Will AI take your job? The early UK data is mixed

TL;DR:

• Stanford analysis of US employment data finds a 2.7% hit to jobs for 22-to-25-year-olds since ChatGPT went mainstream, rising to 12.8% in the most exposed sectors like finance, software and creative work.
• The UK looks particularly exposed on OECD measures, thanks to its heavy concentration in service-sector roles.
• But the economics cut both ways: soaring “agentic” compute bills have pushed major firms to ration AI use, suggesting virtual workers can cost more than human ones.

The BBC’s data-led survey of the augment-versus-replace debate lands on a genuinely uncertain picture. Model capability is climbing fast — systems that three years ago handled tasks taking humans seconds or minutes now complete work taking an hour or more, and the same curve is appearing, earlier, in financial analysis, entry-level legal work and creative jobs. The direction is not in doubt; the scale and speed are.

The cost ceiling

The most important nuance is financial. Firms have burned through trillions — sometimes quadrillions — of tokens on agentic use, running up bills large enough that many have started rationing their most advanced models. That matters because it hints at a limit: if an AI agent costs more than the employee it replaces, the business case for automation weakens sharply. The escape hatch, the BBC notes, is cheaper models — including freely available Chinese ones — which could reopen the cost gap that today constrains automation.

For the UK specifically, exposure is structural. A service-heavy economy sits squarely in the path of the roles most affected, and the OECD data showed a notable UK hit even before last year’s National Insurance rise — making it hard to pin the effect on tax or interest-rate changes alone.

Looking forward

The honest summary is that trends are emerging faster than conclusions. Nobel-winning economists have urged action “now” to steer AI toward rising living standards rather than displacement, and London’s response — a £30m taskforce to retrain exposed workers — reflects that urgency. For UK workers, the near-term reality is neither apocalypse nor non-event: uneven pressure on exposed, junior roles, shaped as much by compute economics as by raw capability.