The transition to a world of artificial intelligence has given a whole new meaning to the concept that capitalism can only renew itself through creative destruction. This is the idea that clapped-out technologies have to be replaced by new ways of doing things, even though the process can be brutal.

That has been the way of things for every new wave of inventions since the dawn of the industrial age in the mid-18th century, but with machines now displaying cognitive skills, able to both think and learn, the potential for economic disruption is all the greater.

In an ideal world, policymakers would have time to adjust and so make the transition smoother and less painful. There are always teething troubles with new technology, which means businesses tend to change their working practices relatively slowly. That gives governments the space to invest in skills and craft industrial strategies. Only by doing so can the full potential of technological progress be realised.

It also helps if economies are growing and jobs are plentiful, as was the case in the 1950s and 1960s. Full employment then made it easier for workers displaced by automation to find another job. That, though, is not the world in which we currently live. Even before the US and Israel launched their attack on Iran, growth was weak and jobs hard to come by.

That said, the war in the Middle East has further complicated matters. The closure of the strait of Hormuz has led to higher energy prices and shortages of raw materials for use in industry and agriculture. A sharp rise in business costs coupled with the ready availability of new labour-saving technology is a toxic mix that has the potential to destroy jobs rapidly and on a grand scale.

The incentive to choose machines over humans will increase because the outlook is a lot worse than it was only a few weeks ago. The International Monetary Fund this week cut its growth forecasts and highlighted the risks of a global recession, and if the warnings prove correct unemployment will rise sharply. Businesses are going to be even more eager to cut costs, especially labour costs, and will be reluctant to start hiring again, even when business conditions improve.

AI optimists acknowledge there will be short-term problems but say that in the longer term there is nothing really to worry about. In the past, every wave of new technology has been accompanied by predictions of machines supplanting humans, but the doomsday scenarios have never materialised. Some sectors suffered and even disappeared altogether but, ultimately, more jobs were created than were destroyed. That’s because humans have used new technology to find more efficient ways to do things, thus raising productivity and boosting growth. An expanding economy provides job opportunities, even if it is impossible to say at the moment exactly where those job opportunities will be.

There are two potential problems with this analysis. The first is that there is no guarantee that history will repeat itself. AI’s impact may prove to be more transformative and much more disruptive than, say, the internal combustion engine. It may be different this time.

The second potential problem is that the jobs destroyed by AI may prove to be better paid than those created. In the past, this has not been the case, with labour-saving machines freeing up humans to do more creative tasks. The opposite could happen this time, with machines doing the clever stuff and humans left with more menial tasks.

In the end, it will depend on whether AI allows men and women to do their jobs better or whether it makes them redundant. In the first case, the AI optimists will be proved right. In the second, they will be proved spectacularly wrong.

Before the Iran war started, the research company Citrini came up with a possible scenario in which there was an AI-driven economic and financial crisis in 2028. Automation wipes out well paid white-collar jobs, which reduces consumer spending power. Machines can work 24/7. They don’t take holidays and they don’t call in sick. But, on the other hand, machines don’t spend money on a new car or a round of drinks in the pub.

So, while it makes sense for each individual company to accelerate its use of AI, the drop in demand for goods and services across the economy as a whole hits corporate revenues, which in turn leads to pressure for further cost-cutting, and leads to more automation and a fresh wave of redundancies. The doom loop created by the adoption of AI eventually leads to a stock-market crash.

Until now, the worry on Wall Street has been of a crash caused by AI failing to live up to the hype. The Citrini report shows what could happen if the claims for smart machines prove to be right. In which case, the countries that have been quickest to embrace AI – such as the US – will be the most vulnerable.

The war is a wake-up call. Policymakers have less time than they might think to ready their economies and their societies for the challenge posed by AI. They need to concentrate on the three Rs: reskilling, reindustrialisation and redistribution. What’s more, they need to act fast. Otherwise any benefits from AI will be captured by a tiny minority, while the majority battles with the consequences of mass unemployment.

• Larry Elliott is a Guardian columnist