The AI Savings Are a Forecast. The Job Cuts Are Real
Boards are acting on projected efficiency, and in South Africa, the human cost of getting it wrong is far higher.
Sign up for my Substack daily AI newsletter here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
More than half of the UK’s large-company finance chiefs now expect AI to reduce their headcount within twelve months. In the same Deloitte survey, two-thirds admit their organisations are only about a quarter of the way through the AI transformation they expect. Read those two numbers together. Companies are planning job cuts on the strength of savings that, by their own account, have mostly not yet arrived. The AI savings are a forecast. The job cuts are real. In South Africa, where a lost job is far harder to replace, that sequence deserves much closer scrutiny.
CONTEXT AND BACKGROUND
The UK is not alone. In the United States, employers cited AI in more than one hundred thousand announced job cuts in the first half of 2026, close to a quarter of the total. White House adviser Michael Kratsios has suggested some companies attach AI to layoffs they would have made anyway, because it “plays better in the press.”
The evidence that AI delivers the savings behind these decisions is thinner. Lawyers at Webber Wentzel, citing MIT research, note that only about five per cent of enterprise AI pilots reach production with a measurable effect on profit. They warn that job cuts based on assumed quick efficiency gains are poor risk management.
INSIGHT AND ANALYSIS
This is where the comparison with South Africa matters. A worker displaced by AI in the UK enters a labour market with far lower unemployment than ours. Here, the official unemployment rate rose to thirty-three point six per cent in the second quarter of 2026. A job cut made on a forecast that proves wrong is not a temporary inefficiency. For many households, it is a permanent loss of income.
South African law also asks for something the global trend skips: evidence. Dismissals for operational reasons require consultation, disclosure and serious consideration of alternatives. Bowmans warns that AI-driven selection with opaque criteria is “vulnerable to challenge as procedurally and substantively unfair.” That makes AI washing a risky strategy here. A board that blames AI for cuts it cannot support with evidence will struggle to defend them.
I have previously written about this in the context of productivity. I argued that time saved is not value created. Hours freed by AI only become value if they produce more output, so boards should ask where the hours went before banking the savings. The same discipline applies to headcount.
There is also a missing step. Mercer found that seventy-one per cent of South African executives put AI’s returns at the top of their agenda, but only forty-one per cent of HR teams are redesigning work to deliver them. Meanwhile, sixty-five per cent of South African workers would trade a ten per cent pay rise for AI training. Companies are reaching for the cut before they have tried the redesign. When that happens, the job loss reflects a failure of management, not a dictate of the technology.
IMPLICATIONS
For boards, the test is simple: prove the savings before cutting the jobs. Pilot first, measure the actual effect on cost and output, and only then decide whether roles change, move or disappear. Where AI does reduce the need for people, test redeployment and reskilling seriously before retrenchment, as our law already expects.
For policymakers and investors, the warning is to look behind the AI label. If companies blame AI for cuts driven by weak growth, we will misdiagnose the problem and erode trust in a technology that could help.
CLOSING TAKEAWAY
I do not doubt that AI will change the shape of many jobs, and some roles will disappear. The question is the order of events. Cutting first and hoping the savings follow puts the risk on workers and the reputational cost on the company. In a country with South Africa’s unemployment, that is not efficiency. It is a gamble with other people’s livelihoods. The savings should be proven before the jobs are cut. Boards that hold to that sequence will make better decisions, and they will have the evidence to defend them.
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
Comments