Good morning - Nate here. Welcome to the first AI & Global Health Brief! I hope you find it useful. I know I learned a lot making it.
AI Tools Are Amazing, but We Still Need Roads, Internet, and Trained Health Workers
Is AI the answer to healthcare quality and access problems in low-income countries?
The World Economic Forum identified seven breakthrough applications already showing results, including AI tools that read brain scans twice as accurately as specialists, detect minute bone fractures, improve ambulance triage, detect diseases before symptoms appear, improve clinical decisions, evaluate traditional medicines, and automate tedious administrative tasks. Research from the past two weeks shows us real potential, but also reasons to question the hype.
A systematic review of 109 studies found that AI applications like predictive analytics, telemedicine platforms, and automated diagnostic tools show significant promise for improving healthcare accessibility and quality in underserved rural areas. However, successful deployment remains limited by unreliable internet, inadequate infrastructure, and insufficient training for local healthcare workers.
Precision nutrition uses AI to analyze genetics, microbiome, and metabolites for tailored dietary recommendations, showing promising results with microbiota-directed foods for malnourished Bangladeshi children. But implementation faces barriers including lack of digital infrastructure, limited laboratory capacity, and the need for population-specific training datasets rather than Western-based models.
Researchers developed an AI-powered healthcare model for rural Bangladesh using generative AI for disease tracking and pandemic preparedness, though poor infrastructure remains the major barrier.
A survey of Pakistani healthcare providers found that while most had positive attitudes toward AI and 73% were familiar with AI technologies in healthcare, only 34% felt confident operating AI systems, revealing a critical training gap that requires hands-on education and better infrastructure before these tools can improve care delivery.
Kenyan clinics achieved 47% adoption of an AI clinical decision support system over 8 months (up from 4%), with clinicians reporting improved diagnosis and treatment planning. But success required addressing unreliable internet, misalignment with local treatment guidelines, and clinicians' tendency to skip using it for "simple" cases.
Researchers surveying 85 healthcare professionals in Nepal and Ghana found that while AI is already being deployed, 85% identified lack of ethical oversight as critical, 72% emphasized need for localized governance, and 65% cited low AI literacy as major barriers alongside inadequate infrastructure (58%) and limited funding (52%).
Furthermore, this article warns that healthcare organizations rushing to adopt AI without proper infrastructure face significant risks, proposing five foundational areas: enterprise architecture, IT governance, data standardization, knowledge management, and clinical decision support systems.
The big story last week from the AI world offers a sobering parallel: an MIT study found that 95% of enterprise AI projects fail. We can argue about the definition of “failure,” but this highlights a crucial lesson: rushing to "do AI" will probably not yield transformative results. Instead, success requires two fundamental components. First, we must understand what specific AI tools can and cannot do. For example, generative AI is probabilistic, meaning its outputs contain inherent randomness and variability. What are the implications of this uncertainty for your intervention? Second, any health AI intervention exists within a specific country, community, and health system context. Will this intervention work and be sustainable given the local political and economic environment, infrastructure limitations, health workforce capacity, and community norms?
Rather than hoping AI will somehow transcend these realities, we still have to do the hard work of creating environments that enable AI tools to deliver real impact.
Tools & Resources
AI Laws Around the World: Database tracking AI laws and regulations across 70+ countries, showing dramatic disparities in national AI preparedness
That’s it for this issue! If you have any comments or would like to share information for a future issue, you can reach me at: [email protected].