The story we are often told about Artificial Intelligence (AI) is one of replacement, particularly in agriculture. The machines are coming for the tractor, the picker, and then eventually, the farmer. But the real potential of AI in agriculture isn’t replacement but what happens when technology serves the people already in the system, extending their sight rather than overriding their judgment. Successful innovation starts with local knowledge and widens it.
Agriculture, while obviously important to our economies, is one of the slowest to integrate new technology. I learned this lesson in 2017 when I joined an AgTech startup called Artemis Ag, a software company focused on bringing people, processes, and farm data into the cloud. Our initial premise was that software could increase a grower’s view of the farm and drastically improve their efficiency. Initially, user adoption was slow. We learned that we hadn’t fully accounted for the farmer’s experience. We didn’t consider the level of knowledge that comes with years of growing in their local environments, nor the impact that tiny changes to input costs could have on their planning. We especially didn’t account for physical design decisions that affect how and where technology can be used or the fact that farmers, broadly speaking, are more likely to be outside and less likely to be on a computer. And this was just in the U.S.
Globally, it’s a different set of problems. On the ground, technology uptake is uneven and intensely local. In rural areas, 21 percent of people in Africa use the internet, compared with 75 percent in the Americas. Access is only one hurdle. A synthesis of 30 studies found that smallholders run into several barriers at once: capital, credit, weak supply chains, and plain doubt about whether a new tool will ever pay off. The technology itself is rarely the bottleneck.
A food system rests on six interdependent foundations—natural resources, production, manufacturing and supply chain, logistics, a strong consumer base, and effective governance and policy. When one gives way, the others feel it. For example, last year Hurricane Melissa destroyed Jamaica’s agricultural base, destroying many of its farms, roads and logistics infrastructure. In the aftermath, the government went to work to figure out the best policy and long-term plan needed, but it raised questions. How do you restart food production and manufacturing? And how do you rebuild the supply chain and logistics while balancing some of the country’s more foundational needs, such as new, more resilient homes, a better electricity infrastructure, and new roads? These questions present an opportunity for AI and innovation to help connect the dots between the six foundations.
A great example can be found in Aquaconnect, a startup my company SecondMuse worked with in 2019, when they joined our Seafood Innovation Project in Indonesia. We supported them as they explored bringing to Southeast Asia what it had built for shrimp farmers back home in India. Most of the shrimp eaten in India are not caught at sea but raised in farm ponds, a livelihood that supports whole coastal communities. The farmers who run those ponds carry real expertise, much of it learned by hand and kept in notebooks. When pond conditions slip, yields fall and the runoff can harm nearby waters. The farmers were not short on knowledge or effort. They were short on sight: a way to see what was happening in the water before it cost them a harvest.
Aquaconnect set out to close that gap by building on what farmers already knew. The company pairs local, on-the-pond expertise with AI and satellite remote sensing, using geospatial data and deep learning models to read a pond’s condition and stage remotely, without a site visit. The technology widens what the farmer can see and act on. Aquaconnect has grown from roughly 3,000 farmers when we first met them in 2019 to more than 90,000 aquaculture farmers across six Indian states today.
It’s here that AI proves its worth, not by doing the work, but by showing the worker the whole field and helping to close the trust-and-risk gap. It offers a bird’s-eye view across foundations, connecting a decision in one part of the system to its consequences in another.
AI and prediction technology lets us ask sharper questions of messy data. This was the power of Artemis when it was acquired by Iunu. Iunu wrapped Artemis’s technology into a Machine vision product that helped growers make stronger, more insight-driven decisions.
So how can we use AI to build a better future in agriculture? AI is not a substitute for engaging trusted local institutions and local growers. They are the key to providing the right mix of meaningful data. Often, farms have outdated legacy software. Be willing to do the hard work of collecting and cleaning up the data before introducing AI. It becomes the difference between building something truly useful and sustainable versus not.
In addition, don’t forget the humans behind the work and the expertise they’ve cultivated over generations. This is the step that enables u to bring in insights that may be obvious to a 30-year veteran farmer who is irreplaceable when they retire.
This idea builds on years of groundwork. In 2019, The Rockefeller Foundation launched the Food System Vision Prize, inviting teams from around the world to propose transformational visions for the year 2050. These were visions that would enable local communities to describe, and ultimately attain, a regenerative, equitable, and nourishing food future. The Foundation distributed US$2 million in prizes to 10 winning teams, whose stories are featured in a series of documentary shorts. The Prize proved that communities everywhere, given the right framework, could imagine their own food futures with rigor and ambition.
We are working together on the next step in enabling ambitious ideas for future food systems. It is a platform SecondMuse is building, supported through a grant from The Rockefeller Foundation. It takes a proven food-system visioning methodology and puts it directly in the hands of communities, so they can build hope-filled visions of their own food future without needing outside facilitators or institutional backing. It is the thesis in working form: communities author their own visions; the structure helps them see the whole system rather than one slice; and it adapts to local context by design.
Aquaconnect works because it did not ask farmers to abandon what generations had taught them. It read their notebooks, added satellites and algorithms, and handed the result back to the same people who had been managing those ponds all along. Iunu did the same thing from a different angle, wrapping machine vision around what Artemis had already built, so growers could see patterns in their own fields they’d been too close to notice.
Do that across a whole food system and the picture changes. The grower sees not just the field but the weather bearing down on it and the market waiting past it. The founder fighting food waste sees whether the real problem is spoilage or a selling radius drawn too tight. The advisor sees which pond’s trouble is about to become the problem of the district. None of them stops relying on what they know. Each of them can finally see past it. That is the food system worth building with AI. Not one run by machines, but one where the people already inside it are no longer working in the dark.
Articles like the one you just read are made possible through the generosity of Food Tank members. Can we please count on you to be part of our growing movement? Become a member today by clicking here.
Photo courtesy of Kat Kelley, Unsplash
