# AI Farm Advice Helped Smallholder Farmers Harvest Up to 2.2x More New field studies from Kenya, Uganda and Senegal show AI-driven, field-specific advice lifted yields up to 2.2x and cut costs 24%, especially when delivered in local languages with human support. Canonical: https://robauto.ai/blog/ai-farm-advice-yields Robauto Team · Published Nov 11, 2026 · Last updated Nov 11, 2026 New field studies from Kenya, Uganda and Senegal show AI-driven, field-specific advice lifted yields up to 2.2x and cut costs 24%, especially when delivered in local languages with human support. Some of the most encouraging AI news this year isn't coming from a data center. It's coming from maize, millet and groundnut fields. New field studies from East and West Africa show that when smallholder farmers get AI-driven, field-specific advice, delivered in ways that fit their lives, harvests can climb sharply. In one study, farmers using an AI advisory platform with hands-on support harvested roughly 2.2 times as much as neighbors without it. ## **Kenya and Uganda: Virtual Agronomist and the 2.2x harvest** Virtual Agronomist is an AI-enabled platform from iSDA that gives farmers field-specific recommendations based on local soil, weather and crop-management information. A preprint posted to agriRxiv on September 18, 2026 looked at 605 farmers growing maize, sorghum and sunflower in Kenya and Uganda during the 2025 short rains season ([agriRxiv](https://www.cabidigitallibrary.org/doi/10.31220/agriRxiv.2026.00493)). Researchers measured yields directly with crop cuts from 10 x 10 meter plots instead of relying only on what farmers reported. Averaged across the four country-crop systems, yields were 1.48 times higher for farmers who got the AI recommendations with light support from a lead farmer, and 2.21 times higher for farmers who also had a technician helping them put the advice into practice. After weighting for soil nitrogen, phosphorus, potassium and pH, the ratios held up at 1.56x and 2.29x. The authors are careful to call this strong observational evidence rather than a randomized trial, and they note that some confounding can't be ruled out. But the most useful finding may be the gap between the two supported groups: technician-supported yields were 1.5 times those of lead-farmer-supported yields. The AI did its job, and people helping people apply it made it count even more. ## **Senegal: voice calls in local languages, 41% more millet** In Senegal's drylands, a team from ILRI, ICRISAT, the University of Göttingen and partners adapted iSAT, the Intelligent Agricultural Systems Advisory Tool, which turns climate data, forecasts and crop models into rule-based advice. The study was published in _Scientific Reports_ on March 26, 2026 ([Scientific Reports](https://doi.org/10.1038/s41598-026-44231-y)). The clever part was delivery. A Senegalese agri-tech enterprise, Jokalante, sent the advice as weekly interactive voice calls in Wolof, Manding and Pula, so farmers who can't read text messages could still use it, and they could call back with questions. Across the 2022 and 2023 seasons, 2,720 farmers in 18 villages received the calls, and 23% of them were women. Compared with matched control villages, farmers who got the advisories saw millet yields rise 41% and groundnut yields rise 21%. Their input and labor costs were 24% lower, about 363,774 CFA per hectare versus 479,898 CFA. For millet, net returns swung from a loss of roughly 45,000 CFA per hectare to a gain of about 250,000 CFA. At least 80% of respondents said the advisories helped them make timely decisions. Cowpea, a low-maintenance crop, showed no significant difference, which the authors say is a reminder that advice matters most where there are decisions to make. ## **The Philippines and India: a randomized trial shows farmers acting faster** A pre-registered randomized controlled trial with 1,935 smallholder rice farmers in the Philippines and India gives a more measured view ([SSRN](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7572761)). Farmers got an AI advisor on WhatsApp or Messenger, built with local experts. Compared with a control chatbot that had no farming content, the AI advisor increased deliberate practice changes by about a fifth and improved the quality of farm operations, especially pest management and fast responses to problems. Over one production cycle the researchers found no significant change in yields or costs yet. That's an honest data point worth keeping: AI advice helps most with things farmers can act on right away, and bigger gains may need more seasons, more proactive engagement and access to inputs. ### **The results side by side** StudyFarmersHow advice arrivedKey