Artificial intelligence and farmers' trust in agronomic forecasts

In western India, a farmer who contracted with a major food manufacturer received data on the future harvest through a field agent. The system estimated the expected volume and quality of produce based on the applied working methods and inputs used, and also forecasted the labor cost level. Thanks to the explanation by the field agent, who demonstrated agronomic inspection data, satellite indicators of crop health, and quality standards from similar farms, the grower understood the logic of the forecast. After the actual harvest matched the system's calculations, the farmer's attitude toward such forecasts changed.

Forecasting yield and quality allows agribusinesses to plan procurement in advance, optimize processing schedules, and reduce losses three to four weeks before harvesting. Nevertheless, the accuracy of such decisions depends on the quality of the initial data, and farmers' readiness to follow recommendations depends on their clarity and the reliability of the information source. Experience from more than 35 corporate implementations shows that forecasts without explanations arouse suspicion and are perceived as groundless verdicts.

To build trust in artificial intelligence forecasts, three conditions must be met. First, the system must rely on real data from a specific farm, including ground-based records, agronomist observations, and resource accounting logs, rather than just regional averages. Second, the forecasts must be understandable to the farmer in the language of their experience. Third, the information must be communicated by a trusted person, such as an agronomist, field agent, or cooperative worker. If these conditions are met, growers share observations more actively and help improve the system's performance.