Individuals exhibit differences in learning ability, but the reasons for these differences are unclear. We used human functional magnetic resonance imaging (fMRI) to show that dynamic changes in whole-brain networks during the early stages of sensorimotor adaptation predict patterns of learning that emerge across two days of learning and relearning. A clustering of participant behavioural data revealed three distinct profiles of learners: individuals who learned quickly on both days, individuals who learned slowly on both days, and individuals who learned slowly on the first day, but quickly on the second day. These learning profiles were associated with the degree of modular reconfiguration during early learning on the first day, and with the selective recruitment of a cognitive network of regions, including areas in anterior temporal and prefrontal cortex. These findings demonstrate that across-day profiles of adaptation can be traced to whole-brain dynamics that manifest during initial learning.