evidoria

← Back to browse

Good practice Imported

Ulangizi AI — Malawi's Government-Data-Backed Farming Advisory Chatbot

Malawi · Lilongwe · See the Malawi profile · See the Lilongwe profile

Evidence: Descriptive / self-reported Top 83% 40/100 · Ask Evidence Copilot about this practice

Opportunity International's Ulangizi chatbot merges generative AI with Malawi's Ministry of Agriculture guidance, giving farmers Chichewa voice, text and photo advice via WhatsApp. Thousands use it, but yield impact is anecdotal and accuracy is uneven.

2,800
Farmers given preloaded phones (2024 Malawi pilot)
~1,800
Combined active users (Malawi, Kenya, Ghana pilots)
17,000+
Messages generated
Ulangizi AI — Malawi's Government-Data-Backed Farming Advisory Chatbot

Details

Maturity
Pilot
Promoter
Opportunity International (with Malawi's Ministry of Agriculture)
Period
2024–2025
Keywords
agriculture, social protection, digital government

Context

Malawi's roughly 80% smallholder-farmer population is served by a small corps of government-employed farm support agents (FSAs) who cannot reach every household quickly, especially during shocks like Cyclone Freddy. Opportunity International, a UK-based nonprofit, built Ulangizi ('advisor' in Chichewa) to close that gap, combining OpenAI's GPT models with the Malawian Ministry of Agriculture's own extension guidance stored in an Azure database.

Activities

The chatbot, accessed over WhatsApp, uses Azure AI speech-to-text and machine translation to accept voice notes and photos of crops or livestock in Chichewa and respond with tailored advice, diagnosing pest damage or recommending planting adjustments.

Results

GiveDirectly, a cash-transfer charity, distributed phones preloaded with the app to 2,800 Malawian farmers in a 2024 pilot; parallel pilots in Kenya and Ghana brought the combined reach to roughly 1,800 active users generating over 17,000 messages.

Conclusions

The evidence remains early-stage: no controlled study of crop yields has been published, the bot has previously misread flagged terms (mistaking a plant name for witchcraft), and Ministry of Agriculture officials note it sometimes gives advice the ministry does not endorse -- underscoring that its reliability depends entirely on how current and curated its underlying government data stays.

Implementation
Indicative cost
Low (< €50k) — NGO/charity-funded pilot (Opportunity International, GiveDirectly); no aggregate cost published.
Time to results
Short (< 1 year) — 2024 Malawi pilot with parallel Kenya/Ghana pilots; ongoing as of review.
Staffing & skills
Opportunity International, farm support agents (FSAs), Ministry of Agriculture

Conditions for success

  • Combining a general-purpose LLM with government extension guidance stored in a curated database
  • WhatsApp delivery and voice/photo input to overcome literacy and connectivity barriers

Common failure modes

  • No controlled study of crop yields published
  • The bot has previously misread flagged terms (mistaking a plant name for witchcraft)
  • Ministry of Agriculture officials note it sometimes gives advice the ministry does not endorse

Where it fits

Governance type
NGO-led, government-data-backed pilot
Scale
pilot across Malawi, Kenya, Ghana
Income level
low-income

Commonly funded by

Philanthropic / foundation funding

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful