The new GSMA report AI for Impact at Scale: Case Studies from Innovators in Low- and Middle-Income Countries examines how AI is being deployed across development sectors, with agriculture emerging as one of the most advanced application areas.
Drawing on case studies from Africa, South Asia, and Southeast Asia, the report shows that AI in agriculture in LMICs has moved beyond experimentation, with solutions increasingly used to improve yields, detect pests and diseases earlier, and support climate-smart farming practices linked to food security and undernutrition reduction.
In agriculture, the report highlights four initiatives. Digital Green’s FarmerChat operates in India, Kenya, Ethiopia, and Nigeria, delivering AI-generated agricultural advice via mobile phones. Designed for low-literacy contexts, it supports voice, text, and image inputs and builds on Digital Green’s broader reach of more than 8 million farmers.
The value AI adds to FarmerChat
Image credit: GSMA Mobile for Development
Among active users, around 60% report adopting at least one new farming practice, while the cost of delivering advisory services has fallen by up to 100 times compared with traditional extension.
Lersha, based in Ethiopia, applies AI to link agronomic advice with climate risk management and access to finance, while Varaha, operating mainly in India and South Asia, uses AI and satellite data to automate carbon measurement and verification, enabling smallholders to access carbon markets. The report also highlights the World Bank’s Geo-AI and data-driven agriculture initiatives, which embed AI within national data systems for crop monitoring, climate risk analysis, and policy planning.
Key findings
Across these cases, the evidence suggests AI delivers the greatest impact when it is localised, multimodal (supporting voice, text, and image-based interaction), affordable, and embedded in existing agricultural institutions with humans in the loop. Where these conditions are absent, AI tends to expose deeper structural constraints, underscoring that agricultural transformation through AI remains a socio-technical challenge rather than a purely computational one.
Why it matters
The report shows that AI can work in smallholder agriculture, but only under specific conditions: impact depends less on model performance than on data quality, local grounding, and sustained human involvement. Where these elements are missing, AI does not compensate for structural weaknesses in agricultural systems, it makes them more visible.