An AI model that integrates data from continuous glucose monitoring, microbiome analysis, physical activity levels and wearable sensors reveals hidden glycemic heterogeneity among individuals with identical HbA1c values. These nuances could improve the stratification of metabolic risk in decentralized clinical studies.