AI Will Create Jobs — Just Not for the People Losing Them

The World Economic Forum projects 12 million more jobs created than destroyed by 2030. That aggregate is accurate and almost entirely useless as a governance response to the workers, communities, and organisations experiencing displacement right now. The unit of analysis is wrong.

Sign up for my Substack daily AI newsletter here.

See my AI Training course portfolio for corporate Business Leaders here.

Article link: https://open.substack.com/pub/johanosteyn/p/ai-will-create-jobs-just-not-for?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/

Every major technological revolution has created new jobs. The statement is historically true, economically defensible, and increasingly useless as a response to AI displacement. The reason is not that the aggregate projection is incorrect. The WEF’s Future of Jobs Report 2025 projects 170 million new jobs created against 92 million destroyed by 2030 — a net positive of 78 million. The reason is that aggregate job creation is a macroeconomic description being used as a governance answer to a microeconomic problem. The workers losing jobs in 2026 are not waiting for the long-run macroeconomic equilibrium. They are managing mortgage payments, school fees, and household expenses in the gap between the displacement that has already happened and the creation that is still a projection. The job-creation counterargument ends every conversation it enters. It should be starting one.

The question that the counter-argument does not answer is the governance question: who bears the cost of the transition between destruction and creation, what is the plan for the workers who cannot bridge that transition, and what accountability exists for the organisations, policymakers, and board members who approved the AI strategies producing the displacement while citing the aggregate projection as their justification?

CONTEXT AND BACKGROUND

The job creation counter-argument rests on a historical analogy that has three structural assumptions. Previous technological waves satisfied all three. AI satisfies none of them.

The first assumption is about time. The transition from agricultural to industrial employment took generations. The transition from industrial to knowledge economy employment took decades. Those adjustment periods were long enough for education systems to redesign curricula, for social safety nets to be built and extended, and for labour markets to develop the training institutions that connected displaced workers to new roles. The current pace of AI capability development is materially faster than either previous wave. The IMF’s January 2026 Staff Discussion Note on new job creation in the AI age states this directly: disruptions are more pronounced when change occurs over shorter horizons, while slower transformations that occur over the course of a working generation can more easily be absorbed through the turnover of the labour force. The adjustment period is not decades. It is compressed into corporate budget cycles.

The second assumption is about skill bands. Previous automation waves destroyed and created jobs in broadly adjacent skill categories. Industrial automation replaced unskilled agricultural labour with semi-skilled factory roles. Computing automation replaced routine clerical work with knowledge worker roles. The skill step required to move from the destroyed job to the created job was manageable within a retraining timeframe. AI is automating entry-level cognitive tasks — document review, data processing, routine analysis, basic customer interaction — and creating roles that require postgraduate qualifications. The IMF’s analysis found that 85 per cent of workers who possess the new skills employers are demanding hold at least a bachelor’s degree, compared with 60 per cent among workers with only existing skills. The skill step from the destroyed job to the created job is not a retraining programme. It is a multi-year educational commitment that most displaced workers cannot afford financially, temporally, or practically.

The third assumption is about who fills the new roles. The historical analogy holds at the aggregate national level. It has consistently failed at the individual, community, and regional level — and governance operates at the community and organisational level, not the national aggregate. The workers displaced by previous waves of automation were frequently not the workers who filled the new roles. The communities that lost manufacturing employment frequently did not become the communities that gained technology sector employment. AI is reproducing this pattern at a greater speed, with a wider skill gap, and in a context where the safety nets designed to manage previous transition costs are already under pressure.

INSIGHT AND ANALYSIS

The IMF’s 2026 analysis provides the most specific and most alarming data available on who is experiencing the displacement now. Early-career workers aged 22 to 25 in the most AI-exposed occupations have experienced a 13 per cent relative decline in employment since the release of ChatGPT. In the medium term, regions with greater demand for AI-related skills show employment levels 3.6 per cent lower for occupations highly exposed to AI with limited scope for complementarity. Critically, the IMF found that the demand for new AI-related skills has so far not boosted overall employment in local labour markets — while the demand for non-AI new skills has. The new AI jobs are not replacing the lost AI-adjacent jobs at the local level where displacement is experienced.

PwC’s 2026 Global AI Jobs Barometer, analysing more than one billion job advertisements across 27 countries, provides the structural finding that explains why. AI is creating a two-track labour market simultaneously. Professionalised roles — where AI amplifies expert judgment, creativity, and leadership — are growing twice as fast as democratised roles and commanding 42 per cent higher wage growth. The average wage premium for AI skills reached 62 per cent globally, ranging from 118 per cent in consumer markets to 16 per cent in government and public sector work. Jobs requiring specific AI skills grew 69 per cent in 2025 — almost eight times faster than the overall labour market at 9 per cent. AI-exposed entry-level roles are seven times more likely to require traditionally senior-level skills — judgment, leadership, strategic thinking — than non-AI-exposed entry-level roles. These roles grew 35 per cent since 2019. Other entry-level roles declined 10 per cent.

