Companies that adopt generative AI are hiring more people, not fewer. But they are hiring the wrong people for a country like South Africa. A new study of 41 countries found that at firms adopting AI, senior employment rose 6.7% over five years while junior employment fell. The authors call AI “labour saving for junior workers and labour expanding for seniors”.
For a country where almost one in two young people in the labour force cannot find work, that finding deserves more attention than it is getting.
What the study found
The paper, “How Does AI Change Labor Demand? Evidence from 41 Countries”, is by Bharat Chandar of Stanford University and Bouke Klein Teeselink of King’s College London. They analysed 1.25 billion job postings and 154 million employment records between January 2021 and March 2026. They identified AI-adopting firms from job adverts that asked for generative AI skills, such as prompt engineering, large language models or tools like ChatGPT, Claude and Gemini, and then tracked what happened to the workforce at those firms.
The headline results, reported by Bloomberg and carried by Moneyweb:
- Overall employment at adopting firms grew by about 3.3% compared with firms that had not adopted AI.
- Senior employment rose 6.7%.
- Junior employment fell by about 3% (Chandar’s own summary puts it at 2.5%), and the junior share of the workforce dropped by about two percentage points.
- The pattern appeared in 23 of the 31 countries with enough data, including Brazil, Saudi Arabia, the UK and the US. The authors found the losses “run somewhat deeper in richer and more digitised economies”.
The authors also note that their measure probably undercounts AI use, since a company can use AI without saying so in a job advert. In that sense the figures are a floor, not a ceiling.
Why juniors, specifically
The mechanism is straightforward, and anyone who manages a team will recognise it. Much of what entry-level staff do in their first two years is the work generative AI is now good at: first drafts, summaries, basic research, simple code, data clean-up, standard customer replies. A senior person with AI tools can now produce what used to need a senior person plus a junior. The firm grows and hires more experienced people to direct and check the work, but brings in fewer beginners.
The problem is that those first two years were never just about the output. They were how people learned the job. A firm that stops hiring juniors today has fewer mid-level people in five years and fewer seniors in ten.
The South African angle
South Africa is not among the countries the study covers, so there is no local number yet. But the local labour market is exactly where this trend would do the most damage. In the second quarter of 2026, Stats SA’s Quarterly Labour Force Survey put the youth unemployment rate (ages 15 to 34) at 47.4%, up from 45.8% in the first quarter, according to the National Youth Development Agency’s response to the data. Among 15 to 24 year olds, 36.4% were not in employment, education or training. The overall unemployment rate rose to 33.6% in the same quarter.
The jobs most exposed are the ones that have traditionally absorbed young South Africans with a matric or a degree: call centres and business process outsourcing, junior accounting and bookkeeping, paralegal work, junior marketing and content roles, and entry-level software development. These are also the sectors where South African firms are adopting AI fastest. As we reported this week, local businesses are adopting AI faster than they are governing it.
There is one piece of relatively good news in the data. The NYDA figures show graduate unemployment at 12.4%, against 39.2% for young people without matric. Qualifications still matter. The risk is that the first rung of the career ladder, the junior job where a graduate turns a qualification into experience, is the one AI removes.
What employers can do about it
None of this means a business should avoid AI. The same study shows adopting firms growing faster. But how a company adopts it is a choice.
- Keep hiring juniors, but change the job. Instead of cutting the entry-level role, redesign it around checking, correcting and directing AI output. That is a real skill, and someone has to build it.
- Use the learnerships and the tax incentive. The Employment Tax Incentive still reduces the cost of hiring workers aged 18 to 29, and a learnership allowance is available for registered learnerships. Both lower the cost of taking on a junior.
- Measure what juniors learn, not just what they produce. If AI now does the output, the value of a junior hire is the senior they become. Plan for that on purpose.
The study’s own conclusion is that AI is not, so far, destroying jobs overall. It is changing who gets them. In most of the 41 countries studied, that means a harder start for young workers. In South Africa, where the start is already the hardest part, businesses and policymakers should be paying close attention.
