The AI Skills Gap: Why SA Risks Training Workers for Obsolete Jobs

CAPE CHAMBER: "We certify people for occupations whose technologies are already changing."  How far behind the global curve is our skills training when it comes to AI? 

LIONEL ADENDORF: The paradox is that South Africa is both advanced and behind. We have excellent researchers, technology companies, and pockets of sophisticated AI training. Yet these remain concentrated in a few institutions and largely outside mainstream undergraduate and vocational education. Our national AI maturity assessment rated education and workforce enablement at 1.9 out of five, between “limited” and “emerging.” The problem is not a complete absence of capability; it is the failure to make AI competence systematically available across disciplines, institutions, languages, and communities. 

CAPE CHAMBER: You have suggested a national audit of AI-era qualifications. Has this idea gained any traction within Government?  Do we have any data on our AI-readiness outside of the global ranking you spoke of in your article? 

LIONEL ADENDORF: I have not yet made this proposal to the Department of Higher Education and Training, but it is a frustration I have expressed before on different platforms. However, there are important initiatives, including the National Digital and Future Skills Strategy, earlier work on the implications of the Fourth Industrial Revolution, and the national AI policy process. Unfortunately, these have not yet produced a systematic review of every major qualification against AI-driven workplace change. 

The national AI maturity assessment provides valuable domestic evidence: education and workforce enablement scored 1.9 out of five, innovation and research 1.6, and growth of the AI ecosystem only 1.3. We have measurement and several strategies; what remains missing is coordinated implementation. 

CAPE CHAMBER: 'Industry must help design curricula'.  Cape Chamber has made a similar call - for the private sector to be more involved in curriculum development and skills-training. How important is this and why?  

LIONEL ADENDORF: Essential, and it’s exactly where Cape Chamber’s position and mine converge. Industry sees changes in equipment, software, occupations, and required skills before they appear in formal curricula. Employers can provide current demand signals, workplace placements, laboratories, practitioners and real problems for students to solve. 

But industry should not write curricula alone. Educational institutions must safeguard academic quality, critical thinking, and the public interest. The strongest model is genuine co-design, in which educators and trainers provide disciplinary depth and teaching expertise, while industry ensures practical relevance. Chambers are particularly valuable because they can aggregate the needs of smaller businesses that cannot sustain partnerships with colleges and universities on their own.

CAPE CHAMBER: 'South Africa has demonstrated the institutional capacity to implement reform; but our problem is ambition without alignment.' Is this another way of saying we talk the talk, but don't yet walk the walk?   

LIONEL ADENDORF: Not quite. I’d gently push back on that framing. The June TVET-reform genuinely happened at real scale, across institutions, capacity and delivery. That’s walking, not just talking. The problem is we walked in the wrong direction. We’ve proven we can execute reform, but we haven’t agreed what we’re executing toward. Ambition without alignment means DHET, the CHE, QCTO, SAQA, universities, and TVET colleges are all individually capable and nobody has directed that capacity toward one shared question: what does an AI-shaped economy actually require of a qualification? 

Alignment, on the other hand, means connecting education policy, industrial policy, AI policy and labour-market demand so that they reinforce one another instead of moving on separate tracks. 

CAPE CHAMBER: 'AI-shaped societies will need ethics, history, law and philosophy.'  Is an 'AI society' by definition a less humane society? Or not necessarily so?  

LIONEL ADENDORF: Not necessarily. But it will drift that way unless we choose otherwise. AI can expand access to education, healthcare, language services and economic opportunity. On the other hand, it can also scale up discrimination, surveillance and exclusion. The technology does not determine which society emerges; institutions, laws and human choices do. 

That is why ethics, history, law, philosophy, and the arts become even more important, not less. They help us ask who benefits, who bears the risk and what must remain subject to human judgement. The humanities are not ornamental subjects alongside technological education. They are part of the intellectual infrastructure required to keep technological progress human-centred. 

CAPE CHAMBER: As latecomers to the world of AI curriculum development, is there possibly an opportunity to learn from best practice?  Or is there a greater danger here of leaving things too late? 

LIONEL ADENDORF: There is a latecomer’s advantage, but only if we move quickly. That advantage decays fast. Every year without a coordinated audit certifies another cohort into occupations that are already changing under them, and it widens the gap with economies that started years ago. The latecomer’s advantage is real. It’s a narrow window, not a permanent cushion. 

But being a latecomer allows us to learn from systems that already embed AI across disciplines, expand work-integrated learning, and assess students through designs, products and real-world solutions. 

However, we should borrow principles rather than copy syllabi. South African curricula must respond to our languages, inequality, the informal economy, infrastructure constraints and industrial needs. The danger is not merely arriving late. It is importing yesterday’s best practices through approval processes so slow that they are outdated by the time they are implemented. Curricula must be easier to update without compromising quality. 

CAPE CHAMBER: How radical do you think this AI revolution is going to be? Some see it plateauing in the near future. Or do you feel we can't yet imagine just how this is going to reshape our society? 

LIONEL ADENDORF: Individual models may plateau, and the truth is that some current claims will undoubtedly prove exaggerated. But I would caution against confusing a plateau in model performance with a plateau in social and economic impact. Even today’s capabilities have barely diffused across most businesses, public institutions and occupations. 

The deeper transformation will come as AI is combined with robotics, sensors, biotechnology and automated decision-making, and is integrated into everyday workflows. Like electricity or the internet, its most important consequences may emerge from widespread adoption rather than a single dramatic technical breakthrough. We probably understand the direction of change better than its eventual scale. 

CAPE CHAMBER: Would you say the AI threats have been overstated, particularly around how it might negatively impact the labour market? 

LIONEL ADENDORF:   It's how it’s often framed publicly, yes, a little.  

The imagery of sudden, mass unemployment overstates how technological transitions usually happen, which is uneven, sector by sector, over years. That’s not reassuring, though. The realistic South African risk isn’t a single dramatic wave of job losses; it’s a slower one, where the routine, entry-level tasks that used to train juniors into a trade are automated first, and we never adjust how we train people for what’s left. For artisans and technicians specifically, that shows up as a qualification that’s technically valid but practically obsolete. 

But in a country with severe unemployment and inequality, even uneven displacement can cause serious harm. The outcome will depend on preparation, worker participation and policy, not on technology alone. 

CAPE CHAMBER: Lastly, the recent shake-up of TVET curriculum did not, as you pointed out, mention AI. Was this a missed opportunity? Does this issue need to be tackled with more urgency? 

LIONEL ADENDORF: Yes, it was a significant missed opportunity. South Africa was already undertaking the difficult work of retiring and replacing legacy vocational qualifications. That was precisely the moment to assess how AI, data, robotics and automation are changing the occupations those qualifications serve. 

This does not mean turning every student into a programmer. It means equipping artisans, technicians, agricultural workers, and administrators with relevant AI literacy, sound data judgement, exposure to new tools, and an understanding of safety, ethics and human oversight. New qualifications should be reviewed now, before they become the next generation of legacy qualifications. The same urgency must extend to universities.