The two-track finding is the most precise available description of why the aggregate projection and the individual experience diverge so dramatically. The worker displaced from a democratised role — routine data processing, basic document management, transactional customer support — cannot step directly into a professionalised role requiring judgment, leadership, AI fluency, and postgraduate qualifications commanding a 62 per cent wage premium. The aggregate is net positive. The individual transition is structurally blocked. The WEF’s 78 million net positive is real — and it describes a future that the majority of workers currently experiencing displacement cannot access without support that most organisations and governments are not yet providing.

For South Africa specifically, the IMF’s analysis includes a finding of direct significance. South Africa is identified as an emerging market economy that demands new skills at about half the rate of advanced economies and experiences average lags of eight to nine months in adopting new skills relative to the United States. This means that the adjustment mechanisms available to advanced economies are less developed, and that the displacement cost will be borne more directly by workers and communities without the institutional infrastructure to support the transition. The Stats SA Quarterly Labour Force Survey for Q1 2026 confirms the baseline against which AI displacement lands: the national unemployment rate reached 32.7 per cent, while youth unemployment among those aged 15 to 24 reached 60.9 per cent

South Africa’s public sector — its largest formal employer — sits at the bottom of the global AI skills wage premium table at 16 per cent.

Its BPO sector sits squarely in the democratised category, facing the slower-growth track. The entry-level collapse the IMF and PwC have documented is not an abstract labour market trend for South Africa. It is the destruction of the primary mechanism through which young South Africans access formal employment, in a country where 60.9 per cent of those aged 15 to 24 are already unemployed.

IMPLICATIONS

The governance implications of the job creation argument’s inadequacy is specific and have three dimensions for South African boards and executives.

The first is about the unit of analysis. Boards approving AI strategies on the basis of aggregate job creation projections are using the wrong metric. The metric that governs — the one that determines whether the organisation has met its obligations to its stakeholders under King IV’s stakeholder-inclusive approach — is not the national net employment figure. It is the specific impact on the specific workers in the specific communities affected by the specific AI deployment being approved. The national net is someone else’s accountability. The organisational displacement is the board’s.

The second is about the transition plan. The organisations that deploy AI tools, eliminating entry-level and mid-level roles while investing the productivity savings in genuine, funded, accessible retraining programmes — not the notional upskilling commitments that appear in AI governance documents and are not resourced in HR budgets — are managing their governance obligation. The organisations that cite the aggregate job creation projection as justification for displacement without a transition plan are not. The IMF’s analysis makes the practical requirement specific: policies that support worker mobility across occupations, expand retraining access, and strengthen transitions from displaced roles to available ones are the governance infrastructure that the jobs creation argument implicitly assumes exists and that most South African organisations have not built.

The third is about the entry-level pipeline. The PwC finding that AI-exposed entry-level roles now require skills that were previously senior-level, combined with the 10 per cent decline in non-AI-exposed entry-level roles, means that the traditional career development pathway — enter at an accessible level, develop capability through experience, progress over time — is being dismantled at its base. For South African organisations dependent on that pipeline for their mid-level and senior talent in ten years’ time, the entry-level collapse is not a social responsibility concern. It is a strategic capability risk. An organisation that eliminates its entry-level pipeline today is defunding its senior talent supply for the next decade.

CLOSING TAKEAWAY

The job-creation counterargument is not wrong. In the long run, across the economy, at the aggregate level, AI will almost certainly create more jobs than it destroys — as the WEF’s 2025 Future of Jobs Report confirms with a net positive of 78 million. That projection is useful for macroeconomists, for long-run policy design, and for technology executives who need a response to displacement concerns that ends the conversation quickly.

It is not useful as a governance framework for the workers experiencing displacement in 2026, the communities where that displacement is concentrated, or the South African boards that have approved the AI strategies producing it. A net positive of 78 million jobs at the global aggregate level means nothing to the 60.9 per cent of South African youth aged 15 to 24 who are unemployed, to the early-career workers who have experienced a 13 per cent relative employment decline since ChatGPT’s release, or to the workers in democratised roles who cannot access the professionalised track where the premium is concentrated. The question that the aggregate projection cannot answer is the one that governance requires: if the jobs AI creates arrive later, farther away, and at higher skill levels than the jobs it destroys, what exactly is the accountability plan for the transition in between? Every South African board that has approved an AI strategy should be able to answer that question. Most of them cannot. The job creation argument has allowed them not to try.

Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net