Agriculture is ready for AI, but its data isn’t
MIT Technology Review Carole Hill, Manish Sood
AI could boost crop yields big time, but only if farm data isn't a mess. Most ag AI pitches skip that inconvenient detail entirely.
Based on reporting by MIT Technology Review, Carole Hill, Manish Sood — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Farming has never lacked for promises about technology. The latest one is AI models that claim to lift crop yields by over a quarter, cut water use by 41%, and slash chemical spraying by a third. Those numbers, cited in a new industry piece from data-management vendor Reltio, sound great on a slide. The catch, which vendors tend to leave out of the pitch, is that none of it works without clean, connected data underneath it.
And agriculture's data problem is genuinely gnarly. A single operation might be pulling in feeds from autonomous tractors, drone imagery, soil sensors, irrigation controllers, USDA reports, and third-party market data, none of which were built to talk to each other. Layer on top of that the need to track GPS coordinates, field boundaries, and soil variation within a single property, because fertilizer rates that make sense in one corner of a field can be wasteful or damaging fifty yards away. Get that context wrong and the AI doesn't just underperform, it actively recommends the wrong thing, at scale, with real chemical and financial consequences.
Reltio's argument, built around its work with 104-year-old distributor Wilbur-Ellis, is that governance and a unified data model are the boring prerequisites nobody wants to fund before the flashy AI project. Knowing which customer farms which field, which supplier sold what input at what price last season, and how that ties to margin sounds like plumbing work, not innovation. But stale plumbing produces stale answers. A pricing relationship that was accurate six months ago and never updated will feed an AI system that's confidently making decisions about a business that no longer exists.
Worth flagging: this piece was produced by Reltio itself, an SAP company that sells exactly this kind of data-unification software. So take the specific product claims with the appropriate grain of salt. But the underlying warning holds regardless of who's selling the fix — agricultural margins are thin, chemical misapplication has real environmental and legal weight, and an AI hallucination in a spreadsheet is annoying while one in a field is expensive and possibly illegal. The industry's AI enthusiasm is real. Whether the data underneath it is ready is a separate, less exciting question that a lot of buyers are skipping.
My take — AI-written commentary, not fact-checked reporting
This is a vendor writing a trend piece to sell you the fix for a problem it just described, which doesn't make the diagnosis wrong, it just means you should read the yield and water-savings stats as marketing math rather than peer-reviewed fact. The real lesson generalizes way beyond farming: every industry rushing to bolt AI onto messy, siloed legacy systems is going to get confident-sounding garbage out, and agriculture just has the misfortune of finding that out with real chemicals in real soil.
Read more about this at: MIT Technology Review