Smallholder farms span less than two hectares or about five acres in area. Combined, these individual farms represent 80% of farmland in Asia and sub-Saharan Africa and provide 80% of the food supply in those regions. The 600 million smallholder families worldwide make up a significant portion of the 1.1 billion people who live in poverty.

Smallholders consistently face several challenges in their ability to earn a living income. They deal with low-quality seeds and animal feed, lack access to financing, and have limited market access. Recently, producers have faced difficulty in implementing initiatives such as updated farming practices and new data collection efforts required by regulations. These initiatives aim to ensure long-term sustainability by complying with regulations, such as the European Union Deforestation-Free Regulation (EUDR) and the Corporate Sustainability Due Diligence Directive (CSDDD). Smallholders also face a digital divide due to inadequate digital infrastructure, services, and training, exacerbating their challenges and precluding them from equal participation in the digital economy.

To support smallholders and guide them toward economic growth, several digital intervention focus areas have been identified. These include access to advisory services, land registration and titling, risk management, and supply chain transparency. Multiple digital technologies are widely used in solutions that provide financial value to farmers and connect them with other value chain actors. Some examples of digital technologies are farm geolocation, remote sensing, satellite data, private blockchains for traceability, and mobile money. To continue closing the digital divide for farmers, we must ensure the paradigm-shifting digital innovation of our time — AI — works for them.

The AI Ecosystem

Two types of solutions dominate and determine the AI ecosystem most of us experience daily: 1) personal productivity and 2) enterprise transformation. Personal productivity solutions — such as those from OpenAI and Google — enhance our private and professional lives with embedded assistance within content creation, planning, and transaction applications. Large enterprises not only create applications that leverage commercially available AI software but also craft bespoke AI models for their digital transformation. These models represent some of their most closely guarded intellectual property, as they provide a competitive edge. Examples span all industries: company valuation and risk assessment in finance, damage detection and route optimization in logistics, and improved diagnosis and access in healthcare.

In agriculture, smallholder farmers do not typically employ the two types of AI solutions discussed above, but other members of their value chains or market systems do. Food and beverage companies optimize their market research and product formulation. Retailers improve inventory and space management while attracting consumers with personalization. Biochemical companies use computer vision to monitor how the germination of different seeds is affected by environmental conditions. Insurers automate risk assessment and reduce human error.

Agriculture market players understand the importance of smallholder farmers in addressing supply volatility and growing companies’ customer bases. For example, Bayer has a target of reaching 53 million smallholders by 2030. However, agriculture ecosystem participants highlight a recurring challenge when engaging smallholders with AI solutions: data availability and fragmentation. Large datasets are a required component of any AI solution, whether its goal is to provide production advisory services, identify or mitigate risks for loan applicants, or identification of candidates for climate-smart agriculture investment. Many large agriculture technology vendors offer AI-powered software platforms that deliver production recommendations or control precision spraying systems. However, these platforms are supported by data from thousands of Internet of Things (IoT) sensors on farming equipment used at larger farms in developed countries. Meanwhile, data on smallholder farming practices is either not collected or exists only in paper form, making it unusable by AI platforms.

Smallholder-focused AI Solutions

Digital solutions aimed at smallholders fall into three main categories, aligning with traditional development intervention areas: advisory, access to finance and access to markets. Advisory platforms provide value chain-specific, agronomist-created farming training, as well as connection to experts for questions. Access to finance solutions enables improved access to capital at lower costs and business plan creation. Access to market applications offers updated commodity price information to improve bargaining power and expand the buyer pool.

Advisory. Digital interventions in advisory services for smallholders face three obstacles: 1) lower levels of digital literacy, 2) local language barriers, and 3) lack of domain knowledge. First, a farmer would need additional training on digital tools in general and specific applications. Second, they would expect to use technology in their local language. Third, they might not necessarily know which of the available training resources is the right one for their farm. Digital Green’s Farmer.chat solves these inherent challenges, while delivering immediate value for the farmer and enhancing data availability. Farmer.chat is an AI-powered chatbot that leverages Digital Green’s extensive library of value chain-specific training videos to deliver tailored assistance to extension workers and smallholders. Users access the bot directly through WhatsApp or Telegram via voice or text, in a local language, to ask questions on best practices or problems faced. The tool then uses multiple natural language processing (NLP) and generative AI (GenAI) models to interpret the question and provide the correct answer via video, voice, and text. In deploying the platform, Digital Green has not only addressed the challenge of localized data availability through its own data collection and content creation but is also expanding existing regional datasets of farmer problems based on their questions, with their informed consent.

Access to Finance. A fitting example of a multi-stakeholder approach that benefits smallholders while meeting the needs of other market actors is Amini. Insufficient environmental data across Africa has impeded both producers, who seek to optimize their output and access regulated markets, and their potential commercial partners, such as financial institutions looking to evaluate farmer loan risk. Amini aims to solve this challenge by combining raw data from multiple sources, including publicly available datasets, remote sensing, and private client data, and processing it with AI models to create actionable data and insights. These insights are provided to smallholders via text message or WhatsApp and to companies engaging with farmers via custom dashboards or application programming interfaces (APIs). One example of how that actionable data benefits multiple stakeholders is Amini’s agreement with Aon and the African Development Bank (AfDB). Under the agreement, Amini’s AI-supercharged data will support Aon and AfDB in de-risking farmers and creating precision crop insurance products.

