The Food System Doesn't Have a Data Problem. It Has a Timing Problem.
At TEDxBerlin's AI Business Day earlier this month, Catharina van Delden, co-founder and CEO of Finches, told a story that most people in this industry already know but rarely say out loud. The signals are there. They're just useless by the time anyone reads them.
Catharina's journey that brought Finches to life started in Uruguay. In 2022, she was standing in her own pecan field during the worst drought in the country's recorded history. More than a billion dollars wiped out across the agricultural sector in a single season. Farmers who had worked the same land for forty years were suddenly facing decisions they could no longer make from experience alone.
That gap between what the data was saying and what anyone could act on is the problem Finches was built to close. But to understand why it matters to manufacturers, not just farmers, you need to follow the supply chain downstream.
The Shock You Didn't See Coming Was Never About Energy
When the US/Iran crisis hit the Strait of Hormuz in early 2026, the coverage was predictable: energy prices, electricity subsidies, governments scrambling. What got far less attention was what happened to fertilizer.
Natural gas isn't just fuel, it's also the primary feedstock for nitrogen fertilizer production. When gas prices spike, fertilizer plants don't just get more expensive to run. They shut down. That's what happened across Europe. Fertilizer prices tripled in a matter of weeks and global food commodity prices followed.
The FAO Food Price Index has shown this dynamic before, after the 2021-2022 energy crisis, and the pattern is structurally identical: an energy shock becomes a fertilizer shock becomes a food input shock, and by the time it reaches a procurement desk in Frankfurt or Antwerp, the damage is already priced in.
Here's the uncomfortable part: this was not unpredictable. The inputs, the dependency chain, the regional concentration risks were all documented. What was missing was not information: it was a system that connected the information fast enough to matter.
This is the distinction that most risk conversations in manufacturing never quite reach. Companies invest in scenario planning, commodity hedging, supplier audits. All of it is oriented toward known risks and historical patterns. The problem is that the supply chain failures that actually cost money rarely announce themselves through the channels those systems are watching.
What Running Short Actually Looks Like
There's a version of supply disruption that everyone prepares for: the empty shelf, the force majeure clause, the sudden shortage that makes the news. But in practice, disruption often shows up much earlier, in smaller signals: a delayed shipment, a quality issue, a price movement, or a supplier quietly struggling to deliver as expected.
A hot, dry summer in central Europe creates the right conditions for aflatoxin-producing mould in maize. The mould grows in the field, but the harvest looks fine and the grain moves through the supply chain. Months later, it shows up in dairy feed, and then in milk, and a toxin that the EU classifies as a Group 1 carcinogen (meaning directly linked to cancer in humans) is found at levels that trigger a product recall. The cause is a weather event from the previous growing season; however the loss lands in this quarter.
That specific scenario is not hypothetical, as it has played out across multiple European markets in recent years, and the Austrian food safety authority flagged aflatoxin on domestic maize for the first time in 2025, a direct consequence of climate shifts pushing conditions that used to belong to southern Mediterranean agriculture into central Europe.
The point is not that this is a new risk. It's that the causal chain is long, the lag is measured in months, and the systems most manufacturers rely on are not built to see across that distance.
A World Economic Forum analysis of food system resilience found that the majority of supply disruptions affecting manufacturers are traceable to early signals that existed in the data but were never aggregated or acted on. The signals weren't hidden. They were in separate systems, separate formats, read by separate people with no mechanism to connect them.
The result is a specific kind of loss that doesn't appear in force majeure statistics: the crop that arrives on time, in full volume, and is completely unusable. Wrong moisture content, wrong grade, contamination that a lab will flag next month. A harvest that is commercially identical to one that never came.
TEDx Berlin: A World with AI
The Original Problem Was Never "More Data"
This is where the industry's instinct tends to go wrong. The response to supply chain opacity is usually more instrumentation: more sensors, more supplier audits, more data collection. And there is genuine value in that, but it addresses the wrong bottleneck.
The agricultural supply chain already generates an enormous amount of signal. Weather station networks, satellite imagery, lab results from commodity testing, agronomist field notes, trade flow data, regional disease and pest monitoring. The data exists and it's produced continuously. The problem is structural: it lives in incompatible systems, arrives through different channels, and is read by people who have no common frame of reference.
A lab result in one system, a weather anomaly in another, a note from a field agronomist that nobody put next to either.
None of that is intelligence on its own. Intelligence requires synthesis, and synthesis requires a system that can read across all of it simultaneously and recognize patterns that no individual data stream would reveal. The combination of a specific temperature range, at a specific growth stage, following a specific rainfall deficit, in a specific region, is what predicts aflatoxin pressure. Any one of those variables is noise. Together they're a warning.
This is exactly the problem Finches was built to solve. The platform takes signals that already exist across the supply chain: years of lab results, field notes, weather records, satellite data, geopolitical developments, and learns the combinations that ended badly in the past. Not in aggregate, across a region or a commodity category, but at the level of a specific supplier, a specific crop, a specific shipment. It then watches live supply chains for those same patterns forming again.
What's changed recently is that multimodal models can now process those different data types simultaneously and in context. A voice note from a field agronomist, a PDF lab report from three seasons ago, a satellite-derived soil moisture index, a line in a trade publication about port congestion in a key transit corridor. Different languages, different formats, different cadences. For the first time, that can be read together, at scale, fast enough to be useful.
The output isn't a regional risk report. It's a specific alert: this contract, this volume, this season, this risk. And it speaks in the language of procurement, not agronomy. Not "aflatoxin pressure is elevated in the region," but: this shipment is likely to come in off-spec, here is the volume exposed, and here is what it costs you if it does. Early enough to re-test, re-source, or work directly with the grower, months before the problem shows up in a lab report.
What This Means for Manufacturers, Specifically
The framing of food supply risk as a farmer's problem has cost the manufacturing sector significant margin over the past decade. Because the risk starts outside the factory, it is often left out of the systems manufacturers use to manage operational risk. It sits in procurement, where it's tracked through supplier relationships and contract terms, not through the underlying agricultural dynamics that actually determine whether those contracts will hold.
This creates a structural blind spot. Procurement teams are usually focused on whether suppliers can deliver on price, volume, and timing. But the signals that shape those outcomes often appear much earlier: in weather patterns, crop disease risks, fertilizer costs, and regional quality forecasts. In many manufacturing organizations, no team clearly owns that layer of risk.
The consequence, which shows up in earnings calls as raw material variance and in operational reviews as quality failures at intake, is that manufacturers keep absorbing risks that could have been visible earlier. The signals may exist, but they often sit outside the systems used to make sourcing decisions.
The useful reframe here is not "supply chain resilience" in the abstract. It is the ability to look at each contract before the season plays out and ask a more practical question: is this volume, from this region, under these conditions, still likely to arrive on time and on spec?
Forward
Catharina's talk in Berlin was less about technology itself and more about the challenge behind it: making confident sourcing decisions when the supply chain depends on crops, weather, and growing conditions that are becoming harder to predict.
The global food system produces more calories today than at any point in human history. The constraint isn't production capacity. It's predictability. And predictability, unlike yield, is not a function of what happens in the field. It's a function of whether the right information reaches the right people before the decision window closes.
For manufacturers whose margins depend on agricultural inputs arriving on spec and on time, this is not a distant supply chain issue. It affects costs, delivery, quality, and ultimately quarterly performance. Companies that build around that reality will not avoid every disruption, but they will have a better chance of seeing the risk early enough to respond.
Sources
FAO Food Price Index - Monthly tracking of international food commodity prices, including the March-April 2026 rises driven by the Near East conflict and Strait of Hormuz disruptions. https://www.fao.org/worldfoodsituation/foodpricesindex/en
FAO: "FAO Food Price Index rises in March as Near East conflict raises energy costs" (April 2026) - Documents the fertilizer and energy price shock following the Strait of Hormuz closure, including urea prices reaching their highest level in over three years. https://www.fao.org/newsroom/detail/fao-food-price-index-rises-in-march-as-near-east-conflict-raises-energy-costs/en
FAO: "Global Agrifood Implications of the 2026 Conflict in the Middle East" - Estimates that approximately one-third of global fertilizer trade was stalled, with 3-4 million tonnes per month not reaching buyers following the effective closure of Hormuz. https://openknowledge.fao.org/server/api/core/bitstreams/1aafb5d8-39d1-481a-b1f8-25facaec3051/content
World Economic Forum: "The next food crisis is already in motion" (May 2026) - Argues that early warning systems monitoring climate, market, and trade disruptions can help companies anticipate risks before shocks become crises, and that legacy supply chain models can no longer respond effectively to the current pace of disruption. https://www.weforum.org/stories/2026/05/strait-hormuz-food-security-crisis-fertilizer/
WEF: "First Movers Coalition for Food: CEO Lessons for the Future of Food Procurement" (January 2026) - Draws on interviews with Chief Procurement Officers to argue that the era of stability in food supply chains is over and that procurement can no longer function as a back-office function. https://supplychaindigital.com/news/wef-future-food-supply-chains
Rennhofer et al., Food Control (2026): "Mycotoxin field trial uncovers first aflatoxin B1 occurrence in maize in Lower Austria" - Peer-reviewed study documenting the first detection of aflatoxin B1 in Tulln, Lower Austria, linked to the unusually hot and dry conditions of 2024. https://www.sciencedirect.com/science/article/pii/S0956713526003038
Battilani et al., Scientific Reports (2016): "Aflatoxin B1 contamination in maize in Europe increases due to climate change" - Modelling study predicting that aflatoxin contamination in European maize will increase significantly under +2°C climate change scenarios. https://www.nature.com/articles/srep24328
MDPI Toxins (2025): "Pre-Harvest Aflatoxin Contamination in Crops and Climate Change Factors: A European Overview" - Reviews evidence that rising temperatures, droughts, and shifting rainfall patterns increasingly favour aflatoxigenic fungi across European crops. https://www.mdpi.com/2072-6651/17/7/344
