FarmerChat: AI-Powered Agricultural Advisory at Scale

India's 150 million smallholder farmers are vital to rural livelihoods and national food security, yet face persistent information gaps due to climate volatility, emerging pests and diseases, fragmented markets, and limited access to extension services. Nearly 58% of farm households lack formal support, and productivity per unit of land lags behind other major producers. Digital Green's theory of change holds that timely, localised, and trusted agricultural advisory strengthens farmers' self-efficacy, improving decision-making, adoption of good practices, and ultimately productivity, resilience, and incomes.

FarmerChat is an AI-powered agricultural advisory platform designed for smallholder farmers and frontline extension workers. It uses Generative AI, including large language models, to provide real-time conversational guidance through text, voice, and images. Farmers can ask questions in local languages, share photos of crops or livestock, and receive actionable recommendations adapted to their context. Because agriculture is high-stakes, FarmerChat does not rely on generic AI outputs. Responses are grounded in curated, expert-verified agricultural knowledge using a retrieval-augmented generation architecture. Domain-specific enhancements—including fine-tuned speech-to-text for local accents and strengthened image diagnostics—ensure accuracy even with noisy inputs. Continuous reinforcement learning from human feedback, led by agricultural experts, improves relevance and safety over time.

Beyond individual queries, FarmerChat functions as a learning knowledge system. Aggregated interaction data surfaces emerging trends such as pest outbreaks or misinformation, informing iterative model updates and advisory refinement. Deployment follows a dual strategy: institutional partnerships integrate the platform into public extension systems through state agriculture departments, rural livelihoods missions, animal husbandry departments, FPOs, SHGs, and community organisations; while digital acquisition expands direct farmer reach through social platforms and telecom partnerships.

FarmerChat is active across India, Kenya, Nigeria, Ethiopia, and Brazil. In India alone, more than 0.4 million users have joined the platform. The goal is to reach 1 million users by April 2026 and 5.5 million by April 2028 while maintaining free access and reducing delivery costs below USD 1 per user. Independent evaluations demonstrate strong impact: a 60 Decibels survey found 74% of farmers rated guidance as highly relevant, 83% reported easier access to information, and 80% considered it comprehensive. IDinsight's evaluation showed 60% of users acted on advice, primarily in pest management, fertiliser use, and livestock care. Delivery costs fell from USD 35 per farmer in traditional extension to under USD 1, with adoption rates up to 10 times higher. The platform demonstrates that AI advisory is most effective when embedded in trusted local institutions and continuously expert-validated, rather than deployed as a standalone product.
For additional context and detailed documentation of this use case, please refer to pages 19-20 in the attached Casebook.