AI job replacement hits cost barrier as UK service sector suffers
Companies are scaling back their use of expensive AI agents after racking up massive bills, even as data reveals a growing hit to young employment in Europe's service-heavy economies.
The corporate rush to replace human workers with AI agents is hitting a financial wall. After deploying quadrillions of text tokens to power automated virtual workers, major companies are now rationing access to these models because the costs have become unsustainable.
This presents an immediate challenge for corporate bosses who have pumped vast sums into AI with the expectation of cutting headcount. The prevailing sentiment that "flat is the new up" in workforce size is being tested by the reality that virtual workers are, for now, often more expensive than human ones depending on the task.
The soaring use of these advanced models in 2026 vastly outpaced the declining cost per token. To maximise productivity before bills became untenable, top companies implemented internal "token leaderboards" to drive employee usage. Ultimately, the incredible expenses forced a halt, proving there are clear financial limits to how much work can be automated.
While the financial limits of automation are emerging, the disruption to European labour markets is already measurable. OECD analysis of job postings shows a notable decline in the UK for highly exposed sectors like telemarketing and legal services. This drop occurred while interest rates were stable or falling, and before last year's National Insurance rise.
The UK's heavy reliance on the service sector leaves it uniquely vulnerable to these shifts. US data from Stanford University over four years highlights the specific demographic bearing the brunt of this transition. Employment among 22 to 25-year-olds dropped by 2.7% following the widespread adoption of ChatGPT, spiking to 12.8% in highly exposed fields like finance, software and creative industries.
The emerging data validates recent warnings from Nobel prize-winning economists. They stressed the world "must act now" to ensure AI raises living standards rather than causing mass displacement. Last month, London businesses also reported struggling to find the necessary skills as AI disrupts the jobs market.
The capabilities driving this shift have advanced rapidly in a short timeframe. Three years ago, large language models could only reliably handle tasks taking humans seconds or minutes. Today, they can spend hours finding problems in cryptocurrency contracts or streamlining their own code, with the next generation potentially able to develop themselves entirely within a year.
Faced with the high costs of Western models, many companies are changing tactics. Western firms are increasingly pivoting to cheaper alternatives, including freely available Chinese models, to sustain their automation ambitions without racking up unsustainable bills.
