What if, through generative AI, a farmer could ask a chatbot any question, and instantly get an answer tailored to their specific community and context? This is exactly the experience that Digital Green has built with OpenAI. Called Farmer.Chat, this product supports the essential work of agricultural extension programs in countries including India and Kenya.
For farmers navigating a changing climate, agricultural extension (also called agricultural advisory services) is critical. Extension agents teach farmers best practices for growing their crops, help them connect with local suppliers, and provide market and pricing information. But especially in rural, remote communities, giving every farmer the support they need is a huge challenge. Even India’s network of over 400,000 agents has an agent-to-farmer ratio of only 1:650.
Over the past 15 years, Digital Green has been dedicated to solving this problem through digital agricultural extension. Working with the Ministries of Agriculture in India, Ethiopia, and Kenya, Digital Green has created nearly 8,000 farmer-to-farmer training videos in more than 50 languages, enabling extension agents to provide timely, locally-relevant information to more farmers. These videos have increased farmer income by an average of 24 percent.
Digital Green saw a chance to make an even bigger difference in farmers’ lives by harnessing generative AI. “With videos, we’ve been able to replicate success from farm to farm within a region, enabling farmers to learn from one another—while also growing a database of locally-derived agricultural best practices,” said Rikin Gandhi, CEO of Digital Green. “With OpenAI, we saw an opportunity to leverage this database to help farmers learn from one another at a previously unimaginable scale.”
With this vision in mind, Digital Green began developing Farmer.Chat on top of OpenAI’s technology suite. The first Farmer.Chat pilot was built on GPT-4, which significantly reduced hallucinations relative to previous models. Farmer.Chat uses Retrieval Augmented Generation (RAG) to integrate Digital Green’s vast library of agricultural information with those of government partners, including training video transcripts, annotated call center logs, and crop research factsheets. India’s Ministry of Agriculture validates all documents in the knowledge base to ensure accuracy and reliability.
Additional precautions were built into the initial rollout. “To safeguard against the risk of the chatbot advising farmers in error, we carefully curated the chatbot’s knowledge base and deployed the chatbot as an assistant to extension agents, rather than deploying the chatbot directly to farmers. This enables another stage of human review,” Gandhi shared.
Building on Farmer.Chat’s success, Digital Green recently also launched a Farmer.Chat GPT for ChatGPT, which adds several features that make the original chatbot even more powerful. With a new option for multimodal input, extension agents can take a photo of a farmer’s crop and get a more accurate, timely diagnosis of any issues. The chatbot also delivers real-time weather and market information.
Early estimates suggest that by leveraging generative AI, Farmer.Chat can bring down the cost of traditional extension services 100x, from $35 per farmer to $0.35 cents per farmer. One key driver of this productivity increase is that Farmer.Chat is accessible across a wide range of languages, including Hindi, Swahili, and regional languages, using an agile tech stack that integrates with local language translation datasets and services in each country.
Extension agents are finding that Farmer.Chat doubles as a self-teaching tool to grow their own expertise. “The chatbot is very informative and user-friendly. It enhances my knowledge and builds my confidence by helping me provide real-time solutions to support more farmers’ concerns in a single day than I could before,” said Raju Kumar, an extension worker with the Government of Bihar’s Rural Livelihoods Promotion Society.
As of January 1, 2024, over 4,500 extension agents are using Farmer.Chat across both Kenya and India. Digital Green is continuing to refine Farmer.Chat and expanding rollout in India to twelve states.
Given the promising results of Farmer.Chat, Digital Green is exploring how enhancements to the product can make it even better. To refine the chatbot’s ability to handle agricultural-specific nuances, and enable questions and answers in local dialects without needing to translate to English, they’re testing the impact of fine-tuning an “Agri-LLM.” This model would be trained on data contributed through a data trust, to ensure farmers maintain oversight of the management and use of data most relevant to them. Digital Green is also starting to use the Assistants API to bring the functionality of Farmer.Chat to interfaces already used by extension workers, such as WhatsApp and Telegram.
From the seed of an idea in 2023, Digital Green’s experience and vision continues to grow at a rapid pace. “As we have built iterations of Farmer.Chat using GPT-4, GPTs in ChatGPT, and now the Assistants API, we have gotten incrementally better at enabling multimodal prompts and context-aware guidance to catalyze the impact of agricultural extension workers,” Gandhi said. “Farmer.Chat can become a truly versatile agricultural companion that leverages farmer expertise to enable extension agents and farmers to build a thriving agricultural ecosystem."