Which jobs is AI already taking — and what should teenagers do about it? The 2026 data

Two headlines, one reality

In autumn 2026, the news about artificial intelligence and work sounds like two different worlds. Some write about “the end of entry-level jobs”; others say there is still no mass unemployment from AI. The paradox is that both are right. Let's look at the numbers — no panic, no rose-tinted glasses.

First, it's important to separate two ideas. A job being “exposed to AI” means AI could, in theory, do some of its tasks. That's not the same as “the job will disappear”. Most often the job changes: some tasks get automated, while others become more important.

The scale: how many jobs will AI touch

The International Monetary Fund estimates that AI will affect almost 40% of jobs worldwide in one way or another, and about 60% in advanced economies. Roughly half of those workers stand to gain from higher productivity. For the other half, AI may take over key tasks — which means less hiring and pressure on wages.

In 2025 the International Labour Organization, together with Poland's NASK institute, analysed almost 30,000 work tasks. The finding: one in four workers globally is in an occupation with some exposure to generative AI, but only 3.3% of jobs fall into the highest category, where most tasks could in theory be done without a person. Clerical and administrative jobs are the most exposed. The ILO's main scenario is that work will be transformed, not that people will be replaced.

The first “canaries”: young people at the start of their careers

The most convincing evidence isn't forecasts but real payroll records. Stanford economists Erik Brynjolfsson, Bharat Chandar and Ruyu Chen analyse data from payroll company ADP — millions of workers at tens of thousands of US companies. They called their study “Canaries in the Coal Mine?” — after the birds that were first to sense toxic gas and warn the miners.

The key detail: experienced workers in the very same jobs show no such gap. AI hits hardest where work relies on “book knowledge” — things you can learn from a textbook. Tacit knowledge that comes with experience — understanding people, context and consequences — still protects, for now.

The second key detail: employment falls where AI does the whole task (writing standard code, answering a routine customer query), and holds up or even grows where AI supports a person.

“Young workers who learn how to use AI effectively can be much more productive.”

The official forecast: which jobs are shrinking

And what's growing? Data scientists — about +35%, nurse practitioners — +41%, and home health and personal care aides will add 847,300 new jobs. Overall, the US economy is projected to add 5.9 million jobs over the decade.

BLS is candid that not all of these declines are caused by AI alone, and its new “AI exposure” ranking is not a layoff forecast. But the pattern matches Stanford's: the roles under pressure are those built on standard operations with text and data. Growth comes where human care, complex judgement or building AI itself is needed.

No panic: what the wider statistics show

Does this mean a wave of mass unemployment? Not yet. The Budget Lab at Yale checks US labour-market data every month, and in September 2026 it again concluded that there are no clear signs of economy-wide disruption from AI. Companies are hiring less right now, but there are many reasons — from tariffs to general economic uncertainty.

At the same time, demand for people who can work with AI is surging. PwC analysed more than a billion job ads across 27 countries:

That last number matters most for teenagers. The “first job” isn't disappearing, but the bar is rising: beginners are now expected to show what used to be asked of experienced staff — judgement, communication, ownership of results.

What teenagers and parents can do: 7 recommendations

We wrote in more detail about the skills employers will value through 2030 in “Skills you can't automate”.

Sources