Artificial intelligence in support of smallholder farmers in Africa and the Digital Green Experience in Ethiopia

In practical terms, AI enables farmers and extension agents to access more personalized and continuous advisory services – faster than ever before. Farmers can ask questions at any time, in their local language, through text or voice, and even share images of crops or livestock to receive more accurate support. They can get immediate, context specific guidance on key issues such as pest management, crop diseases, agricultural inputs, weather conditions, and increasingly unpredictable climate risks.

The reliability and responsibility of artificial intelligence are grounded in local realities. To ensure this, the tool is developed and deployed in local languages and are continuously improved to reflect the specific agricultural, cultural, and environmental contexts. The models are trained and validated using locally relevant content, including recommended agricultural practices, guidance from national research institutions, and data tailored to local conditions. Where available, existing decision-support tools developed by agricultural research institutes are also integrated to provide location-specific, personalized advice.

In addition, a “human-centered” approach is central to ensuring responsible and high-quality outputs. Agronomists and crop and livestock specialists regularly review the AI-generated responses based on thousands of questions asked by farmers. Their feedback is used to correct errors, refine recommendations, and continuously strengthen the system over time.

One of the key challenge relates to data. The availability of local, reliable, and usable datasets remains limited, and data are not always structured in formats that can be directly integrated into AI systems. Agricultural market and weather information, for example, may exist but often lacks clear indicators or regular, real-time updates at the local level. Digital Green works closely with governments, research institutes, and other stakeholders to enrich these databases, but this remains a gradual and complex process.

Another major challenge concerns linguistic diversity and infrastructure constraints. In many countries, a large share of farmers speak only local languages that are underrepresented in existing language models, requiring prioritization choices that often focus on dominant national languages. These challenges are compounded by limited access to technology, particularly smartphone availability and the quality of connectivity in rural areas.

Finally, for artificial intelligence to have a lasting impact on agricultural systems, it must rely on strong local institutions, especially well-established and adaptable agricultural extension services. AI can only be effective if it complements and strengthens these systems by being embedded in structures aligned with national and local realities. Public institutions, development agencies, and technical partners therefore play a central role in ensuring ownership, governance, and the long-term sustainability of these solutions.

More details on Digital Green's project