The transition from foraging to plant cultivation represents the most important shift in the economic history of early Holocene societies. This process unfolded independently in different regions of the globe, resulting in varied plant assemblages, cultivation strategies, dietary practices, and landscape modifications. To investigate the drivers of this transition, we employed a machine-learning approach. Using Random Survival Forest, we analyze a comprehensive dataset of radiocarbon dates linked to the first adoption of domesticated plants, coupled with environmental predictors. Our findings indicate strong spatial autocorrelation in the timing of agricultural adoption, underscoring the role of diffusion and contact between regions. Region-specific bioclimatic factors emerged as influential: in the Americas, mean temperature and temperature seasonality were critical, while in Southwest Asia and Europe, seasonal variation in precipitation relative to temperature held greater importance. These results suggest that diffusion facilitated the spread of agricultural practices in a process shaped by local environmental conditions, as it was not possible to determine a set of universal drivers. Thus, the emergence of food production was influenced by a combination of local factors and cultural transmission, leaving the specific determinants for each region's first transition an open question for further study.