One Million Plots. Just Getting Started.

That moment captures what iSDA is building towards. Not just better yields, but a farming life where every decision — what to plant, when to apply, how much to use — is grounded in real information about real soil, not a passed-down habit or a neighbour’s guess.

iSDA, supported by the Gates Foundation, has built Virtual Agronomist to do exactly that. In under two years, our Virtual Agronomist has been used by farmers on over 1.3 million plots across 8 African countries. In the first quarter of 2026 alone, we nearly doubled the total plots reached across all of the previous year, with individually tailored nutrient management plans delivered to more than 98% of them.

That pace did not happen by accident. And it did not happen because the technology worked perfectly from day one.

The Gap Virtual Agronomist Was Built to Close

Agricultural extension has long been proven to work. When a trained agronomist visits a farm in person and the farmer is able to act on their advice, yields can rise by 50 to 100%. The problem is cost: between $30 and $60 per farmer per year. Across sub-Saharan Africa, where extension ratios run as high as one agent to every 10,000 farmers, that cost makes anything close to full coverage impossible. Billions have been spent, but the yield gap has barely moved. 

Our premise was not to replace agronomists. It was to take what agronomists know and build a system that could deliver it to any farmer, at any location, at a cost that actually works at scale.

What Virtual Agronomist Is 

Virtual Agronomist is not a general-purpose AI that happens to know about farming. It uses a large language model as the conversational layer on top of deterministic agronomic models, which means the underlying recommendations are scientifically grounded and consistent rather than generated afresh each time. Every recommendation is calibrated to a farmer’s specific plot location, soil nutrient status, crop type, and target yield. 

That precision starts with iSDAsoil, one of Africa’s most detailed soil datasets, mapped at 30-metre resolution across the continent. When a farmer registers a plot, the soil data is already there. The system knows the land before the farmer types a single question.

Virtual Agronomist is delivered via WhatsApp, with no app to download and no new platform to learn. Interestingly, that decision came from a single conversation.  Abigail Wairimu, one of our software engineers, had been talking through logistics with Makena, a young farmer and entrepreneur in Kenya who bought produce from other farmers and sold it on to local traders. Makena was, in many ways, a lead farmer before we even had a name for the role: someone other farmers already trusted and turned to. She told Abigail that 70% of her coordination with other farmers already happened over WhatsApp. From then, we cut out the custom app we had been building and moved to the channel farmers were already using. 

Across rural sub-Saharan Africa, smartphone ownership among smallholders remains well below 50% in most markets, so for farmers without smartphones, lead farmers bridge the gap. These are community farmers with better access to digital tools and networks, trained to carry VA’s recommendations to their neighbours. It is this combination of precise technology and human trust that makes both the reach and the cost achievable.

What We Learnt Along the Way

Our assumption at launch was that the main challenge would be technical: get the models right, optimise for low bandwidth, improve the interface. To our surprise, that assumption did not survive long.

“The technology is not the main concern,” says Bruce Kisitu, iSDA’s Director of Operations. “The issue is getting it in the hands of the users.” Network instability, plot registration complexity, translation friction, digital literacy: these were the real friction points.

Soon we realised that scaling fast requires engineering resilience we hadn’t fully anticipated. Matt Miller, our Director of Engineering, describes watching registrations and nutrient plans “increase by hundreds” in real time, in a way that quickly shifted the team’s focus from refining the product to keeping the infrastructure standing.

What has kept adoption growing, even where recommendations sometimes challenge years of habit, is something simpler than any feature update. As Bruce often puts it: farmers trust the change they can see from the road. A neighbouring field that is visibly doing better is the most persuasive thing iSDA has. All demos, in a sense, are done by the roadside.

For Jamie Collinson, the operational challenges are a reminder of what the plots are actually producing. Across cohorts, our farmers have seen an average 60% yield increase. In Uganda’s maize cohort specifically, that figure was 66% — 1.8 tonnes per hectare. 48% of VA users achieved at least 4 tonnes per hectare, compared with 10% using typical practice. 

“When I joined iSDA,” Jamie says, “our agronomists told me that an expert could generally help a farmer get 50 to 100% more yield if they were in the field with them. The problem is effective delivery at scale, and AI is demonstrating a real difference there.” 

Reflections From the Team

1.3 million plots is a milestone worth marking. But it is also a starting point. We’ve learnt what works and have come to understand where the friction lives, and it’s clear that the farmers still outside our reach outnumber the ones inside it by a long way. That’s exactly what drives what comes next.