In an unusually positive assessment of AI for the world's poorest economies, the world's largest development lender has announced that the technology will bring "new opportunities" to the world's poorest nations.The world's biggest development lender has just issued an unusually positive verdict on artificial intelligence, for the world's poorest economies at least, announcing that the technology will bring "new opportunities" to the world's poor.

The world has never experienced a more transformative technology than AI, but despite its potential to replace many jobs, the World Bank's recently released World Development Report 2026, “The Promise of Artificial Intelligence,” asserts that while high-income countries are more vulnerable to job displacement, low- and middle-income countries, which host over 6.8 billion people, are significantly less affected by existing and emerging risks to work from AI and other technologies. The bank's chief economist, Indermit Gill, put it simply: "Developing economies today have more to gain and less to lose from AI than richer ones do. For the countries that are in the midst of their worst period of average growth in three decades, it's a rare dose of good news.

The figures that support this assertion are of interest. More than three times as many positions in high-income economies are highly exposed to automating with generative AI (14.2 percent) versus developing economies (4.5 percent). The potential upside for productivity, by contrast, is more closely divided: 16.2 percent of developing economies' jobs have the potential for significant gains from AI tools, hardly behind the 18.7 percent figure for workers in the rich world. The logic is very simple when you think about the way these economies are organized. Agriculture and small business are still prevalent in poorer countries, and there simply isn't “as much” routine cognitive work to be automated. The relative losses of richer economies are much greater, as they have many more office jobs, customer service positions, and administrative functions.

However, the same 600-page report contains a warning that contradicts the headlines of optimism, and it falls on the group of countries that have created entire growth strategies around exactly the type of digital work that AI is currently eating away at. For 20 years India, the Philippines and a swarm of new outsourcing centers in Africa and Eastern Europe prepared themselves as a back office for their richer counterparts to run call centers, process invoices, staff help desks, and do entry-level software and finance work. The bank now says that that model is seriously being stressed. The report says outsourced digital jobs in developing countries dropped 39 percent in 2025 alone, and that the sharpest contraction was in those jobs most likely to be affected by generative AI. AI wouldn't mince words: It may open the door to the loss of a viable pathway to the middle class in most developing economies, notably in call centers and some entry-level jobs in software, financial and business services.

The real story here isn't the average level of structural protection in the labor market, but rather a special, high-value part of it that is being eaten away one that has been followed by business analysts for months. According to AlixPartners and other consultancies, generative AI is turning the fundamentals of the outsourcing business on their head. The scheme for thirty years went like this, simple logic: shift standardized, rules-based work to an "affordable" labor market. In an increasing amount of that same work, however, an AI system can perform the same task at almost no additional cost, and without moving any object.

But executives within the outsourcing and staffing community generally have been reacting to the more dire interpretation. The industry's prevailing counter-argument is that the jobs being the most lost are the jobs that are most repetitive, standardized, and less complex, whereas demand is growing for roles that require them to oversee the operations of AI systems, engage in more judgment calls with clients, or work more in the messy, less standardized tasks that AI has not yet mastered. In this perspective, outsourcing doesn't disappear, but rather moves upmarket and is rewarded by countries and companies that invest early in reskilling a workforce towards AI-adjacent rather than towards pure headcount arbitrage as a result of which it becomes more valuable. Whether that will come quickly and widely enough to catch the workers who will be pushed out of the jobs that were previously considered routine is an open question that no one, not even the World Bank, has completely answered.

The report's more positive message is that developing economies have no need to join the U.S. and China in a race to the data center to acquire funding for data centres and training cutting-edge models, which are data intensive and require hundreds of billion dollars. “Low-cost AI tools can have a significant positive impact in healthcare, education and judicial work, as well as in agricultural extension work even in environments where the electric power is scarce and internet connectivity is intermittent,” the bank notes. It refers to the quick spread of the technology, with middle-income countries receiving about half of ChatGPT's global traffic over six months of its launch, the bank says faster than any previous general-purpose technology, such as electricity or the internet, which took decades to reach the same countries.

The bank has a three-stage playbook that it calls adopt, adapt and advance: deploy existing frontier infrastructure and tools, tweak them to local languages and conditions, and only then, when infrastructure and institutions are in place, try to develop homegrown frontier tools. The window of opportunity for this is narrow and countries which lag behind in investing in power, connectivity, skills and institutions are running the risk of falling even further behind the curve, said Gaurav Nayyar, the report's director, in his bluntness.

The report doesn't ignore the geopolitical aspect either. The most powerful AI systems are created and managed by a handful of companies that are mostly based in the U.S. and China, meaning that developing countries that rely on imports of AI face becoming dependent on another type of dependency. A suggestion from the bank is that it should be ‘procurement diversity' procurement models, cloud capacity and tools from multiple vendors (and demand interoperability), so there is no single country's technology stack that becomes a choke point.