Linda Eedes, Investment Executive at Foord Asset Management
The US economy added 162,000 jobs in August 2026, well ahead of forecasts of around 65,000, while unemployment remained at 4.1%. Businesses are still hiring. AI has helped drive growth through enormous investment in semiconductors, data centres and power infrastructure. But what of the fear that it will eliminate millions of jobs?
Barely a week goes by without another headline about AI-related job cuts. Yet there is still remarkably little evidence of a broad employment shock.
If AI were already displacing workers at scale, employment should be weakening most sharply in occupations with the greatest exposure. Research by the Stanford Institute of Economic Policy Research finds that employment in highly exposed occupations, such as travel agents and customer service personnel, remains relatively stable. US job postings for software developers have grown faster than those for other occupations over the past year. Youth unemployment, often seen as vulnerable to the automation of entry-level tasks, is down more than a percentage point.
Radiology shows how difficult it is to translate technological capability into employment forecasts. In 2016, Geoffrey Hinton, often described as the ‘godfather of AI’, said people should stop training radiologists. Within five to ten years, he predicted, AI would outperform humans at reading medical images. Ten years later, AI has become exceptionally good at analysing scans. But the number of radiologists has grown by around 10%, and the profession faces a significant shortage.
Reading scans is only part of the job. Radiologists also consult with other doctors and monitor patients. Some perform hands-on procedures. AI can automate scan interpretation and report writing without eliminating the broader role. Meanwhile, faster and cheaper imaging has increased demand for scans, adding to the workload.
This distinction between tasks and jobs matters. Economists Daron Acemoglu and Pascual Restrepo describe how automation can displace workers, make them more productive and create new tasks for people. Software development illustrates the point. AI can write large amounts of code, but developers must direct the tools, check their work and fit the output into larger systems. New roles are also emerging in building AI models and helping businesses use them.
Company surveys offer further clues. In an August New York Fed survey, more than 60% of service-sector firms reported using AI, but only 4% said they had laid off workers because of it. More than a third were training staff to use the tools. So far, many firms appear to be adapting their workforce as they adopt AI.
None of this means AI will not destroy jobs. Technology has eliminated occupations before and will almost certainly do so again — but its effects on employment are more complicated than the headlines suggest.
For investors, the lesson is how hard it is to forecast what follows a technological breakthrough. Hinton was broadly right about AI’s capabilities, but much less accurate about their effect on employment. What happens to productivity and wages when millions of people can do their jobs faster and better? How will they use the time freed up? The answers will help shape economic growth.
The gap between expectations and reality reinforces the need for investment portfolios that can perform across a range of outcomes. Valuation is always central. The price paid determines how much good news an investment already allows for. Where prices assume near-perfect execution, even a small disappointment can prove costly.
At Foord, diversification reduces our dependence on any one technology or forecast. We also scrutinise the economics of the businesses we own. Enormous spending on AI does not automatically mean attractive shareholder returns. Companies still need durable competitive advantages and adequate returns on the capital they invest.
AI is likely to be transformative, but its path will be unpredictable. Investing requires conviction; it should not be confused with certainty. Our aim is to compound capital at an acceptable rate even when some assumptions prove wrong, while capturing the upside when we are broadly right. As the old saying has it, it is better to be roughly right than precisely wrong.
ENDS
