A farmer destroyed a crop the size of 20 football fields on AI advice: the danger of answers without your data

A farmer destroyed his own crops covering roughly the area of 20 football fields after taking AI advice. The model recommended a radical measure — from the described symptoms it judged the situation hopeless — but knew neither the real state of those specific fields nor the local conditions. By the time it became clear the harvest could have been saved, it was too late.

The culprit in such stories is not "bad AI" but the gap between the question and the data. A general-purpose chat answers from a world-average picture: it never saw your fields, your soil tests, your treatment history. The confident tone of the reply masks the fact that it is, at bottom, an extrapolation of other people's experience. The more expensive the decision, the more dangerous that gap — and the farmer made a very expensive decision.

The lesson transfers to any business. Rule one: decisions with irreversible consequences — destroy, fire, terminate, dump stock at a giveaway price — are not taken on the advice of a tool that cannot see your data. Rule two: if you want AI to genuinely help decide, give it data — connect it to your accounting, sales, and operations history. An AI agent answering from your 1C and your CRM can still be wrong, but at least it is wrong about your reality, and its reasoning can be checked. Advice out of thin air cannot.