In December 2025, Bangladesh signed the National Health Compact at the UHC High-Level Forum in Tokyo, committing to scale up AI-enabled tools in its health system by 2030. The compact names six pillars for health reform, and buried in the sixth is a line that matters more than its placement suggests: the government will support "the scale-up of digital health solutions, AI-enabled tools, and other emerging technologies that enhance equitable, efficient, and high-quality service delivery."
It's the first time AI has appeared directly in a Bangladeshi health policy document. But a compact is not a blueprint. It says AI tools should scale up; it doesn't say which tools, for which pillar, or solving which problem.
Why targeting matters in a constrained system
Bangladesh's public health spending was 0.40 per cent of GDP as of 2021, behind Bhutan at 2.21 per cent, Nepal at 1.80 per cent, and Pakistan at 0.84 per cent. Out-of-pocket costs make up 73 per cent of total health spending, one of the highest shares in South Asia. Health costs pushed 3.74 per cent of the population below the $2.15-a-day poverty line in 2016, up from 3.11 per cent in 2010.
In a system this constrained, a badly targeted tool doesn't just underperform. It wastes money the country doesn't have to waste. That's the argument for getting the "how" right before scaling anything up.
What the compact's pillars point toward
The second pillar calls for expanding community-based and telehealth service delivery, especially in hard-to-reach areas. Researchers have already built a chatbot for dengue symptom triage that runs on simple decision-tree logic in low-bandwidth clinical settings, in Bengali and English. Tested on 2019-2023 case data and piloted with 50 users, three in four said they were satisfied with it. It's a small example, but it points at triage tools that work where the internet is slow and the nearest doctor is an hour away.
The third pillar focuses on redesigning the Essential Service Package around life-course prevention, including maternal health. A tool for maternal health risk prediction combines medical rules a doctor would recognise with a machine-learning model trained on patient data. When 14 doctors reviewed the combined version - one that explains its reasoning in terms they recognise rather than just a risk score - more than half said they would trust it enough to use in practice.
The fifth pillar is where things get more complicated. It commits to establishing a National Health Security Office to manage a National Health Fund and lead strategic purchasing, targeting urban slum residents, widows, people over 70, and people with disabilities. This isn't new: a draft National Health Protection Act has existed since 2014, proposing a similar authority, a health card system, and income categories to determine who gets subsidized care. More than a decade later, the compact picks up where that draft left off, this time with a signed commitment and a 2030 deadline.
Any list sorting citizens into income categories is a judgment call. Someone decides who counts as "below poverty level" versus "marginal income," and that determines who gets a subsidized health card. If digitized without checking for bias first, the system risks locking in existing unfairness at a larger scale and faster pace than a paper-based system ever could.
What would make this real
The compact already has a monitoring framework with targets: raising the UHC service coverage index from 54 to 65 by 2030, and cutting the share of households facing catastrophic health spending from 42 per cent to 35 per cent. A few additions would make the AI-enabled tools commitment more than a sentence in a document.
Any tool used to determine eligibility or priority for the National Health Fund should be checked for bias before it goes live, the same way financial and clinical outcomes are already tracked. Fairness and explainability deserve their own line in the compact's indicators. What gets measured tends to be what gets built.
The government should also draw on research already happening inside the country. Bangladeshi researchers have working models for triage, maternal risk prediction, and privacy-preserving collaboration across institutions, at different stages of readiness. None of this needs to be imported or reinvented from scratch. The long-dormant National Health Protection Act should be revisited with explicit rules for checking algorithms for bias.
On fraud detection, the source is candid: "my own research on this, which mixes rule-based logic with pattern recognition, has so far been tested only on simulated financial data built for a research paper, not on real claims from any health system." It's a promising direction, not a finished tool. The more basic work of auditing eligibility data for bias should come first. For professionals working in healthcare administration and AI for Healthcare, the lesson is that pilot-tested tools exist and can be deployed now, while eligibility and claims systems need bias audits before they touch real patients.
Why this matters for healthcare professionals
For clinicians and health administrators, the compact's AI commitment creates a concrete opening: tools that have already passed pilot testing - like the dengue triage chatbot and maternal risk predictor - can be adopted without waiting for imported solutions. But the same window creates risk for anyone whose work touches eligibility determinations or claims processing. Those systems will need bias checks before going live, and professionals should ask whether the tools they're asked to use explain their reasoning in terms a doctor would recognise. The gap between a signed compact and a working system is where the real work happens.
Your membership also unlocks:
