By Clara Cecil, Wageningen University & Research

During the recent RFSI Europe conference, I led a breakout session on artificial intelligence (AI) and agriculture. Lately, it seems that everywhere I turn there is a conversation about AI. What are the latest models? What new investment opportunities are being created? What are the implications for the future of work?

Honestly? I’m getting tired of the constant chatter and speculation. The United Nations recently asserted that we’ve entered a new era of global water bankruptcy (United Nations University, 2026). At the same time, generative AI is linked to high electricity demand and water consumption (MIT, 2025). AI is being touted as a solution for everything from healthcare to water management to climate resilience. While some of the potential advancements are genuinely exciting and potentially transformative, it is important to discern reality from hype. I keep engaging in discussions around AI because if voices from farming, ecology, and food systems aren’t at the table, we already know who will be.

I come to this topic from two directions. I am currently studying for a PhD in Regenerative Agriculture at Wageningen University in the Netherlands, through the ReGeNL program. Before that, I researched the water and land use impacts of data centers, the physical footprint of AI. I’ve also spent five years working at Proof, a technology company. I consider myself both a techno-optimist and a techno-skeptic. Approached through this dual lens, we had a vibrant discussion including the perspectives of farmers, investors, researchers, and tech entrepreneurs.

 

The Case for Optimism

The potential applications of AI in agriculture are genuinely exciting. We’re talking about cameras that can detect crop diseases before a farmer can see them with the naked eye, soil monitoring systems, weather and rainfall analysis, and robots that can augment physical labor in the field. AI can make sense of the vast quantities of data that modern farming generates – connecting soil to biodiversity to nutrient density to human health in ways that were previously impossible to track at scale.

At its best, AI could meaningfully reduce the burden on farmers. It could serve as a powerful decision-support tool, bringing together satellite data, weather patterns, and field observation to offer recommendations that a single advisor could simply not synthesize alone. One participant described the potential for open-source data infrastructure – something that could disseminate good practices across agro-ecosystems in local languages, without requiring a business model built on extracting value from farmers.

There’s even a vision emerging of a new data economy for farmers: one where the knowledge embedded in their soil records, yield data, and field observations is recognized as valuable and compensated accordingly, whether through carbon credits, supply chain premiums, or direct payment.

 

The Case for Skepticism

The risks are just as real, and the session surfaced them candidly.

The most immediate concern is dependency. In a similar way to pesticides, the more we have these tools, the more we rely on them, and the more we potentially lose the instincts and embodied knowledge that farming regeneratively requires. Remote sensing gives you a partial picture of a landscape. Self-reporting systems may under-report the things that farmers are reluctant to share. As several participants point out, AI doesn’t tell you when it doesn’t know something. It hallucinates. It reflects the biases of its training data.

Then there’s the question of data ownership. Some large agricultural companies were cited as using field data to improve their machines. Other large companies are navigating EU data law in ways that favor their interests. Farmers are sharing information and, in many cases, receiving nothing in return. One participant put it plainly: data sovereignty matters. If farmers don’t own their data, they don’t own their future.

The structural risk runs deeper still. A few large technology players now have significant interests in agriculture, and their incentive is not necessarily to support regenerative transitions. If the data these systems are trained on reflects decades of conventional, input-heavy farming, the recommendations that come out will likely reinforce those same patterns. The AI will optimize for the world it was trained on.

Finally, there’s the environmental paradox: we’re using energy-intensive AI systems to address the very environmental crises that energy intensity helped create. Data centers – the physical backbone of digital technologies – are taking up agricultural land in countries like the Netherlands where land is already scarce. In my home country of the United States, technology companies are purchasing farmland – already vulnerable due to aging farmer populations and challenging operating margins – to construct hyperscale data center facilities spanning multiple football fields in size (FarmProgress, 2026). The water and energy footprints of large AI systems are contributing to the utility pressures that farmers already face. How do we leverage these tools without compounding the damage?

 

Being Radically Human

I’ve been thinking a lot about a piece written by Finn Harries, who started Juntos Farm in Spain (When Machines Grow Faster Than Seeds). His suggestion for navigating AI anxiety, to paraphrase, is to be radically human. Use your hands. Smell the soil. Cultivate ancestral knowledge. Build community in a physical sense.

This framing landed differently for me after listening to a recent episode of The Great Simplification podcast with Nate Hagens, where he draws a striking parallel: AI-driven content is doing to our minds what industrial food has done to our bodies. Just as processed food offers cheap calories without real nutrition, the flood of AI-generated content gives us fast, persuasive, hyper-palatable information that often lacks substance. He calls it “ultra-processed information,” and the risk, he argues, is a kind of cognitive atrophy. The more we outsource our thinking to algorithms, the less we exercise the human judgment that makes us good at anything.

In agriculture, this parallel feels especially sharp. The knowledge that makes a good farmer isn’t just data – it’s pattern recognition built over years, embodied intuition, and the kind of wisdom that doesn’t live in a training data set. One participant noted that their father-in-law, a retiring farmer, only started recording things digitally a few years ago. Before that, it was notebooks. And before that, memory. That knowledge doesn’t automatically transfer into a dataset, and if we’re not careful, it disappears entirely.

Being radically human, then, doesn’t necessarily mean being anti-technology. It’s a commitment to being the kind of thinker and farmer that technology should serve – not replace.

 

Principles I’m Holding Onto

Coming out of the session, a few principles feel important to me as I think about what responsible AI in agriculture actually looks like:

Farmers as decision-makers, not data providers: The value of agricultural AI should flow back to farmers, not be extracted from them. That means fair compensation for data sharing, meaningful input into how systems are designed, and tools that augment farmer judgment rather than replace it.

Local over generic: The risk of global, aggregated models is that they flatten the specificity that makes good farming possible. Recommendations need to be grounded in local ecosystems, local conditions, and local knowledge. A tool that works in Iowa shouldn’t be assumed to work in Andalusia.

Transparency and open access: There’s a real opportunity for open-source infrastructure in agriculture – data that is collectively owned, well-documented, and accessible across languages and geographies. The alternative is a future where a handful of platform providers become gatekeepers to the knowledge that farmers need in order to farm well.

Accountability for the footprint: If we’re going to use AI in service of more sustainable food systems, we have to be honest about its environmental costs. That means pushing for renewable energy-powered infrastructure, advocating for land use policies that prevent data center expansion on prime agricultural land, and being skeptical of claims that the environmental benefits of AI in farming automatically outweigh its costs. This advocacy burden should be a shared responsibility, not something that falls on land stewards alone.

Keep the human in the loop: AI should be a resource for advisors and farmers, not a replacement for the conversation between them. Intuition, site visits, and relationships are critical and need to be protected.

I left the session feeling genuinely energized. Not because we necessarily solved anything, but because the people in the room were asking thoughtful questions. Technology will keep moving fast. What we build around it – governance, ownership structures, and values – has the potential to move at the speed we choose. It’s critical that the agricultural community is involved in these conversations, rather than watching from the sidelines.

---