Don’t fear the AI ‘jobpocalypse’
Fears of an AI-driven “jobpocalypse” are rising. Public figures warn that AI could hit labour markets like a tsunami, and the timing feels uncomfortable: unemployment is edging up in many advanced economies, entry-level roles are harder to find, and tech redundancies keep making headlines.
But the big claim — “AI is already destroying jobs at scale” — doesn’t hold up well under scrutiny. Labour markets did cool after ChatGPT’s release in November 2022, yet a slowdown that happens after a technology launch doesn’t prove the technology caused it.
Take the US, where AI investment has been most visible. Some observers argue that booming stock markets and falling job openings since around 2023 are proof that AI is boosting capital returns while squeezing workers. Zoom in, though, and the story shifts: job openings were already declining before ChatGPT arrived. A more straightforward explanation is macroeconomics. The Federal Reserve raised interest rates aggressively, which cools demand — and hiring — across the economy whether AI exists or not.
There’s also a simple post-pandemic effect. Hiring and vacancies surged unusually high after Covid. When that spike fades, the drop can look like a crisis even when it’s partly a return to normal. Similar patterns show up across several G7 countries: cooling labour markets alongside tighter monetary policy.
Some weak spots in the jobs market also have nothing to do with AI. In the UK, for instance, youth unemployment has been linked in part to domestic policy changes like higher payroll taxes. And entry-level positions are often the first to suffer in any hiring slowdown, because companies reduce “training capacity” before they cut core roles.
Another structural factor is “degree inflation”. As more young adults earn university degrees, a degree becomes less distinctive. That intensifies competition for graduate roles — and can push graduate unemployment up faster than overall unemployment. In the euro area, the rise in the share of university-educated twenty-somethings between 2019 and 2024 may be part of that story.
Even corporate layoff headlines can be misleading. Some layoffs are labelled “AI-related”, but that label can serve a purpose: it sounds more forward-looking than admitting weak demand or that the company over-hired earlier. In the data cited, AI-linked layoff plans were only a small fraction of total job-cut announcements.
What about research claiming AI-exposed sectors are already suffering? The evidence is mixed. One study using monthly US vacancies data found no significant impact on job openings or total jobs in sectors more exposed to AI since the technology emerged. And despite constant warnings about white-collar wipeouts, employment in professional, management, and office roles has actually risen overall in the US and euro area since ChatGPT’s release.
Looking ahead, the picture is still uncertain — but it isn’t automatically bleak, especially for younger workers. Historically, disruptive technologies have often favoured younger, well-educated people who can retrain faster and adopt new tools more easily. By contrast, previous tech waves show that older, less tech-fluent workers can face harder adjustments.
More importantly, technology has usually been a net creator of jobs over time. Many modern occupations simply didn’t exist decades ago. New work appears directly (new roles), through specialisation (new sub-fields), and indirectly when productivity gains raise incomes and create demand for new services.
AI may be more transformative than past technologies, but that doesn’t guarantee mass redundancy. Most jobs are bundles of tasks. AI can automate repetitive parts while making other parts more valuable — like judgment, accountability, communication, and human interaction. That “task reshaping” lens helps explain why we’re seeing complex changes rather than a clean “replace humans” pattern.
In practice, AI both substitutes and amplifies. A project manager might spend less time scheduling and more time driving strategy. A financial analyst might build fewer models from scratch and spend more time interpreting, validating, and stress-testing AI-generated output. The role doesn’t vanish — it evolves.
The realistic conclusion is balanced: some jobs will shrink, and disruption can arrive before new roles fully emerge. Skills will shift, entry-level routine-heavy roles may be exposed, and transitions will be uneven. But there’s also time to adapt, because widespread adoption and optimisation take time and the technology is still evolving.
The bigger risk may not be AI itself. It may be whether our education and skills systems move fast enough to help people capture the opportunities AI creates.