Access to Markets. To improve smallholders’ bargaining power and access to buyers, current pricing information is crucial. Gramhal's mobile platform, Bolbhav, collects real-time price information in multiple value chains from individual farmer sales receipts and makes the unified data available back to farmers. Smallholders get access to price data for multiple value chains and multiple markets either by paying a very small annual fee or by contributing sales data, thus becoming members of a data cooperative. While Gramhal initially used its farmer-facing mobile platform to collect sales data, it has since incorporated computer vision AI models to automatically import the data from photos of receipts, reducing the barrier to entry for farmers.

Beyond the Present

To continue augmenting the three development categories above with the next wave of smallholder-focused AI solutions, listed below are potential focus areas.

Farmer data monetization. As discussed previously, the data generated by the 600 million smallholder farmers, combined with historical data they have accumulated, represents significant value for AI ecosystem members working in agriculture. More importantly, it could and should provide an additional revenue stream for the farmers themselves. A crucial requirement is that data use is not extractive and that farmers continue to own their data in perpetuity, licensing its use as they see fit. Non-profit organizations can play a unique role in boosting this opportunity for farmers by 1) leveraging the local infrastructure and social capital they have helped create and 2) adding data infrastructure and training to their programming. We envision a smallholder-owned, localized, agriculture-focused AI data marketplace, in which cooperatives play a core organizational and capacitation role. Key building blocks for data ownership and transactions by farmers and cooperatives, such as digital identity not tied to a commercial provider, already exist — for example, Bluenumber.

Everything via chatbot. Leveraging a messaging platform that farmers already use, such as WhatsApp, as the interface to provide services relevant to them removes multiple barriers to scale. These obstacles include app-specific training, regional language support, and the costs associated with localized platform customization. A potential growth area for this approach is access to finance. While AI-powered chatbots might not be ideal for loan applications — where very high accuracy is vital — other areas of access to finance programming might be better suited for the technology. Examples include delivering a financial and business skills curriculum and providing localized, expert-derived answers to money management questions.

Capacity creation and youth engagement. In addition to data availability, the importance of data quality in AI is well documented. Vital requirements for high-quality, representative data are diverse samples and culturally sensitive data labeling, performed by an inclusive set of annotators. This market need offers an opportunity for additional skill-building and a higher potential income for smallholders, especially for the younger generation. Data annotation could provide an initial entry into the AI ecosystem and its concepts for youth, leading to further skill advancement and higher-value roles in data science. Initiatives pursuing this approach are already in place, such as the work of Adanian Labs in Africa.

The focus areas above share an emphasis on smallholder data availability and quality, which are also requirements of the broader AI ecosystem to fully engage with farmers. We will conclude with a concrete example of AI bringing smallholder data into the digital economy while empowering farmers to benefit from their data.

Case Study: From Paper Records to Digital Value

The Deep Social Entrepreneur Women Cooperative in Chitwan, Southern Nepal, sits in a white two-story house surrounded by fertile fields growing corn (maize), rice, and vegetables. The cooperative provides multiple services to its smallholder members: agricultural training and input supply, loans, social capital, and collective bargaining. In the second-floor office, the cooperative maintains well-organized records, which represent valuable data that could assist members with advisory services, and loan repayment histories for formal financing or pricing strategies. However, three formidable obstacles stand in the way of farmers leveraging their own data: 1) the records are paper based, 2) they are in Nepali, typed and handwritten in Devanagari script, and 3) they cannot be analyzed easily or included in digital market systems. This is a prime example of the digital divide that impedes smallholders from benefiting from the digital economy. What if cooperative staff could just take a photo of a paper record with their phone, and AI software would recognize the Devanagari text on the photo, translate it into English, and structure it into an appropriate digital file?

Global development organization Heifer International, in collaboration with FruitPunch AI, is testing a new method of digitizing and valorizing farmer records using Optical Character Recognition (OCR) AI models. Mobilizing the global AI for Good community of volunteers, Heifer is enabling smallholder farmers to benefit from their own data through the AI for Nepali Farmers Challenge. In two ten-week sprints, several teams of volunteers skilled in AI, data science, and coding are making an impact on smallholders while also improving their own skills.

The success of the AI for Nepali Farmers Challenge carries multiple benefits for smallholders, the AI ecosystem, solution providers, and development organizations like Heifer International. Farmers and agricultural cooperatives can use digitized historical data to access capital at a lower rate by demonstrating loan repayment histories and easily importing the data into a future enterprise resource planning (ERP) system. Importantly, the data generated by millions of smallholders represents an additional revenue source for farmers and significant value for the AI ecosystem. Because any tooling created in the Challenge, though not the data used, will be available as open source, solution developers can incorporate it into their products and deliver additional value to farmers and other users. For non-profits like Heifer, the Challenge represents a tactical effort with low financial and personnel time investment that demonstrates the value of technology for our stakeholders, paving the way for including it in larger-scope programmatic efforts.

Smallholder farmers face multiple challenges on the road to achieving a living income, including the inability to benefit from their valuable data. This data also represents an opportunity for the ecosystem of AI solutions used by their market partners. We must ensure that farmers are compensated for the value of their data. AI can play a significant role in closing the digital divide and supporting economic growth for smallholders. To deliver on that promise, we must empower them to become equal participants in the AI ecosystem by designing and deploying culturally sensitive, localized AI solutions that focus on their needs and engage them from the start. 

About the Author:

Vesselin (Vess) Natchev is the Global Technical Lead at Heifer Labs, a digital technology unit within Heifer International that collaborates with country programs to co-create technology interventions that deliver farmer value and accelerate Heifer’s organizational impact. His active area of interest is in deploying AI-powered solutions that directly benefit smallholder farmers and bridge the digital divide. Before joining Heifer in 2022, Vess spent 20+ years at IBM designing, creating, and implementing digital solutions worldwide.