World Bank Says AI Could Compress a Century of Development Into a Decade

The World Bank’s World Development Report 2026, released August 4, makes a claim that should sound either thrilling or terrifying depending on your vantage point: artificial intelligence could let developing countries accomplish in a decade what might otherwise take a century. The catch is the one nobody wants to hear — they need electricity, internet, and competent institutions first.

The report is the first comprehensive assessment of AI’s implications for developing countries, and it arrives at a moment when the global AI conversation has been dominated by frontier model arms races, trillion-dollar infrastructure plans, and regulatory turf wars between the US, EU, and China. The World Bank’s argument cuts across that grain. Developing countries don’t need to build frontier models. They need to adopt small, low-cost AI tools and adapt them to local conditions.

The Numbers That Matter

The report’s headline finding is that 14.2 per cent of jobs in high-income countries are at risk of automation by generative AI, compared with just 4.5 per cent in low- and middle-income countries. That gap isn’t because developing countries are ahead — it’s because they have fewer formal, automatable jobs to begin with. The more striking number is on the upside: 16.2 per cent of jobs in developing economies could see productivity meaningfully boosted by AI, close to the 18.7 per cent expected in high-income countries.

In other words, the asymmetry favours the developing world on the margin. Less to lose, almost as much to gain. The World Bank’s Chief Economist Indermit Gill put it plainly: “They do not need large models or big data centers to reap its benefits. By adapting small, low-cost AI tools to local conditions, they can bring better medical care, education, judicial services and agricultural extension within reach of millions.”

That framing matters. It directly challenges the narrative that AI supremacy requires frontier model development, exascale compute clusters, and the kind of capital expenditure only a handful of companies and countries can muster. The World Bank’s report argues the opposite: the greatest returns come from adaptation, not invention.

The Three-Step Path

The report sets out a sequenced approach: adopt available tools, adapt them to local conditions, and over time advance toward frontier AI development. This is not a suggestion that developing countries should aim lower. It’s an argument that attempting to replicate frontier AI before the foundations exist is wasteful and inefficient.

The examples are concrete. AI tools can help doctors diagnose patients in countries where there are few physicians. They can improve crop decisions for farmers who don’t have access to agricultural extension services. They can help governments improve tax collection, social programmes, and disaster response — functions that in many developing economies are hampered by limited public capacity and unreliable records.

The report also cites the Stanford HAI 2026 AI Index for context on the pace of AI diffusion, noting that AI is spreading faster and is more context-specific than earlier general-purpose technologies like electricity and the internet.

The Foundation Gap

None of this works without the basics. In Sub-Saharan Africa, nearly one-third of rural schools lack reliable electricity, and more than two-thirds lack dependable internet access. The World Bank is working through Mission 300 to provide energy access to 300 million people across the region by 2030, which would lay the groundwork for broader digital and AI inclusion.

The report also highlights the need for local data, including in local languages, so AI tools can be tailored to serve specific populations rather than defaulting to the cultural and linguistic assumptions baked into models trained primarily on English-language data. This connects to a broader concern about AI monoculture — if every model reflects the same training distribution, the tools don’t work as well for the people who need them most.

Gaurav Nayyar, director of the report, said: “The window to get this right is narrow. AI presents a once-in-a-lifetime opportunity to solve problems that have resisted solutions for generations.”

The Risk of Widening Gaps

The report doesn’t sugarcoat the downside. The most advanced AI systems are being built by a small number of countries and companies. Without deliberate action, AI could widen gaps between countries, increase inequality within them, concentrate market power, weaken trust in public institutions, and create new risks for safety, rights, and social cohesion.

This is the tension the report holds without resolving. The technology that could compress a century of development into a decade is controlled by a handful of actors who have no particular incentive to prioritise developing-world needs. The adaptation-first strategy is a pragmatic response to that reality — but it depends on the tools being available, affordable, and adaptable enough to work in low-resource settings.

What This Means for New Zealand

New Zealand sits in an unusual position relative to this report. We’re classified as a high-income country, which means the 14.2 per cent automation risk figure applies to us. But we’re also a small economy that has never built a frontier model and likely never will. The adapt-don’t-build strategy the World Bank recommends for developing countries is, in practice, what NZ is already doing — adopting overseas-built AI tools and adapting them to local conditions.

The difference is that NZ has the electricity, the internet, and the institutions. The question is whether we have the local data, the local language models, and the policy framework to make sure the adaptation actually serves New Zealanders rather than defaulting to whatever the largest foreign models produce. The sovereign AI conversation has been circulating in NZ policy circles for over a year. The World Bank report gives it a new frame: adaptation isn’t a consolation prize. It’s the strategy that works.

The productivity angle is relevant too. NZ’s productivity growth has been anaemic for decades. The report’s finding that 16-18 per cent of jobs could see meaningful productivity gains from AI is, for a country that has been searching for a productivity lever since the 1990s, worth paying attention to. The ASB productivity bootcamp for SMEs is one early example of how that might work in practice.

The Pattern Beyond the Report

What stands out in the World Bank’s framing is the explicit rejection of the idea that AI value flows from the frontier. The report’s argument aligns with a broader shift in how AI’s impact is being understood — that the greatest gains may come not from the most powerful models, but from the most appropriate ones, deployed in the contexts where the gap between current capacity and potential improvement is largest.

This connects to work on Maori data sovereignty and indigenous AI voice models, where the challenge is not access to frontier models but the ability to shape AI tools around local data, local language, and local governance. The World Bank report doesn’t name that work, but its logic points in the same direction.

The report also implicitly pushes back against the concentration narrative in AI policy — the idea that a few countries and companies will control everything, and everyone else will be a consumer. The adaptation-first strategy says the value is in the deployment, not just the model. That’s a different power structure, and one that developing countries can actually influence.

❓ FAQ

Does the report say AI will replace jobs in developing countries? No. It says only 4.5 per cent of jobs in low- and middle-income countries are at risk of automation, compared with 14.2 per cent in high-income countries. The bigger opportunity is productivity gains — 16.2 per cent of jobs in developing economies could be meaningfully boosted.

What does “adopt, adapt, advance” mean in practice? Start with existing AI tools that are cheap or free. Modify them for local languages, data, and needs. Over time, build toward domestic AI capability. The report argues this is more effective than trying to build frontier models before the infrastructure exists.

Is the World Bank saying big AI models don’t matter? Not exactly. The report acknowledges frontier models are where the capability originates. Its argument is that the economic value for developing countries comes from adaptation, not from competing with OpenAI or Google on model development.

🔍 THE BOTTOM LINE

The World Bank’s report is a counter-narrative to the frontier-model obsession. The countries with the most to gain from AI aren’t the ones building the models — they’re the ones that could use small, adapted tools to close gaps in healthcare, education, and public services that have persisted for generations. Whether they get the chance depends on whether the foundations arrive in time. The window, as the report says, is narrow.

📰 Sources

— CJ Murden, editor of Singularity.Kiwi. Former digital technologies teacher, author of AI-focused books. Writing with a New Zealand focus.