IMF Cautions on Pace of AI’s Economic Impact
Artificial intelligence (AI) may be on course to join steam, electricity and computing as one of history’s most transformative technologies, but its economic impact is likely to materialise more gradually than the hype suggests. This is the latest caution from the International Monetary Fund (IMF), which is urging policymakers and the public to temper expectations with lessons from previous industrial revolutions.
The Fund argues that throughout history, major technological breakthroughs have triggered both optimism and anxiety, and AI is no different. While the technology has been celebrated for its potential to fast-track scientific discovery, others fear a future defined by job displacement and inequality. Yet, the IMF notes that steam engines, railroads and later information and communications technology (ICT) provoked similarly intense debates in their time.
According to the IMF, economic history provides a consistent message: transformative technologies lift productivity and living standards overall, but the benefits often take time and the transition can be turbulent. Basic economics, the Fund says, suggests that by boosting worker productivity, new technology should raise labour demand and wages. Over the past two centuries, this pattern has broadly held true, economies have grown, and employment has not collapsed under successive waves of innovation.
However, the timeline for impact has often been slower than initially expected. A section of the analysis revisits the work of economic historian Nicholas Crafts, who found that the influence of steam power during the First Industrial Revolution was “slower and smaller than previously believed”, with meaningful gains showing only after 1830. Early adoption was limited, and the wider economy needed time to reorganise before reaping the full benefits.
This historical insight, the IMF suggests, helps explain the modern version of a similar puzzle, the so-called Solow paradox. As growth economist Robert Solow famously remarked, “you can see the computer age everywhere but in the productivity statistics.” Computing and ICT did eventually deliver substantial productivity gains, but not overnight. The Fund argues that AI may follow the same pattern: promising, powerful, but gradual in its measurable economic effect.
The report also emphasises that industrial revolutions are characterised not only by powerful “general purpose technologies” like steam, electricity and ICT, but also by the “invention of a method of invention” breakthroughs that dramatically accelerate how new ideas are generated. AI, the IMF says, appears to meet both of these conditions.
Even so, debates persist over whether today’s innovations can match the world-changing impact of earlier breakthroughs such as electric lighting, sanitation or mass manufacturing. Some economists, including Robert Gordon, argue that technological “low-hanging fruit” has already been harvested.
Beyond productivity, the IMF points to the uneven social consequences that accompany major technological shifts. Drawing on Friedrich Engels’ observations of the early Industrial Revolution, the analysis recalls how some workers prospered while others saw their livelihoods collapse. Mechanisation lowered the cost of yarn and benefited weavers initially, but later waves of automation devastated hand-loom weavers, many of whom endured severe poverty.
Modern economic research echoes these patterns. Studies by Daron Acemoglu and Pascual Restrepo highlight how new technologies often trigger a “displacement effect”, reducing labour demand as machines take over tasks once performed by humans. Yet technologies can also create new types of work the “reinstatement effect” as happened through the 19th and 20th centuries with the rise of engineers, machine technicians, and later, software specialists.
The challenge today, the IMF notes, is that much of recent AI and ICT development has leaned heavily toward automation rather than task creation. This trend, economists warn, risks entrenching wage stagnation, widening inequality and slowing labour demand. Acemoglu and Restrepo even argue that economies may be experiencing “excessive automation”, which could undermine productivity itself unless innovation shifts toward technologies that complement human labour, especially in sectors like education and health.
The report also addresses speculation about an AI “singularity”, a point where machines improve themselves so rapidly that human labour becomes obsolete. The IMF references economist William Nordhaus, who tested the likelihood of such a scenario and concluded that most required conditions remain unmet. Human-centred jobs, from cooking to caregiving, remain beyond the reach of even the most advanced AI systems.
Ultimately, the IMF stresses that unlike societies in the early 1800s, modern governments possess policy tools capable of shaping how technology evolves. Yet the Fund warns that the direction of AI development is currently being set largely by corporate priorities, not broader societal goals. With lessons from earlier industrial revolutions now clearer than ever, the IMF says policymakers have both the responsibility and the means to ensure that AI’s benefits are widely shared, but only if the political will exists.
