By Dr Girmaw Abebe Tadesse, who leads Microsoft’s AI for Good Lab in Nairobi
For millions of African farmers, an unpredictable rainy season can decide whether a harvest succeeds. Plant too early, and seeds may fail before the rains arrive, while planting too late can leave crops short of time. These choices shape household income, school fees, food security and the livelihoods of whole farming communities. The FAO says about 309 million Africans faced hunger in 2025, so AI for farmers matters more as COP31 opens in Antalya, Türkiye.

How AI for Farmers Turns Forecasts into Advice
Microsoft Research scientists developed Probabilistic Bias Correction, a machine-learning method that improves forecasts two to six weeks ahead by correcting repeated errors in weather models. Research involving Kenya suggests this early warning could help farmers adjust planting dates and plan seed purchases. A forecast alone cannot tell a farmer what to do, however, because good advice must reflect soil, crop varieties, irrigation, costs and local experience. That makes AI for farmers a question of practical guidance for each region, not only of prediction.
Governments face a related task that AI for farmers can also support: knowing where crops struggle and who needs help. In Kenya, ADAPT-Kenya aims to build a shared, government-owned information system, led by the Ministry of Agriculture and Livestock Development with NASA Harvest, KADIC and Microsoft’s AI for Good Lab. During the 2026 long rains, partners supported drought-response assessments, though the author says this is not yet evidence of better yields. The longer-term goal is a set of tools that government teams can maintain, test and improve.
Local Languages and Lasting Systems
Advice only helps when farmers can understand it, so AI for farmers must speak local languages. LINGUA Africa, led by the Masakhane African Languages Hub with partners including the Gates Foundation and Google.org, supports African-language AI tools. Its awardees include MlimiVoice in Malawi, for Chichewa and Tumbuka speakers, and SemaShambani, which aims to offer spoken advice in Kiswahili and Dholuo. These approaches look promising, although the author notes their wider impact remains to be demonstrated across many communities.
The larger obstacle is institutional, because records are fragmented, connectivity is uneven and public agencies lack specialist staff. At COP31, governments and partners should treat agricultural data systems as essential climate adaptation infrastructure, not optional technology projects. Adaptation finance can fund crop surveys, shared databases, affordable connectivity and training for public analysts and extension workers. With that base in place, AI for farmers can help communities see risks earlier and respond with greater confidence.





