What challenges did Digital Green face before AI implementation?

Farmers in India and Kenya struggle to access accurate advisory content due to language and digital divides. Existing AI systems are poorly adapted to local languages and cultural realities. Lack of digital infrastructure, low bandwidth, and limited device capability further hinder adoption. Available content is often not actionable, culturally grounded, or situation-specific.

What was the impact of the AI solution?

Enabled voice, video, and text interaction under low-bandwidth constraints. Blueprint for domain-adapted, inclusive generative AI.

Source

MicrosoftEvidence: Medium50/100

Digital Green transforms agricultural advisory with multimodal AI agent

Use case typeAgronomic advisoryUpdated Jun 13, 2026

Microsoft Research Africa and India, in collaboration with Digital Green, developed Project Gecko to improve the adoption and effectiveness of generative AI for smallholder farmers in India and Kenya. The project addresses significant barriers for farmers, such as language, cultural context, and lack of digital infrastructure. At its core is MMCTAgent, a multimodal AI agent built with Azure AI Foundry, using the VeLLM platform to generate multilingual, culturally relevant content and advice. The agent processes queries via text, voice, or video, and responds with actionable guidance through those same modalities—optimized for low bandwidth and minimal computing resources. The system integrates community-sourced agricultural videos and knowledge provided by Digital Green, allowing the AI to surface advice in the farmer's language, with content referencing the specific steps relevant to their problem. Initial field results confirm improved usability, trust, and farming outcomes over generic AI products. The solution is designed to be extensible to healthcare, education, and retail, reflecting its potential as a blueprint for inclusive, domain-specific, agentic AI applications.

Organization: Digital Green
Industry: Agriculture
Location: Kenya
Published: November 2025

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

Customer: Digital Green
Provider: Microsoft
Maturity: Unknown
Linked source: techafricanews.com

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 3

1Multimodal AI agent for agricultural advisory
2Culturally adapted AI content delivery in low-resource settings
3Voice, video, and text agricultural guidance

Developed a multimodal critical-thinking AI agent (MMCTAgent) on Azure AI Foundry.
Trained VeLLM models (Small/Local Language Models) for multilingual, domain-specific applications.
Integrated Digital Green's library of 10,000+ videos and knowledge assets.
Delivered step-by-step, multimedia responses tailored for low-resource environments.

Technologies

Azure AI Azure AI Foundry MMCTAgent LLM Copilot

Enabled voice, video, and text interaction under low-bandwidth constraints.
Blueprint for domain-adapted, inclusive generative AI.

Architecture

Project Gecko's MMCTAgent leverages Azure AI Foundry to combine speech, computer vision, and text understanding. The agent analyzes agricultural video content and farmer speech queries, retrieving relevant guidance from the Digital Green library. Responses are generated using domain-specific VeLLM models, and communicated back via the appropriate modality (voice, text, or video), all optimized for low bandwidth. Data and continuous user interactions improve models and expand language support.

Sources & evidence3

Evidence: Medium50/100Evidence strength