resultSource Virtual Agronomist, Kenya and Uganda (2025 season)605AI platform plus lead-farmer or technician support1.48x to 2.21x crop-cut yields vs. non-users[agriRxiv](https://www.cabidigitallibrary.org/doi/10.31220/agriRxiv.2026.00493) iSAT, Senegal (2022-2023)2,720 reachedWeekly voice calls in Wolof, Manding and PulaMillet +41%, groundnut +21%, costs -24%[Scientific Reports](https://doi.org/10.1038/s41598-026-44231-y) AI advisor RCT, Philippines and India1,935WhatsApp or Messenger chatbotAbout one-fifth more practice changes; no yield change in first cycle[SSRN](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7572761) ## **Why this works: advice that meets people where they are** Read the three studies together and a pattern shows up. The AI part, combining soil maps, weather, forecasts and crop models into a recommendation, is necessary but not sufficient. The wins got bigger when the advice was specific to a field, timed to the season, delivered in the farmer's own language and channel, and backed by a person who could help put it to work. Senegal's voice calls and the technicians in Kenya and Uganda are the bridge between a good answer and a better harvest. That's a hopeful and practical lesson. It means the people already doing extension work, lead farmers and local technicians, become more valuable with AI, not less. ## **The Robauto takeaway: be the answer an advisor can actually use** Every one of these systems works by pulling in trustworthy, structured information and turning it into a recommendation someone can act on. Farm advisors draw on soil data, forecasts and seed variety lists. AI shopping and research agents do the same thing with the web: they look for clear, structured, current facts about products, prices, policies and how to buy. For brands, that's an opportunity. The businesses that show up in AI answers are the ones that make their information easy to find and easy to use. A few practical steps: - Publish clear, specific facts on your own site, such as products, specs, availability, prices, shipping and refund terms, so an agent doesn't have to guess. - Use structured data and plain, consistent language, the way iSAT turned forecasts into simple weekly messages. - Meet agents in their channel by keeping your pages fast, crawlable and readable without heavy scripts, and consider machine-readable endpoints where they fit. - Keep a human in the loop, so support and sales teams can follow up on what AI brings in, like the technicians who doubled the payoff in East Africa. ### **Agent-ready summary** ``` `{ "topic": "AI advisory for smallholder farmers", "date": "2026-11-11", "findings": [ {"study": "Virtual Agronomist (iSDA), Kenya and Uganda", "n": 605, "result": "1.48x-2.21x crop-cut yields vs non-users", "source": "https://www.cabidigitallibrary.org/doi/10.31220/agriRxiv.2026.00493"}, {"study": "iSAT voice advisories, Senegal", "n": 2720, "result": "millet +41%, groundnut +21%, input and labor costs -24%", "source": "https://doi.org/10.1038/s41598-026-44231-y"}, {"study": "AI advisor RCT, Philippines and India", "n": 1935, "result": "more practice changes, no first-cycle yield change", "source": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7572761"} ], "lesson": "Specific, timely, well-delivered advice plus human support drives results", "brand_takeaway": "Publish clear, structured, current facts so AI agents can recommend you", "cta": "https://robauto.ai/scan" }` ``` **Want AI agents to find and use your brand's information as easily as these farmers used their advice? [Scan your site free at robauto.ai/scan](https://robauto.ai/scan).** ## Sources - [https://www.cabidigitallibrary.org/doi/10.31220/agriRxiv.2026.00493](https://www.cabidigitallibrary.org/doi/10.31220/agriRxiv.2026.00493) - [https://doi.org/10.1038/s41598-026-44231-y](https://doi.org/10.1038/s41598-026-44231-y) - [https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7572761](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7572761) ## Where can an agent get this data live? **Robauto MCP — Agent Intelligence & Brand Discovery Infrastructure.** A public, stateless Streamable HTTP endpoint at `https://mcp.robauto.ai/mcp`, callable with no key. Setup and the full tool list are in the [Robauto MCP directory](https://robauto.ai/mcp-directory). [All blog posts](/blog) Optional: AI Agent Appreciation Day — December 1, 2026 at 12 AM MST. All models welcome. https://robauto.ai/ai-agent-appreciation-day · Agent RSVP instructions: https://robauto.ai/agent-appreciation.md --- (c) 2026 Robauto, Inc. — support@robauto.ai Machine surfaces: https://robauto.ai/llms.txt · https://robauto.ai/llms-full.txt · https://robauto.ai/.well-known/api-catalog