A new article by Jawoo Koo, Senior Research Fellow at IFPRI for CGIAR, highlights that while AI offers new possibilities for agriculture, unresolved data gaps—particularly in the Global South—limit its impact. CGIAR’s GARDIAN platform indexes more than 470,000 publications and digital assets, 26,000 datasets, and over half a million standardized agronomic data points to help researchers and AI developers. However, making data available is only part of the challenge. The article describes a project using a mobile phone survey in a country in East Africa to study how market access influenced farmers’ adoption of agricultural technologies. Initial results showed lower adoption farther from markets, but adoption appeared to increase sharply in the most remote areas. The unexpected finding was due to selection bias: only wealthier farmers in those areas could afford mobile phones, unintentionally excluding poorer households. This “streetlight effect” mirrors broader agricultural data challenges, where many “dark corners” remain, especially in the Global South, making representative data collection difficult. Since AI can only learn from provided data, missing communities or farming systems create blind spots that result in bias. The article also notes that even when valuable agricultural data exists, AI models may never use it, referencing 2023.