For a community health worker, the value of an AI tool depends on the job it helps them do. An Innovations in Healthcare study and webinar examined that question through the work of Audere, Simprints and ThinkMD in low- and middle-income countries.
Three different uses
The examples covered distinct tasks. ThinkMD used machine learning to support clinical logic in Nigeria. Simprints applied facial biometrics to patient identification in Uganda. Audere used computer vision to help interpret rapid malaria diagnostic tests in Rwanda. Each addressed a specific part of a care pathway rather than offering a single solution to every problem.
Start with the setting
Eliab Mwiseneza of Jhpiego Rwanda emphasised the importance of local context. A tool has to fit the people using it, the community being served and the practical constraints of the service. Training and partnerships with local organisations were recurring themes in the discussion.
Keep people responsible for decisions
Participants distinguished practical supervised machine-learning applications from wider excitement about generative AI. The discussion also referred to the WHO principle that AI should support human decision-making. Data governance, appropriate safeguards and clear responsibilities were presented as essential parts of implementation.
Build for continued use
ThinkMD's Dr Barry Finette stressed the need for accurate, clean datasets. Sarah Grieves of Simprints argued for beginning with a defined problem, while Audere's Natalie Maricich highlighted local partnerships and capacity building. Together, those observations point to a demanding test: whether a tool remains useful, supported and workable after its initial introduction.
