The Jobs Data Has Stopped Agreeing With Itself

One credible study shows AI growing headcount, another shows it erasing jobs by the thousand, and the contradiction itself is what leaders need to understand.

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/the-jobs-data-has-stopped-agreeing?r=73gqa&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

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

The debate about AI and employment has reached a strange place. The data no longer points in one direction, and the more carefully you read it, the more it seems to contradict itself. The temptation is to pick the study that confirms what you already believe. The more useful response is to ask why two credible sources can look at the same technology and reach opposite conclusions, because the answer is the most important thing a leader can grasp about AI and work.

CONTEXT AND BACKGROUND

Consider two recent findings. A study by Ramp and Revelio Labs, drawing on workforce records from nearly 22,000 companies, found that heavy AI adopters grew their headcount by 10.2 percent, with entry-level roles rising 12 percent, directly countering the fear that AI is wiping out junior jobs. Set against that, Goldman Sachs payroll analysis found that AI has erased roughly 16,000 net United States jobs a month, with workers under 30 hit hardest and the wage gap between entry-level and experienced staff widening sharply in the most exposed occupations. Both cannot be wrong, which means the simple story, that AI either creates jobs or destroys them, is the wrong story.

INSIGHT AND ANALYSIS

The contradiction resolves once you stop treating AI as one thing doing one thing. For well-resourced, fast-growing firms, AI is an expansion tool: it makes core work cheaper and faster to produce, which raises the return on growing the whole business, so these firms hire more, including at junior level. For mature firms under economic pressure, the same technology is a substitution tool, used defensively to compress costs and hold headcount flat. The Ramp data even concedes this, noting it skews toward tech-forward firms that were expanding anyway, and that companies which merely run pilots without sustained investment see no headcount gains at all. The dividing line is resources and strategy, not technology. Goldman’s own measured view supports caution over panic, estimating that current AI uses, if generalised across the economy, would put only about 2.5 percent of employment at near-term risk, with displacement of perhaps 6 to 7 percent over a decade.

IMPLICATIONS

This is where the South African picture turns sharp, and not in our favour. The International Labour Organization, in a joint study with the World Bank covering 135 countries, warns that in developing economies disruption may arrive faster than the productivity gains, because digital gaps mean the substitution lands before the expansion does. South Africa is acutely exposed to this disruption without the dividend. It has the white-collar, business-process and entry-level administrative roles that AI substitutes, but not the dense, well-capitalised technology sector that turns AI into headcount growth. The local consequence is already visible: rather than dramatic retrenchments, companies grow output without growing headcount, with one of the country’s largest banks freezing vacancies as AI absorbs the workload, against a youth unemployment rate of 43.8 percent. The danger is the quiet removal of the first rung of the career ladder, the very roles through which young people have always entered work.

CLOSING TAKEAWAY

For a South African board, this reframes the question entirely. The useful question is not the abstract one of whether AI destroys or creates jobs in general. It is the concrete one of which kind of firm you are running: one with the resources and strategy to use AI to expand into new work, or one reaching for it to cut costs because the alternatives have run out. The data has stopped agreeing with itself because the technology does genuinely different things in different hands. The contradiction is not noise to be resolved. It is a mirror, and what a leader sees in it is their own strategy.

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