Despite all the noise around artificial intelligence, one of the most important questions that remains difficult to answer is whether AI is actually taking people's jobs in the labour market.
Anthropic Report Findings
A new report from Anthropic, the company behind Claude, concludes that there is no clear evidence that AI has led to a rise in unemployment among the workers most exposed to it. However, there are early signs that hiring in some AI-exposed occupations may already be slowing for younger workers.
That distinction matters because, for much of the past two years, the debate about AI and employment has been between those who believe it will unleash a productivity revolution and those who fear a wave of white-collar automation, with millions of workers displaced by systems that can write, code, analyse, and communicate at ever-increasing speed.
Complex Reality
The truth, as usual, is much more complicated. What makes the Anthropic report interesting is that it goes beyond what AI could theoretically do. Just because AI can help complete a task does not necessarily mean that task is being automated in the workplace.
Businesses do not change overnight. Software has to be integrated, managers have to trust it, employees have to use it, customers have to accept it, legal and regulatory issues have to be addressed, and human judgement still has to be applied. In many cases, the technology may be available long before the organisation knows how to use it.
Anthropic has looked not only at where AI is theoretically capable of undertaking tasks, but also at where it is already being used in real, work-related and more automated ways. This gives a clearer picture of where the labour market may be heading and shows that AI is still far from reaching its full theoretical capability.
Occupations at Risk
That should calm some of the more dramatic predictions of immediate mass unemployment, but it should not make us complacent. The occupations with the highest observed exposure include computer programmers, customer service representatives, data entry workers, medical records specialists, market research analysts, sales representatives, financial analysts and software quality assurance analysts.
This is a revealing list because it is not only about repetitive manual work or low-skilled occupations, but also about office and administrative work, and some of the professional tasks that have traditionally formed the first rung of the career ladder for graduates and younger workers. That may be where the real challenge begins.
Impact on Young Workers
The report does not find a clear increase in unemployment since the launch of ChatGPT, but it does find suggestive evidence that young workers aged 22 to 25 are becoming less likely to start jobs in highly exposed occupations.
That is not the same as mass layoffs, but it could be just as significant over time. AI may not appear in the economy as a wave of redundancies, but rather in the guise of fewer junior hires, a trimmed graduate intake, or an entry-level customer service or analyst position that disappears before anyone notices it has gone.
That matters because the first job is not simply a job. It is where people learn how work actually works, develop judgement, confidence, habits, networks and commercial understanding, and begin to turn qualifications into experience. If AI weakens that first rung, the consequences could be profound.
Implications for Wales
This is especially relevant for Wales, where we already face long-standing challenges around productivity, graduate retention, and access to high-quality professional opportunities outside the strongest labour markets. If AI accelerates the advantage of firms and places that adopt it quickly, then the gap between leading and lagging economies could widen.
Larger firms with the skills, capital and management capacity to integrate AI properly may become more productive, while smaller firms that lack the time, confidence or support to adopt it may fall further behind. Graduates in places with strong professional labour markets may still find routes into work, while those in weaker economies such as Wales may find opportunities narrowing.
With a new Welsh Government in place for the next four years, dealing with this issue could be nation-changing. There is an urgent need to understand where AI exposure is greatest in our own economy, which occupations are most vulnerable, which businesses are using AI to improve productivity and which young people are being prepared for a labour market that is already shifting beneath them.
Skills for the Future
The best workers will not be those who pretend AI does not exist, nor those who use it uncritically, but those who can ask better questions, interpret better answers, spot mistakes, understand customers, and turn information into action. The same applies to businesses, where the biggest opportunity for SMEs is not replacing people but reducing administration, improving sales processes, strengthening customer communication, and freeing owners and staff to focus on higher-value work.
Yet that will not happen automatically. Badly used AI will simply produce bad work faster, creating poor marketing, shallow analysis and false confidence. The firms that benefit will be those that combine technology with good management.
Conclusion
That has always been the real productivity challenge. For Wales, the choice is clear. We can either treat AI as another distant technological fashion, something discussed by academics but not embedded in economic policy, business support or education, or we can recognise that the country cannot afford to lose more of its talent, ambition or productivity potential, and treat it as one of the defining economic issues of the next decade and do something to maximise the opportunity it presents.
