South Africans are embracing AI, but turning that appetite into economic opportunity will require trained educators, reliable internet, affordable access, recognised skills pathways, scaling of existing financing models and investment in the country’s own languages, according to a newly published research paper commissioned by Collective X.
According to Statistics South Africa, the country’s official unemployment rate stood at 33.6% in the second quarter, with 8.5 million people out of work. For youth aged 15 to 34, the rate is much higher at 47.4%, among the highest in the world. At the same time, AI is already reshaping industries such as call centres that have traditionally created large numbers of entry-level jobs.
Collective X is the NPO that coordinates South Africa’s National Digital Skills Plan for entry-level ICT roles, bringing together employers, funders, training providers and government behind a single, shared approach to closing the country’s digital skills gap, eliminating youth unemployment and boosting economic growth for the country.
“The AI conversation cannot be separated from the jobs and skills conversation,” says CEO Evan Jones. “We already have much of the talent, infrastructure and institutional capacity we need. The priority now is to connect those assets and turn AI literacy into real pathways to work.”
The report, Walking the Tightrope: South Africa’s AI Future, was conducted by Materra Research and sets out two possible futures for the country, based on a review of Africa-focused scientific literature, an assessment of the current AI model landscape and interviews with more than a dozen experts. AI could deepen South Africa’s employment challenges or become a powerful tool for expanding access to skills, work and entrepreneurship. Which path the country takes will depend less on whether South Africans adopt AI and more on the decisions made now about education, access, training and inclusion.
The appetite already exists
South Africans are adopting AI tools ahead of formal policy, curricula and institutional frameworks. The challenge is therefore not persuading people to use AI, but converting that informal adoption into recognised skills and employable capability.
The report identifies educator capacity, data affordability and the limited representation of South African languages in global AI systems as major barriers to doing this at scale. That is why the recommendations begin with a simple principle: train the trainers first.
Teachers, TVET instructors, workplace mentors and other facilitators need the skills and support to help learners use AI effectively and responsibly. Training should then be delivered through the infrastructure people already use, including mobile phones, existing education institutions and zero-rated platforms, rather than waiting for new campuses or expensive technology to be built.
Turning AI skills into economic opportunity
AI training must also lead to outcomes employers and workers can use. A key priority is scaling practical micro-credentials that show what people can actually do with AI through portfolios, workplace verification, practical assessment and other recognised forms of evidence. These credentials should be integrated into existing education and training structures and aligned as closely as possible with employer demand.
But the report cautions against defining success only as formal employment. AI can improve the capabilities of freelancers, independent workers and small businesses. The researchers therefore recommend that entrepreneurship, self-employment and small-business productivity should be treated as important economic outcomes alongside salaried jobs.
South Africa’s languages could become part of the solution
Among the report’s most distinctive findings is the opportunity presented by South Africa’s indigenous languages. Leading AI models still perform significantly worse in African languages than in English, with one benchmark recording gaps of up to 31 percentage points. IsiZulu, isiXhosa and Sesotho each have only around 230 to 290 megabytes of monolingual text available online, compared with roughly nine terabytes for English.
The report argues that this data shortage could be seen as an employment opportunity. Developing high-quality South African language datasets requires people to transcribe, translate, evaluate and create language data. A coordinated public programme could generate income for people while building a national asset that allows more South Africans to interact with AI in their own languages.
A jobs plan, not a technology plan
The findings suggest that South Africa does not need to build an entirely new institutional architecture for AI skills development. It needs to connect and scale what already exists, funded through instruments such as SETA discretionary grants, the Skills Development Levy, the National Skills Fund, Digital Skills Impact Fund, and B-BBEE skills-development spend, combined with industry support for connectivity, devices and other barriers to access. Equally important is being deliberate about which AI systems the country adopts, weighing cost, capability, POPIA compliance and dependency on foreign providers.
“Ultimately, these new opportunities will not emerge automatically or be accessible to everyone,” says Dr Rodney Manyike of the Human Resource Development Council. “The objective should therefore be to ensure that the AI transition happens with the workforce rather than to it.”
For Collective X, the research’s central message is not that AI will inevitably create or destroy jobs, but that South Africa still has the opportunity to shape which of those futures emerges.
The full report is available here .
