Leveraging the transformational
potential of AI to empower small-scale farmers in Ethiopia

In Ethiopia, agriculture is practiced by 97% of rural households. Agricultural productivity remains a significant challenge in many regions, particularly due to low yields and a limited adoption of innovative practices. Evidence shows that providing farmers with information about the benefits and proper use of new practices can effectively increase farmer knowledge, adoption, and, in some cases, yields (Beaman et al., 2021; BenYishay et al., 2020; Chandrasekhar et al., 2022; Manzoor et al., 2020; Duflo et al., 2023; Dzanku et al. 2023). Agriculture extension workers serve as a crucial link in this process. Many extension services, however, fail to reach farmers at scale and may prove ineffective.

Digital Green Foundation has spent over a decade working to improve agricultural extension through technology. Its flagship community video model where local farmers demonstrate successful practices on screen has been evaluated in four randomized trials in India and Ethiopia. These studies showed positive results on knowledge and adoption, though more limited impact on yields.

Building on this experience and feedback from field agents, Digital Green is now launching Farmer.Chat, an AI-enabled chatbot designed to support Ethiopia’s public extension agents, called development agents, in their daily work. The tool uses natural language processing (NLP) to deliver real-time, localized advice in Afan Oromo (other languages to follow), drawing on verified content from Digital Green’s video library and other expert sources. Development agents use the chatbot during farmer group meetings to answer specific questions and provide tailored recommendations on issues such as irrigation, pest control, seed selection, and climate-smart practices.

The app also includes voice and image functionalities and is designed to work with low connectivity to suit rural areas. It is constantly improved based on user feedback, ensuring advice stays relevant over time.

FID is financing two critical components to support the rollout and evaluation of Farmer.Chat:

• From April to September 2025, Digital Green will test how Farmer.Chat performs in the field, explore challenges to adoption, and refine the tool and its implementation strategy in three districts (woredas) of Oromia, covering 32 communities and training 100 DAs (men and women).
• After this first phase, an impact evaluation will be led by the University of Bordeaux and Addis Ababa University in 15 districts of Oromia, involving 240,000 farmers and over 850 DAs. The study will measure how Farmer.Chat affects farming practices, yields, and incomes, and assess its cost-effectiveness compared to traditional extension.

The project is implemented in close coordination with Ethiopia’s Ministry of Agriculture and local government partners. The impact evaluation is led by researchers from the Bordeaux School of Economics, in partnership with the Research Center for Inclusive Development in Africa (RIDA).

This phase will strengthen the institutional capacity of the Ministry of Agriculture and lay the foundation for more effective and inclusive agricultural policies. Key expected outcomes include:

• Improved staff capabilities in data management, analysis, and visualization
• Evidence-informed policy decisions on issues like input distribution, trade flows, and livestock services
• A sustainable model for using administrative and survey data to drive agricultural policy
• Lessons learned on how to institutionalize data-driven policymaking in low-capacity settings

In the long term, this innovation could help increase productivity, improve farmers’ incomes, and strengthen climate resilience.