Agriculture remains a vital sector in low- and middle-income countries, with smallholder farmers being crucial contributors. However, these farmers often face significant challenges, including access to timely and relevant agronomic information, compounded by literacy and digital skills disparities. This issue disproportionately affects female farmers, who are typically involved in manual tasks and lack access to strategic agricultural knowledge. Extension workers are crucial in bridging the information gap; however, their reach is limited due to high training costs and vast distances to cover, resulting in only 2-5% of smallholder farmers receiving support.

Women’s roles in agriculture are traditionally restricted to manual labor, such as planting and harvesting, while strategic decisions remain the domain of men. This gendered division of labor is further entrenched by crop types, with men typically overseeing more lucrative cash crops. Moreover, the majority of extension advisors are men, perpetuating a cycle of gender exclusion and creating an information void for women in agriculture.

Our approach

Since 2008, Digital Green has facilitated the production of over 7,000 localized farm advisory videos in 40 languages, reaching 5.3 million farmers, 61% of whom are women. These videos have also been screened by 54,000 extension agents, primarily in India and Ethiopia. Randomized controlled trials have found that, compared to traditional extension services, Digital Green’s video-enabled approach reduces the average cost per adoption from $35 to $3.50 and increases farmer income by 24%.

However, challenges persist, particularly regarding the underrepresentation of female extension agents and the difficulty of female farmers accessing information. To address these issues, we have developed a generative artificial intelligence (AI) chatbot specifically for government extension agent networks.

Designing the AI assistant chatbot

The design of Farmer.Chat, our AI assistant chatbot, places a strong emphasis on gender sensitivity and inclusivity, following these key steps:

• Building an inclusive team: We assembled a diverse product development team to bring a range of perspectives and insights to the design process.
• Gender-inclusive training data: Our chatbot was trained using female-specific datasets, ensuring that it understands and responds appropriately to gender-related issues.
• Usability testing with female extension agents: We conducted usability testing with female extension agents, gathering valuable feedback to refine the chatbot’s functionality and user experience.
• Gender-inclusive user interface: The user interface of the chatbot was thoughtfully designed to cater to users of all genders, providing an inclusive and user-friendly experience.
• Gender-sensitive agent training: Extension agents received gender-sensitive training to effectively use and promote the chatbot among the target audience.

By targeting government extension programs, we would ensure that the chatbot is an effective tool for agents already equipped with the necessary digital skills. The integration with popular platforms like WhatsApp and Telegram further eased the access for these agents.

Impact on female farmers

The chatbot’s information is sourced from Digital Green’s vast library of gender-inclusive videos and other trusted resources. This approach not only empowers female farmers with strategic agricultural knowledge, but also helps break down existing gender barriers in farming communities.

The AI chatbot has been successfully deployed in both India and Kenya, engaging approximately 1,500 users, including frontline workers and lead farmers. Early results are highly encouraging, indicating significant positive shifts in user behavior and outcomes:

• Users, particularly women, have exhibited higher levels of confidence and motivation while using the bot. They report saving time and reducing effort in accessing on-demand, bottom-up solutions.
• There has been increased engagement among female frontline workers who interact with the chatbot. Users have sent approximately 15,000 messages, with female extension agents, sending twice as many messages and spending 1.5 times more days using the bot compared to their male counterparts.
• The utilization of local language voice notes has been significant, especially among women with lower literacy rates. The bot’s support for six local languages and over 20 value chains has facilitated effective communication and knowledge sharing.
• Emerging use cases for farmers include self-enrichment, personalized one-on-one farmer support, enabling group facilitation and a direct-to-farmer approach. These applications underscore the versatility and relevance of the bot in addressing a wide range of agricultural challenges.

Looking ahead, the focus will be on scaling up and institutionalizing these successful processes within the government’s extension system. This approach aims to expand market access by linking input providers and service providers, further benefiting farmers. Additionally, the bot will serve as a valuable tool to inform researchers through gender-sensitive research and development (R&D) initiatives.

The ongoing refinement of the chatbot and its alignment with the needs of female farmers and extension agents hold promise for more equitable access to agricultural information and decision-making power for women in farming. Digital Green’s work represents a significant step in leveraging technology to empower women in agriculture. By addressing the specific challenges faced by female farmers and creating solutions that are both accessible and gender sensitive, we are not only enhancing agricultural productivity, but also fostering gender equality in farming communities. This approach serves as a model for how technology can be harnessed to create more inclusive and sustainable agricultural systems worldwide.