The Center for Data Innovation recently spoke with Catharina van Delden, CEO of Finches, a Germany-based company using an AI-powered platform to help food and ingredient companies anticipate agricultural supply-chain risks. Van Delden explained how Finches combines weather, satellite imagery, supplier records, and other data to identify potential disruptions before they affect sourcing decisions.
David Kertai: What inspired Finches’ creation?
Catharina van Delden: During my first year on our family farm in Uruguay, the country experienced its worst drought in decades. Many neighboring farmers lost their livelihoods. What stayed with me was that the warning signs appeared long before the losses. Rainfall dropped, soil moisture levels fell, and forecasts pointed to worsening conditions months in advance. Yet the people who needed to act on those signals often lacked a clear picture of what was happening.
The same problem exists further up the supply chain. Companies that source agricultural ingredients often rely on contracts, supplier certificates, previous harvests, and long-term averages to make purchasing decisions. These sources describe past conditions, not what is happening in the field. Meanwhile, early warning signs are scattered across weather reports, field visits, laboratory results, satellite imagery, and email messages.
We created Finches to bring these signals together before problems become crises. Our AI-powered platform analyzes information from across agricultural supply chains to help companies identify risks such as drought, quality issues, crop diseases, and supply disruptions while they still have time to respond. It gives sourcing teams a clearer picture of conditions across their suppliers and growing regions.
Kertai: What types of information do you collect to inform your AI platform’s decisions?
Van Delden: We combine two broad categories of information. The first is a company’s own sourcing data, including supplier records, contracts, crop varieties, laboratory results, and field observations from agronomists, agricultural specialists who advise on crop production and growing conditions. Agronomists can record field visits through our mobile application using voice notes, photographs, and written observations, even in areas without Internet access. By the time they leave the field, our AI system has turned that information into a structured report that the wider team can search and analyze.
The second category comes from outside the company. We collect information from weather models, satellite imagery, soil data, crop-science research, trade regulations, and local news sources, often in the language of the region where crops are grown. Satellite imagery helps us monitor conditions across large agricultural areas, while weather models provide information about current and forecast conditions. Some customers do not employ agronomists, so Finches can also operate entirely with external data.
Kertai: How does your system connect different events, such as unusual weather patterns and crop diseases, to potential disruptions in agricultural production?
Van Delden: At the center of the platform is Bob, our AI system. Bob consists of specialized agents that monitor weather, satellite imagery, scientific research, news sources, and each customer’s internal data. The system continuously compares new information against a detailed agricultural model that maps crops, varieties, regions, suppliers, contracts, and growing conditions.
Consider durum wheat, a variety of hard wheat commonly used to make pasta. Warm, wet weather during flowering, the stage when the plant develops its reproductive structures, increases the risk of mycotoxins, harmful compounds produced by certain fungi. Weather models can identify those conditions early, while satellite imagery can reveal standing water in affected areas. Traditional laboratory testing may not detect the problem until harvest. Finches connects these signals to the suppliers and fields at risk, alerting sourcing teams while they still have time to secure alternative supplies.
Every alert includes supporting evidence and information about its sources. Rather than simply flagging a risk, the system explains what happened, who it affects, and what actions a company might consider, such as conducting additional testing or identifying backup suppliers. Finches supports decision-making but does not replace it. Procurement, quality, and laboratory teams remain responsible for final decisions.
Kertai: How does your system account for differences between crops and varieties in a given field?
Van Delden: Our customers already know which crops and varieties they source from each supplier through contracts, supplier records, and field visits. That information enters the platform directly from their existing operations.
The challenge is understanding what a signal means in its specific context. A frost event has very different consequences for spinach than it does for almonds. Likewise, the same rainfall pattern can create different risks depending on the crop, variety, location, and stage of growth. To support that analysis, Finches maintains a detailed agricultural model that connects crops, varieties, environmental conditions, laboratory measurements, suppliers, and contracts. This allows the system to evaluate events in context rather than treating every signal the same way.
Kertai: Which risks have you found hardest for companies to anticipate, and how does your approach help them identify those risks earlier?
Van Delden: Since 2024, we have spoken with more than a hundred sourcing professionals. We have often heard the same story: early warning signs existed in weather data, field reports, supplier communications, laboratory results, and other sources, but they failed to reach decision-makers in time. The information sat in spreadsheets, arrived weeks late, remained in individual employees’ knowledge, or got lost among competing priorities.
Quality issues are a good example. A poor harvest often reveals itself through rising prices, but quality problems may not emerge until a shipment arrives and fails inspection. By that point, alternative sourcing options are limited, and the market has often already reacted.
We monitor sourcing regions and suppliers to connect these early indicators, or weak signals, before they develop into major problems. By combining weather, satellite, scientific, and field data, we help companies identify emerging risks while they still have time to secure alternative suppliers, adjust purchasing plans, or change production schedules. Our goal is to help customers respond when they first see smoke, not when the fire is already spreading.
