The World Bank's World Development Report 2026, released today, highlights Bangladesh as a test case for how developing nations can use existing AI tools to improve healthcare and public services-while warning that weak infrastructure and poor data governance could undermine those gains.
For healthcare professionals, the report offers concrete evidence that AI can expand access to care in resource-constrained settings. It also shows what happens when these systems fail.
AI expands diabetic retinopathy screening
Clinics in Bangladesh using AI-powered medical imaging to detect diabetic retinopathy-a leading cause of preventable blindness-have increased patient screening capacity by up to 40% per day. The technology addresses shortages of specialist doctors and radiologists by supporting, rather than replacing, healthcare professionals.
The report points to these applications as proof that AI's greatest value in developing countries lies in expanding access to essential services and improving productivity, not in building costly frontier models. It also notes that AI could raise productivity in 16.2% of jobs in developing countries, close to the 18.7% projected for high-income economies. Only 4.5% of jobs in developing economies are considered at risk of automation, compared with 14.2% in wealthier countries.
For clinicians considering similar tools, the AI for Healthcare resources cover implementation approaches used in real-world screening programs.
Welfare targeting raises red flags
The report cautions against relying solely on AI for social protection programs. A study using mobile phone data to identify poor households found the algorithm correctly identified only 32% of eligible families, leaving 68% unrecognised. Around 6% of households could not be matched with any mobile phone records, increasing the risk of excluding some of the country's most vulnerable people from welfare programs.
Public confidence in the technology remains limited. Only 45% of surveyed households considered the AI-based targeting system fair.
The findings suggest that while AI can improve efficiency, poor-quality data and weak system design can deepen inequalities if adequate safeguards are not in place.
Policy push gathers pace
Bangladesh is preparing a National AI Policy for 2026-2030 to promote responsible, inclusive and accountable use of the technology. The draft policy proposes establishing a National Data Governance Authority to oversee AI deployment and data regulation. It builds on the government's Aspire to Innovate (a2i) programme, which has driven digital public services over the past decade.
Authorities are also forming a multi-stakeholder advisory council comprising representatives from academia, industry and civil society to guide AI governance. Bangladesh is moving towards ratifying the Council of Europe Framework Convention on AI, the world's first legally binding international treaty on artificial intelligence. Those tracking public-sector adoption can follow AI for Government updates on policy and governance developments.
Structural barriers remain
The World Bank says Bangladesh still faces major challenges in scaling AI adoption. In the country's garment sector, automated cutting and sewing technologies have been available for years but remain underused because relatively low labour costs reduce the incentive for automation. Businesses also cite regulatory complexity and compliance costs as barriers to adopting advanced technologies.
Inadequate electricity supply, limited digital infrastructure, skills shortages and weak institutional capacity continue to constrain AI adoption. To improve connectivity, Bangladesh has opened its market to satellite internet providers, including Starlink and OneWeb, to expand broadband access in remote areas. Experts say greater investment is also needed in local data infrastructure, computing capacity and digital skills.
Why this matters for healthcare professionals
The diabetic retinopathy screening program is a working example of AI expanding clinical capacity where specialists are scarce. But the welfare targeting failures show the same technology can exclude vulnerable patients when deployed on incomplete data. For healthcare teams, the practical takeaway is to audit AI tools against real patient populations before deployment, not just against training datasets. The report's findings also suggest that AI governance frameworks-like the one Bangladesh is building-will increasingly shape which medical AI tools get approved and how they are monitored.
"AI has thrown developing economies a lifeline, and they should seize it," said Indermit Gill, the World Bank Group's chief economist. "Countries do not need massive data centres or sophisticated AI models to benefit, but must move quickly to strengthen infrastructure, skills and governance."
Your membership also unlocks:
