South Africa’s AI strategy must be pro-employment

South Africa has within its means as a democracy the ability to shape how artificial intelligence will impact employment.

The question is not whether AI will transform our economy. It is beyond doubt that AI, and particularly deep-learning “generative AI”, will reset the benchmarks for human labour. The question is whether that transformation will benefit working people or see them displaced in the interests of short-term profitability. 

There are two fundamentally different paths facing our country, and indeed peoples across the world: an anti-employment model that seeks to accelerate profitability through the automation of human labour, or a pro-employment model that leverages AI capabilities to augment human capabilities.

These are not predetermined technological outcomes. These outcomes will be determined by our political economy choices. It is becoming increasingly apparent that, with the right regulations and collective efforts, we can influence a more positive outcome. 

First the bad news. The AI revolution is not like the Industrial Revolution. The Industrial Revolution took effect from 1760 to 1840 and there were massive job losses, but humans and economies had time to gradually reskill. For example, the steam engine took 80 years to reach 50% adoption. By contrast, AI is advancing much faster. McKinsey finds that businesses’ use of generative AI rose from 33% in 2023 to 79% in 2025. According to Stanford University’s AI Index Report, AI system performance increased by 67% in the year to 2025. 

With AI, unlike with the Industrial Revolution, humanity will not have 80 years to adjust. 

Misplaced optimism based on an Industrial Revolution analogy can be found in the 2025 Future of Jobs Report by the World Economic Forum. It estimates that by 2030 AI will destroy 92 million jobs. But not to worry, it says, AI will also create 170 million new jobs, so we will be 78 million jobs ahead. Unfortunately, the biggest job creation they forecasted was in big data-centre construction and software development. Within a year of that report we see that projection unravelling – with a rapid decline in junior software developer jobs as AI starts to do that work more cheaply. 

And can we simply retrain everyone for this new reality? Certainly, reskilling is critical. But only a fraction of the population can be reskilled at any point in time to match the relentless evolution of AI. 

If this is a process of “creative destruction”, with winners and losers, the distribution of job destruction and job creation will be uneven across countries and demographics. 

South Africa has 34% of working-age people unemployed, rising to 60% for those under 25. The country cannot afford to have AI to deepen unemployment and destroy entire categories of entry-level jobs for young people. 

So, what is the national choice facing us?

The first scenario sees companies using AI to eliminate junior employees, automate customer services and reduce their workforce. Productivity and short-term profits may rise, but employment will decline. As unemployment rises, distressed consumer markets and social instability rise, and the economy suffers in the longer term.

The second scenario sees companies equipping employees with AI tools, enabling inexperienced workers to undertake sophisticated tasks, improving productivity and expanding their business. As workers become more capable, overall economic capability and competitiveness rise.

In both scenarios the country can claim to be “innovating”. But the consequences for society are profoundly different.

Augmentation over displacement

The good news is that the rise of AI can be contested and shaped. The AI industry is in its early and uncertain phase – and it is politically contentious, overindebted, and loss-making. Regulations are being proposed, including from AI companies themselves. 

Globally, there is rising awareness that AI adoption must be steered towards augmenting human labour rather than displacing it. It is becoming clearer that countries can regulate pro-employment AI adoption, where AI tools enable workers to better achieve practical, on-demand competence. 

This process of AI adoption concerns societal “stakeholders” (government, business, social formations) and not merely corporate shareholders.

As with long-term issues such as climate change, shareholder value and short-term profit maximisation fall short. The choices companies make have spillover effects on the rest of society, including future generations. 

All companies therefore operate under de facto social licence. They are required to follow national laws, follow international anti-money laundering laws, pay their taxes, and get fined if they pollute rivers. Beyond compliance, South African business has been a willing leader in corporate governance, such as with the highly adopted King Codes of Corporate Governance. 

This social contract can be incorporated through a combination of laws and regulations (such as an AI Skills Augmentation Act) and corporate self-regulation (such as an AI and Jobs Code of Good Practice). 

There are three main elements to a “Social Contract on AI and Jobs”.

First, the government must treat AI displacement of human labour as a negative externality, like pollution. Where automation causes substantial, avoidable social harm without sufficient compensating benefits, particular AI applications must be restricted, prohibited or taxed. Conversely, pro-employment AI innovation should be incentivised. The tax revenues for such incentivisation could also be raised, at least in part, by the “AI sin taxes”. 

Second, business must take the initiative and define better practice for socially responsible AI innovation. Business needs to show that it prioritises the national imperative for youth employment, and at every opportunity it will augment and not displace workers. Such a voluntary code can be developed promptly and integrated into the King Codes as supplementary guidance. 

Third, employment must be promoted alongside a wider range of income-generating assets, including access to skilling and ownership of one’s own data. South Africa’s AI strategy must incorporate measures similar to Denmark’s recent amendment of its Copyright Act, whereby it gives people “copyright” rights to their face, body, voice, thereby enabling individuals to control commercial or deceptive uses of their identity, receive compensation and pursue meaningful penalties against offenders. 

Waking up

The benefits of responsible AI innovation and particularly AI-augmentation of young people are considerable. It can empower Africa to leapfrog in economic development. For example, while Europe is stuck with traditional old-style banking, banking in Africa operates through smartphones and trading apps. AI combined with the “youth dividend” of Africa could unleash rapid-fire positive changes in the continent.

Internationally, countries are waking up to the fact that it is both possible and essential to regulate AI adoption. China highly regulates what AI companies can do, the US has adopted a laissez-faire approach, and the Europeans are somewhere in the middle. 

Whether South Africa chooses to regulate AI adoption or not, we will be making a choice.

Given that South Africa’s biggest crisis is unemployment, we could become the leading proponent globally of an AI adoption that is explicitly pro-employment. Together, the government and key role players can agree to discourage anti-employment displacement, reward pro-employment augmentation and leverage AI opportunities to build sovereign- and African-competitive advantage. 

With the rise of AI, a collective national commitment to prioritise employment is the most prescient social contract we can make as a young country. DM