AI Job Displacement Hits Young Workers Hardest, New Data Shows

AI capabilities accelerate past human benchmarks

Artificial intelligence models can now complete software development tasks that would take humans several hours, marking a sharp acceleration in capability over just three years. Industry benchmarks show that large language models have progressed from handling tasks measured in seconds to tackling work that previously required an hour or more of human effort, according to data reported by the BBC.

The latest models can identify vulnerabilities in cryptocurrency contracts and optimize their own code—capabilities that signal a fundamental shift in how technical work gets done. Similar patterns are emerging in financial analysis, entry-level legal work, and some creative roles.

Young workers bear the brunt of displacement

Stanford University analysis of four years of U.S. employment data reveals a 2.7% employment decline among workers aged 22-25 in AI-exposed occupations since ChatGPT's widespread adoption. In the most affected sectors—finance, software development, and creative industries—that figure climbs to 12.8%.

The contrast is stark when compared to AI-resistant roles. Health workers, childcare providers, and hairdressers have seen no comparable decline. While some economists argue that interest rate changes could explain part of this trend, the concentration of losses in AI-exposed sectors suggests technology is playing a significant role.

Job postings decline in exposed sectors

Online job listings tell a complementary story. OECD analysis comparing highly exposed sectors like telemarketing and legal services against less vulnerable fields such as construction and food preparation shows notable drops in the UK. The timing is significant: these declines occurred during periods of stable or falling interest rates and preceded recent National Insurance increases, the BBC reported.

The UK's service-sector-heavy economy may leave it particularly vulnerable to AI-driven job displacement on international measures.

Token usage explodes, then hits economic limits

Global AI usage, measured in tokens—small text chunks AI systems process—has surged dramatically in 2025. Major corporations deployed "token leaderboards" to drive employee productivity gains from advanced models, with some companies consuming quadrillions of tokens for "agentic use"—AI agents performing tasks autonomously.

But economics is imposing constraints. The resulting bills grew so large that many companies have begun rationing access to the most advanced models. Some Western firms are shifting to cheaper Chinese-derived models distributed freely to the market. This cost barrier may limit how much work can practically be automated, at least in the near term.

Why it matters

The employment data suggests AI displacement is not a distant threat but an ongoing reality, concentrated among young workers entering the labor force. For business leaders, the tension between AI capability and deployment cost will shape workforce planning decisions. Nobel-winning economists have warned that action is needed now to ensure AI raises living standards rather than simply eliminating jobs. The evidence suggests that window for proactive response is narrowing.

These findings were first reported by the BBC, with analysis from Economics Editor Faisal Islam and the data journalism team.