There is little doubt that artificial intelligence (AI) qualifies as a general-purpose technology (GPT) that is set to profoundly reshape our contemporary economies, just as the steam engine and electricity did in their time. AI can automate not only routine tasks but also some complex cognitive functions: AI agents can analyse huge volumes of data, engage in reasoning, draw conclusions and, to some extent, independently arrive at decisions. Unlike previous industrial revolutions, this technological wave is unprecedented in that it replaces not just physical labour (brawn) but also human intellectual capability (brains).

The effects of AI on productivity and growth will depend on a number of determining factors: the pace of adoption and dissemination; the economic importance of sectors transformed by AI; countries’ preparedness in terms of both digital infrastructure and human capital; and ease of access to key resources, including in particular energy and compute.

For techno-optimists, AI promises to drive productivity gains by enabling a wide range of tasks to quickly be automated. It will also profoundly transform innovation processes by speeding up research and development cycles: the ability to more cheaply test a large number of hypotheses will make AI an endogenous driver of growth.

Meanwhile, more conservatively minded observers anticipate a limited impact on overall productivity. In this view, AI-powered technologies will spread gradually and be concentrated in a few already highly productive sectors. At the same time, the workforce will be redeployed into local service activities, which are labour-intensive but hard to automate and relatively low in productivity. As a result of this structural bias, aggregate productivity gains will remain modest. Furthermore, the cost of AI could hinder the dissemination of these technologies both within and among countries, at the risk of deepening economic and technological inequalities. 

As regards employment, the general consensus is that AI will transform the content of jobs by automating certain tasks, but will not eliminate them altogether. The main expected effect is thus a transformation of roles, with the focus shifting to hard-to-automate skills like creativity, discernment, empathy and relational intelligence. AI thus appears to be a technology that can augment human labour than a pure substitute for it, while giving rise to new roles in the future (prompt engineers, AI solutions architects, algorithm auditors, etc.).

In knowledge-intensive sectors (banking, finance, consulting, legal services), automation will affect analytical and content production functions, while customer relations and high-value advisory services will remain human-driven. In industry, the combination of robotics and AI will make it possible to optimise production, maintenance and quality control processes. In public services such as healthcare and education, AI could play the role of an assistant, taking on some administrative tasks. Conversely, jobs requiring direct physical or human interaction – such as in trades, hotels and catering and the caring professions – will be less affected.

However, AI risks polarising the labour market between, on the one hand, highly qualified jobs that require scarce skills and are hard to automate and, on the other, a multitude of local service jobs that are relatively unproductive and poorly paid. Productivity gains could be mainly captured by the owners of capital, through profits and dividends, and by a minority of talented individuals with highly sought-after skills.

Beyond these economic challenges, AI also raises some major concerns. On the climate front, AI models consume large amounts of energy and scarce resources, which can put pressure on energy infrastructure and lead to resource use conflicts. AI is now concentrated in the hands of a small number of key tech players who hail the coming of a society “governed” by algorithms while fuelling dystopian visions of a future beset with risks. There are also major ethical questions around privacy, transparency of decision-making and the prevention of algorithmic bias. Lastly, the malicious use of these technologies – for example, to engage in large-scale misinformation, political interference and cyberattacks – threatens social stability and disrupts democratic processes.

AI promises unprecedented progress while threatening to destabilise our societies. Only a clear framework for its use and appropriate safeguards will ensure that AI works for humanity and not against it.