As technology and artificial intelligence continue to become more prevalent in the fresh produce industry, Francisco Martin-Rayo, co-founder and CEO of Helios AI, an AI-powered intelligence platform that forecasts climate risk and predicts commodity prices, sat down with The Packer to share how AI can transform reporting from reactive to a proactive risk management strategy.
How does the Helios AI model help open-field produce shippers and buyers prepare for severe microclimate shocks before they hit the field?
Martin-Rayo: Most of the industry runs on hindsight. A field report tells you what already happened, and by the time it circulates, the market has moved and the alternative supply has been booked by someone else. We work the other direction.
We’ve mapped 90% of the districts in the world growing crops for export down to 14 million hexagons. We run a separate machine learning model for each commodity, and we reforecast every 24 hours. That gives a shipper or a buyer lead time measured in weeks and months rather than days.
We predicted the Brazilian citrus crisis eight months before Reuters and Expana and 12 months before USDA. When cherries froze in Michigan and Turkey at the same time, our customers already knew what that meant for a fruit that isn’t substitutable. Lead time is the whole product. Everything else is commentary.
Open-field specialty crops are far more sensitive to weather anomalies than broad-acre row crops. What specific climate metrics are proving to be the biggest drivers of open-field yield loss today?
The single biggest driver is heat outside a crop’s optimal range during a narrow reproductive window — not seasonal average heat, which is what most models look at. A stone fruit orchard doesn’t care what July looked like on average; it cares about the four days during flowering when temperatures sat where pollination fails. That’s why we don’t run one weather model. We run one per crop, because the temperature that ruins a bloom in a cherry orchard is fine for a tomato field.
After that, three metrics do most of the damage in open-field specialty crops: chill hour accumulation in winter, which sets the ceiling on next season’s crop before you’ve planted anything; rainfall intensity rather than rainfall total, because 60 millimeters in an hour is a disaster and the same volume over a week is a good year; and consecutive dry days during fruit sizing, which is where you lose grade and packout rather than the crop itself.
The reason specialty crops get hit harder isn’t that they see worse weather; it’s that the damage windows are dayslong, the recovery options are limited, and with a perennial you’re not replanting next spring — you’re rebuilding for five years.
How do early-warning yield alerts give both open-field growers and retail buyers a buffer to adjust supply commitments and pricing fairly before a regional shortage hits?
Panic buying isn’t caused by shortage. It’s caused by everybody discovering the shortage on the same morning. When discovery is simultaneous, every buyer bids at once, freight tightens at once and price gaps rather than moves. That’s when contracts break.
Early warning breaks the simultaneity. A grower who can see a regional problem forming eight weeks out goes to their buyer with a forecast and a proposal. That’s a very different conversation from a force majeure letter after the harvest is already lost. The buyer, on the other side, can qualify a second origin while there’s still fruit and still freight to book, instead of paying a spot premium for whatever is left.
The fairness piece matters more than people think. Most contract fights are really information fights, where one side knew something the other didn’t. When both sides are working off the same forecast, you’re negotiating a schedule instead of assigning blame.
How does predictive climate intelligence help an open-field grower make practical operational calls ahead of time?
Our customers are largely on the buy side, procurement teams and retailers, so I want to be careful not to claim we’re an irrigation controller. We’re not. But the calls a grower makes ahead of a dry spell are mostly allocation calls, and allocation is where a seasonal forecast beats a 10-day one.
Water is a budget, not a tap. If you know in April that August is going to be dry, you spend water differently in May. You decide which blocks you carry to full quality and which you take to a lower grade on purpose. You move harvest crews and cold chain capacity forward. You decide whether the second cycle goes in the ground at all, which is a decision that costs real money to get wrong in both directions.
Those are business decisions made months ahead, and you can’t make them off a 10-day forecast.
How does Helios demonstrate a clear, measurable financial ROI to an open-field operation using Helios insights to hedge against regional crop failures and protect their margins?
We benchmark against what the market knew and when it knew it, because that’s the only honest way to price a forecast. Bain evaluated 13 platforms in this space and ranked us first. Against a 700-event benchmark, we catch roughly 90% of disruptions.
In practice, the return shows up in three places. Avoided premium buying, which is the biggest — we saved one importer-exporter over $2 million by calling the Peruvian mango disruption five months before prices went up [fivefold]; contract timing, meaning locking volume before a market reprices rather than after; and working capital, because knowing which risks are real lets you stop carrying inventory against the ones that aren’t.
I’d add a caveat any serious buyer will appreciate. You can never fully prove a counterfactual. What you can do is put a date on the call, put a date on when the market moved and let the customer price the gap themselves. Ours is usually measured in months.
What else would you like our readers to know about technology and AI’s role in climate-smart farming?
Two things. First, climate-smart farming has been talked about mostly as a set of practices, and it’s really a risk-management discipline. Cover crops and drip irrigation are good. They don’t help if you’re growing the wrong crop in a district that’s shifted archetype over the last decade. We classified every one of our 14 million hexagons into 81 climate archetypes precisely to answer that question, which is a strategic question about where you farm, not just how.
Second, a grower in Salinas is now competing in a market set by weather in Almería, Michoacán, and the Nile Delta. You can farm perfectly and still get repriced by a heat wave 6,000 miles away. That’s not a reason for pessimism; it’s a reason to stop treating global supply intelligence as something only the big buyers need.
Farming is the hardest business in the world. If everything is perfect, you make a few points. If anything goes wrong, you lose money. Our job is to make sure that when something goes wrong, you saw it coming.
