Reliable power, digital infrastructure and effective adoption by businesses will determine how widely AI’s economic gains spread, IMF and OECD officials told Reuters. Their 22 October interviews described a gap between widespread consumer use and the deeper changes in workplaces needed to lift productivity.[1]
Era Dabla-Norris, deputy director of the IMF’s Fiscal Affairs Department, said effective diffusion could add as much as 0.5% to annual global growth over the medium term. Measured economy-wide productivity gains remain much more modest. The distinction is between the potential after broad adoption and the changes already visible in economic data.[1]
Different estimates answer different questions
OECD chief economist Stefano Scarpetta said use in core business functions lags consumer adoption. The OECD’s estimate of a possible 0.4%–1.2% contribution to annual productivity growth applies to its member economies. It covers a different population and outcome from the IMF estimate, so the two ranges should be read separately.[1]
Three numbers that need different labels
| Measure | Reported estimate | Population and meaning |
|---|---|---|
| IMF potential contribution | Up to 0.5% annually | Global economic growth, medium-term scenario |
| OECD potential productivity addition | 0.4–1.2% annually | OECD-member labour productivity growth, depending on assumptions |
| IMF job exposure | About 40% | Jobs worldwide potentially affected, not automatically displaced |
The IMF’s estimate that roughly 40% of jobs worldwide are exposed to AI includes work that could be complemented or reorganised as well as automated. Dabla-Norris highlighted entry-level tasks through which workers traditionally build expertise. When software performs the initial analysis or routine first pass, employers also need a way for newer staff to learn the judgment required for more advanced work.[1]
Infrastructure determines where adoption can spread
For lower-income countries, unreliable electricity, expensive connectivity and weak digital infrastructure can prevent ordinary firms from integrating AI into production. Access to a chatbot is easier to provide than reliable systems for planning, inventory or service delivery. The officials’ concern is that these practical constraints could widen the gap between countries able to deploy AI broadly and those left behind.[1]
The same distinction applies within richer economies. A firm can distribute licences while leaving its core processes largely unchanged. To assess productive adoption, it needs to examine which activity changed, what implementation cost and whether output, service quality or time improved. User counts alone answer a narrower question.[1]
The OECD’s September outlook identified AI-related investment as a support for the global economy. Building computing facilities contributes to demand now; using the resulting capacity to improve production across many firms is a further transition. The investment cycle and the productivity gains therefore need to be tracked separately.[2]
That leaves governments and employers with concrete priorities: dependable infrastructure, access for smaller firms and training that preserves routes into skilled work. Spreading useful applications through ordinary institutions is what turns technological capability into a broader economic gain.[1]
Sources & further reading
- IMF, OECD warn AI could widen rich-poor country divide Reuters / Ankita Yadav, via Euronext · 22 Oct 2026 · Accessed 25 Oct 2026
- OECD Economic Outlook, Interim Report September 2026 OECD · 23 Sep 2026 · Accessed 25 Oct 2026