Responsible AI solutions for global health: Local research to guide us through an evolving landscape

 
A mother holds her baby and a cellphone.
IDRC / Fredrik Lerneryd, Panos Pictures
Jacaranda Health’s AI-enabled digital health service PROMPTS  provides real-time, personalized support to new and expecting mothers in English and seven African languages in Kenya.

The world is experiencing a polycrisis — a perfect storm of climate catastrophes, conflicts, economic turmoil and disease outbreaks — that has disrupted and displaced populations, impacting their health and wellbeing on a broad scale. In response to these growing pressures on health systems, health actors (individuals, clinicians, public health practitioners and policymakers) are turning to artificial intelligence (AI) technologies for solutions.

AI’s rapidly evolving landscape presents both promise and peril. When developed responsibly, AI applications can inform decision-making, streamline logistics, improve treatment, reduce costs, enhance patient experiences and guide more equitable responses to outbreaks of communicable and non-communicable diseases. However, AI has many limitations and challenges, including underdeveloped regulatory and policy mechanisms, biased datasets, unaddressed privacy threats and an energy-intensive environmental footprint.

Health actors need locally relevant and rigorous research to generate the evidence to guide them in making the best choices for the use of AI technology for the health and wellbeing of populations across the globe.   

Research highlights

  • The development of responsible AI needs to rely on evidence-based solutions applied in real-world contexts by Global South researchers and innovators to address the root causes of health disparities in developing regions. 
  • Implementation research is an integrated approach that bridges the gap between the design of AI solutions and deployment in real-world settings. It is needed when focusing on the health needs of populations experiencing deep vulnerability and deprivation. 
  • When done responsibly, scaling offers extraordinary opportunities to address vulnerabilities and improve lives. The choice to scale AI solutions in different contexts should be carefully considered and informed by grounded, localized research and implemented to redress health inequities.  

An emergent research landscape

As a leading supporter of responsible AI for development, IDRC released a discussion paper, Strengthening Health Systems by Leveraging Responsible AI Solutions: An Emergent Research Landscape, in February 2025. The paper makes a case for ways to design, develop and deploy responsible AI solutions that are safe, inclusive, rights-based and sustainable.

It features a series of 12 case studies from across Africa, Asia, Latin America and the Caribbean, and the Middle East, funded by IDRC and the United Kingdom’s Foreign, Commonwealth and Development Office under the AI4GH initiative and broader AI4D partnership. From air-quality monitoring in South Africa to tracking polio cases in Ethiopia to providing rural women in Kenya with better health-care access, each case study presents real-world examples of how pressing global health needs are being addressed by locally designed and deployed responsible AI solutions.    

Case study: Improving maternal care in Africa

The problem: Women in Kenya and other sub-Saharan African countries experience numerous barriers to receiving quality health care due to cultural, socio-economic and demographic disadvantages.

Research for improved outcomes: Kenyan-based non-profit organization Jacaranda Health launched its AI-enabled digital health service PROMPTS to empower new and expecting mothers to seek and connect with the best care via mobile phone messaging. The platform uses a customized large language model, UlizaMama, to provide real-time, personalized support to mothers in Swahili, Sheng and English. The tool has since been expanded into five other African languages. 

The AI solution: The tool collects valuable information on how education, household composition, religion, the impact of climate events, the use of cooking fuel, technology ownership, sanitation, hygiene, age and geolocation create a vulnerability profile. 

Results: Since it started, almost 3 million mothers have been supported by PROMPTS. The platform is distributed to women in more than 1,000 public hospitals and health centres across 23 Kenyan counties. Beyond Kenya, Jacaranda Health is piloting PROMPTS in Ghana, Nigeria and Eswatini, with plans to scale across sub-Saharan Africa. It is also being piloted in Nepal.  

Guiding responsible AI and global health research

The discussion paper proposes implementation research as the lens through which potential AI rewards and risks can be addressed in a health system context. As an integrated approach designed to bridge the gap between what is expected to happen when promising AI solutions are designed and what actually happens when they are implemented in real-world settings, implementation research is especially needed when focusing on the health needs of populations experiencing the deepest forms of vulnerabilities and deprivation.    

The paper draws on an extensive literature review, an analysis of existing AI and global health projects, and key informant interviews and expert reviews. The landscape proposes three interconnected health system entry points: services, community and individual health. Each entry point is grounded in five cross-cutting prerequisites: i) regulation, policy and governance, ii) data quality and representation, iii) gender equality and inclusion, iv) ethics and sustainability, and v) Global South-led equitable partnerships.  

A graphic displaying emergent research landscape for responsible AI and global Health

Given the complexity of the challenges to health systems, there is an understandable desire for one-off, big solutions. However, solutions must be thoroughly evaluated and carefully positioned for scaling up. Evaluation is critical at every stage of design, development, deployment and assessment of AI solutions in global health.

In addition, the choice to scale AI solutions in different contexts should be carefully considered and informed by grounded research. Scaling offers extraordinary opportunities to address vulnerabilities and improve lives on a large scale when done responsibly. Ultimately, the emergent research landscape for AI and global health is designed to target impact with evidence that leads to stronger, more resilient health systems.

On the horizon: Rigorous and localized solutions

Leveraging responsible AI for better health systems requires the collective efforts of diverse stakeholders across disciplines, regions and sectors. As a potential contributor to addressing the root causes of health disparities in developing regions, AI use needs to rely on evidence-based solutions applied in real-world contexts through implementation research championed by Global South leaders. IDRC’s discussion paper is a starting point, providing a foundation for AI solutions that build stronger and more equitable health systems.