Using routine school data to support adolescent well-being in resource-limited settings

Artificial intelligence (AI) is increasingly promoted as a tool to strengthen health systems, yet its use in low- and middle-income countries (LMICs) remains largely concentrated in hospitals and specialized settings, overlooking schools that reach most adolescents and often represent the first point where changes in health and well-being become visible. Despite routinely collecting data on attendance, meals, and basic health indicators, schools rarely use this information for prevention or early action. This paper presents a conceptual and practice-oriented framework for using routine school data to support early identification of emerging adolescent health risks. We describe a workflow that standardizes fragmented and partially digitized records and applies lightweight, task-specific AI-supported tools to detect patterns and deviations that may warrant attention, including in settings with intermittent data collection and limited digital infrastructure. The approach prioritizes low complexity, minimal infrastructure requirements, and integration into existing school routines, with outputs serving as advisory signals to support contextual interpretation by educators and school health staff. We emphasize that these signals are based on proxy indicators and require careful human review to avoid misinterpretation. We argue that school-level, privacy-preserving AI can strengthen prevention by enabling more timely, context-sensitive, and community-led responses, while its effectiveness depends on data quality, local capacity, and appropriate governance to ensure ethical use, transparency, and trust across diverse LMIC contexts.