contending with high unemployment, weak economic growth, and an education system that too often leaves school-leavers ill-prepared for the labour market. Now artificial intelligence adds another layer of uncertainty.
Will AI take their jobs?
That is probably the wrong question. The better question is: what happens when AI removes the entry-level tasks through which young people once learned how to work?
The public debate about artificial intelligence and employment usually swings between two extremes. In one version, AI will cause mass unemployment and render most human workers obsolete. In the other, AI is simply another tool — like the calculator or the spreadsheet — and the labour market will adapt as it always has.
The emerging evidence suggests both views are too simplistic.
The more useful way to think about AI is to start not with jobs, but with tasks. Jobs consist of bundles of tasks. A journalist conducts interviews, reads, writes, edits, fact-checks, and exercises judgement. A software developer writes code, but also interprets a client’s needs, debugs, documents, collaborates, and makes trade-offs. A teacher explains, assesses, motivates, and manages a classroom. AI does not simply arrive and take over “the job” in one clean sweep. It enters specific tasks: writing, coding, summarising, image generation, translation, data analysis, search, and administration.
This is already happening. Evidence summarised by Marginal Revolution blog points to the rapid adoption of generative AI, particularly in writing and software-related tasks. The important distinction is between augmentation and automation. In some cases, AI helps a worker do a task better or faster. In others, it performs the task with limited human involvement.
So far, the evidence does not support a simple story of mass unemployment. Studies on the use of AI chatbots in occupations exposed to them find surprisingly small effects on wages and hours worked, even where firms have actively encouraged their use. Productivity gains appear real, but modest, and do not automatically translate into higher compensation. That is a useful caution against the exaggerated claim that AI has already transformed the entire labour market.
But it would be equally wrong to conclude that nothing is happening.
The pressure appears greatest at the edges of the labour market. Freelancers in exposed occupations have already experienced lower demand and income. In some US payroll data, young workers early in their careers in AI-exposed occupations have seen relative declines in employment, while more experienced workers in the same broad fields have been less affected. Evidence from startup firms also suggests that companies exposed to generative AI can grow faster while hiring fewer junior and implementation-level workers.
That is the part of the story that deserves more attention. AI may not take all the work. But it may remove some of the tasks through which young workers once learned.
That is a serious problem. Entry-level work has always involved a degree of inefficiency. The junior analyst prepares the first draft. The intern cleans the data. The young lawyer reviews the documents. The trainee accountant checks the spreadsheet. These tasks are not glamorous, and many of them can be automated. But they are also training grounds. They teach the habits, context, and judgement that later become expertise.
If AI removes too many of these tasks, firms may become more productive in the short term while weakening their future skills pipeline. The question is not only “how much work will be lost?” It is also “how will people become good at their work?”
For South Africa, this matters more than in countries with tight labour markets. We already have exceptionally high unemployment, particularly among young people. We also have large gaps in digital access, unequal school outcomes, and weak pathways from education to employment. If AI raises the productivity of skilled insiders while making it harder for outsiders to enter, it could deepen existing inequalities.
At the same time, it would be a mistake to reject AI. The technology can help small businesses write better proposals, automate basic administration, improve client communication, analyse data, and access expertise that was previously expensive. It can support teachers, students, researchers, health workers, and entrepreneurs. In a low-growth economy, productivity tools should not be dismissed.
The real question, then, is not whether government can design the perfect AI labour market strategy. In a country with a weak state and limited implementation capacity, that kind of answer can sound hollow. The more immediate question is where young people, educators, businesses, and entrepreneurs can identify the gaps — and turn them into capabilities and opportunities.
The first gap is AI literacy. This is no longer just a technical skill for programmers. Young people need to know how to use AI critically: how to ask better questions, verify outputs, protect privacy, improve drafts, and combine machine speed with human judgement. There is already room for short courses, tutoring services, and workplace training that help people use AI well, rather than merely using it casually.
The second gap is judgement. If AI can produce a first draft, the valuable worker is the one who can tell whether the draft is good. That requires literacy, numeracy, subject knowledge, and communication. These older skills become more important, not less. Here is an opportunity for schools, universities, training providers, and employers to focus less on producing answers, and more on evaluating, improving, and explaining them.
The third gap is work experience. If firms automate junior tasks, young workers will need new ways to build portfolios. Rather than simply waiting for formal employment, they may need to demonstrate evidence of completed projects: a small business’s website improved, a dataset analysed, a proposal written, a marketing plan tested, or a community organisation assisted. AI makes some of this easier. It lowers the cost of producing visible work.
The fourth gap is AI adoption by small businesses. Large firms will experiment with AI regardless. Many small businesses will not know where to begin. This creates an opportunity for young people who can become AI translators: people who help a local business improve their invoicing, inventory management, social media, client communications, or bookkeeping. In a country where formal employment is scarce, the ability to sell practical AI-assisted services could matter considerably.
This is not a reason for panic. But it is a reason to prepare.
Youth Day reminds us that young people are not merely future workers. They are also the people who must navigate the institutions, technologies, and failures of the present. AI will probably not create a future without work. The more likely future is one in which work is reorganised around people who know how to use AI well.