Generative AI and large language models

Health programmes depend on accurate, accessible information. Large language models allow BAO Systems to build tools that analyze complex datasets, support regulatory review, and extend the reach of health communications to populations that manual processes cannot serve.

Drug safety: AI-assisted look-alike, sound-alike detection

Registering and regulating the introduction of new drug products into the market in LMICs can be an arduous task within the backdrop of manual systems for delineating existing and similar products, with increased error margin and risk of inadvertent use of the wrong medications. BAO Systems developed an AI prototype for a national drug regulatory programme, building on a strategic industry partnership. Using large language models, the solution analyzes product datasets to return the top three look-alike and sound-alike name matches for any submitted drug name, alongside image clustering that identifies visual similarities between product packaging.

HIV and family planning: GenAI-powered health chatbot

BAO Systems upgraded a scripted healthcare chatbot for a global health partner in Latin America, integrating generative AI capabilities so that target populations could ask open-ended questions about HIV and contraception rather than navigating rigid menu structures. The AI draws on a curated knowledge base aligned with strict medical guidelines, while recognising local slang and regional expressions to improve comprehension and accessibility among youth audiences.

Agentic AI and automated narrative reporting

Generating a situation report from routine programme data is often a multi-day process involving data exports, manual analysis, and iterative drafting. Agentic AI systems act as co-pilots within existing software, automating the synthesis steps so analysts can focus on interpretation and response.

RISE: automated epidemiological situation reports

In collaboration with global health partners, BAO Systems is advancing an agentic AI solution for the DHIS2 ecosystem to extract greater value from existing systems for proactive decision support. Operating as an intelligent data agent, the system embeds a human-in-the-loop co-pilot calibrated to execute automated trend analysis, surface data anomalies, and produce baseline predictive reports.

By leveraging these analytical capabilities to automatically generate narrative epidemiological reports from live data, drafting turnaround times can be reduced from days to minutes. The technology is planned to be released as a global good for the wider DHIS2 community.

Automated machine learning and predictive analytics

Routine health data contains signals that manual review consistently misses: gradual service delivery disruptions, statistically significant outliers in specific districts, emerging gaps between forecasted and actual programme performance. Automated machine learning surfaces those signals before they compound into crises.

COVID-19: forecasting essential health service delivery

As an awardee of the FCDO COVIDaction Data Analytics and Use Challenge, BAO Systems developed a proof of concept using AWS Forecast’s autoML process to predict the expected volume of essential health services delivered during the COVID-19 pandemic.

Drawing on publicly available PEPFAR HIV/AIDS facility-level data, the deep learning-based forecasting approach outperformed standard baseline models, giving health program managers more accurate estimates of service delivery gaps attributable to pandemic disruption.