What challenges did Farmer.Chat face before AI implementation?

Smallholder farmers and extension agents often lack timely, localized, and personalized agricultural information. Traditional extension services are insufficient and unscalable, especially in remote and low-literacy settings.

What was the impact of the AI solution?

Deployed in four countries. Engaged over 15,000 farmers. Answered more than 300,000 queries. Reported improved productivity, trust in AI advice, cost savings, and higher engagement.

Rank#1,515of 3,449

Source

MicrosoftProductionEvidence: Medium65/100

Farmer.Chat: Scaling AI-Powered Agricultural Services for Smallholder Farmers

Use case typeAgronomic advisoryUpdated Sep 3, 2024

Farmer. Chat is a generative AI-powered chatbot for smallholder farmers that delivers localized agronomic advice through retrieval-augmented generation, multilingual support, and multimodal inputs across WhatsApp, Telegram, and mobile apps. The platform integrates structured and unstructured agricultural knowledge, external services such as weather and disease diagnostics, and user feedback loops to improve response quality and trust. It has been deployed across Kenya, India, Ethiopia, and Nigeria and has handled more than 300,000 queries for over 15,000 users.

Organization: Farmer.Chat
Industry: Agriculture
Location: Kenya
Published: September 2024

Reported outcomes

300,000 queries

queries answeredAdoption & scale

15,000 usersfarmers engaged+75%queries successfully answered+71%context precision9.1 secondsaverage response time+45%follow-up questions share

Strategic outcomes

Other strategic outcomeExpanded access for low-literacy farmersCustomer experience & trustBuilt trust through verified agricultural adviceSpeed & agilitySupported women farmers and gender-responsive advisoryCost efficiencyReduced dependence on extension agents and consultantsOther strategic outcomeAdded multimodal and multilingual agronomic support

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Farmers engaged: 15,000 users increase

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

Deployed in four countries, Farmer.Chat has engaged over 15,000 farmers and answered over 300,000 queries.

Normalized claim

Queries answered: 300,000 queries increase

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

Deployed in four countries, Farmer.Chat has engaged over 15,000 farmers and answered over 300,000 queries.

Normalized claim

Queries successfully answered: 75% increase

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

The platform successfully answers close to 75% of user questions.

Normalized claim

Context precision: 71% increase

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

In our evaluation of over 1,000 queries... we found that the average context precision in Farmer.Chat is 71%.

Normalized claim

Average response time: 9.1 seconds decrease

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

The average response time in Farmer.Chat is 9.05 seconds.

Normalized claim

Follow-up questions share: 45% increase

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

this addition had a significant impact, contributing to over 45% of total interactions during peak periods.

Normalized claim

Faithfulness high accuracy share: 80% increase

arXivSep 3, 2024Research reportInferred claimMedium evidence strength

Farmer.Chat maintains high faithfulness for nearly 80% of queries

Normalized claim

Relevance high accuracy share: 67% increase

arXivSep 3, 2024Research reportExplicit claimMedium evidence strength

Similarly, 67% of responses are highly relevant

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Farmer.Chat
Provider: Microsoft
Maturity: Production
Linked source: arXiv

It has been deployed across Kenya, India, Ethiopia, and Nigeria and has handled more than 300,000 queries for over 15,000 users

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 4

1Agronomic advisory
2Conversational assistants
3Multilingual communication

Smallholder farmers and extension agents often lack timely, localized, and personalized agricultural information.
Traditional extension services are insufficient and unscalable, especially in remote and low-literacy settings.

Built a generative AI chatbot using Azure OpenAI GPT-4 and GPT-3.5, retrieval-augmented generation, multilingual translation, speech recognition, and messaging app integrations.
Ingested expert-vetted agricultural documents, videos, and web sources into a knowledge base and added feedback loops for iterative improvement.
Supported text, voice, and image-based interactions in multiple local languages.

Technologies

Azure OpenAI GPT-4 GPT-3.5 Retrieval-Augmented Generation Google Translate APIs Whisper Telegram WhatsApp Tomorrow.io Plantix

Deployed in four countries.
Engaged over 15,000 farmers.
Answered more than 300,000 queries.
Reported improved productivity, trust in AI advice, cost savings, and higher engagement.

Architecture

Farmer.Chat uses a knowledge base builder to ingest structured and unstructured agricultural documents, videos, and web sources into embeddings for retrieval. A multi-step LLM pipeline performs intent detection, query rephrasing and decomposition, text retrieval and ranking, and response generation. The system supports multilingual processing via Google Translate APIs, speech recognition via Whisper, and multimodal interactions through WhatsApp, Telegram, and a mobile app. A continuous feedback loop captures conversation logs and user ratings to refine outputs and retrain models.

Sources & evidence1

Evidence: Medium65/100Evidence strength