Africa AI Ledger · News
WHO Adds AI Queries To Polio Surveillance Across Africa
A new natural-language layer lets authorised teams question operational surveillance data from 47 countries, while epidemiological judgement and human verification remain essential.

Background generated in ChatGPT; Africa AI Ledger editorial template
The World Health Organization's African region has added a natural-language artificial-intelligence layer to operational polio dashboards, allowing authorised teams to ask questions of surveillance data in ordinary language and receive maps, charts, tables and written summaries.
WHO Africa says its Geographic Information Systems Centre receives weekly surveillance and operational data from 47 countries. The new interface is intended to help programme staff examine detection, vaccination coverage, campaign quality and field-surveillance activity without constructing every dashboard query manually.
The change is less dramatic than replacing epidemiologists with a model—and more useful because it does not try to do so. Staff can move more quickly from a question to a view of the relevant records, but public-health experts still need to verify the result, understand the context and decide what action is justified.
Polio surveillance combines many signals. Teams investigate suspected cases, examine environmental samples, monitor vaccination activity and track whether field visits are being completed. Delays or gaps can matter because the virus may circulate before paralysis appears. A faster way to interrogate operational data can help identify where follow-up is needed.
Natural-language access also broadens who can use complex systems. A programme manager may be able to ask which districts have declining coverage or where electronic-surveillance reporting is incomplete without relying on a specialist to build a new view. The interface should still show the underlying definition and time period so that a simple question does not produce a misleading answer.
Data quality remains the limiting factor. AI cannot recover a case that was never reported or correct a location recorded inaccurately unless the problem is detected and resolved. National programmes need continued investment in field teams, laboratories, connectivity and the electronic surveillance tools feeding the regional platform.
Public-health data are sensitive. WHO and participating countries should maintain strict access controls, audit queries, minimise personal information and define whether prompts or outputs are retained. If an external model or cloud service is involved, the organisations responsible should disclose where data are processed and which information leaves the protected environment.
The system also requires evaluation. Useful measures include the time taken to answer operational questions, the accuracy of generated summaries, the frequency of unsupported results and whether earlier identification leads to faster investigation. User satisfaction alone would not show whether surveillance improves.
Language and terminology deserve attention. Public-health staff across 47 countries work in different official and professional languages. The interface should be tested for the terms people actually use and should ask for clarification when a request is ambiguous. A confident but incorrect interpretation could direct attention away from the real gap.
WHO's approach shows a practical use of generative AI: reducing the distance between a human question and an existing evidence system. The value lies in making validated records easier to examine, not in presenting the model as a new source of epidemiological truth.
If governance and testing remain strong, the same pattern could support other disease programmes. The lesson is to keep the hierarchy clear. Field surveillance creates the evidence, epidemiologists interpret it, public-health institutions remain accountable, and AI helps them navigate the information more efficiently.
Source: WHO Regional Office for Africa.
Source & attribution
Source: afro.who.int