IMAGE: CALEB LUKE LIN
THE BOSSES are at their screens. A large new survey of nearly 6,000 chief executives, chief financial officers and other senior managers across America, Britain, Germany and Australia finds that more than two-thirds of them personally use artificial intelligence in a typical working week. The average is about an hour and a half. Some 7% spend more than five hours with AI tools. The technology has arrived in the C-suite.
The survey, conducted by researchers at four central banks and several universities between November and January, offers the first representative picture of how businesses actually deploy AI. Its findings, published as a National Bureau of Economic Research working paper in February, contain plenty of numbers to chew on.
Around 70% of firms now use AI in some form. Text generation using large language models is the most popular application, followed by visual content creation and data processing with machine learning. Younger firms and those with younger directors adopt AI more readily. So do more productive companies and those that pay higher wages.
Yet the bigger picture is more complicated than the headlines suggest. For all the feverish coverage of AI’s potential, its realised impact on business so far is barely detectable. More than 80% of firms report no effect on either employment or productivity over the past three years. Among the minority that do notice a difference, the effects are subdued: a productivity boost of 0.3% on average, and essentially zero for employment. The revolution, it seems, has not yet arrived.
Ask about the future, however, and the numbers shift dramatically. Over the next three years, executives expect AI to boost productivity by 1.4% while cutting employment by 0.7%. For America and Britain the expected job losses are larger still—1.2% and 1.4% respectively. That implies roughly 1.75m fewer jobs at existing firms across the four countries by 2028 if these forecasts prove accurate. The gap between recent experience and future expectation is stunning. Executives appear to believe that the technology which has done little so far will soon transform their businesses.
The more interesting finding concerns who expects what. The survey also asked employees the same questions, via a separate poll of American workers. The results diverge sharply. Employees report using AI at work about as much as executives do, around 1.8 hours a week. They also agree that the past three years have seen only small effects. But they part company on the future. Workers expect AI to increase employment at their firms by 0.5% over the next three years, a full 1.2 percentage points higher than the executives’ forecast. On productivity, the gap runs in the opposite direction: employees expect a 0.9% boost, half a point less than their bosses predict.
The divergence echoes earlier research on automation, which found that those who control the means of production tend to foresee different outcomes from those who operate them. Executives, looking at AI as a tool to slash payrolls and squeeze out efficiency, see a leaner workforce. Employees, who actually use the technology day to day, view it as an assistant that helps them do their jobs better—and perhaps makes them more valuable to their employers. Both could be right, depending on how firms choose to deploy the technology. But they cannot both be right at the same firm.
The survey also shows who is using AI most intensively. Chief executives lead the pack, spending more time with the technology than chief financial officers or other senior managers. That may reflect the nature of the work: CEOs spend their days on writing, strategising and communicating, the sort of tasks at which large language models excel. Younger executives also use AI more heavily, as do those at younger firms. That suggests that familiarity with the technology, acquired either through age or corporate culture, shapes how readily people adopt it.
None of this settles the argument about whether AI will destroy jobs or create them. The survey captures expectations, not outcomes. And expectations can change: between early 2025 and early 2026, British executives’ forecasts for AI-driven job losses nearly doubled, from 0.8% to 1.4%. The direction of travel is obvious enough. Executives increasingly see AI as a substitute for labour, rather than a complement to it. Whether employees’ more sanguine view prevails depends on how firms actually implement the technology. For the moment the bosses and the workers are looking at the same machines and seeing different futures. ■
