Understanding the neural dynamics underlying motor learning is pivotal for revealing how the brain adapts to acquire new skills. Plasticity in the primary motor cortex is essential for motor skill learning, with Layer 5 pyramidal tract (L5 PT) neurons playing a key role in the process. Previous studies have shown dendritic spines undergo plasticity changes during motor learning, yet the resulting changes at the dendritic level remain unknown. To investigate changes occurring in apical tuft dendrites during motor learning, we performed repeated two-photon calcium imaging of the same L5 PT apical tuft dendrites in mice, as they learned a skilled lever-pull task. We simultaneously recorded activity from multiple dendrites across 5-15 cells. Using our experimental design we reconstructed the morphological structure of dendritic branches in detail within the tuft trees of each recorded cell, which enabled us to analyze both functional and structural aspects of the network. We used a novel methodology based on Riemannian Geometry and wavelet analysis to characterize the transformative process of the dendritic network induced by learning and to reveal the components driving it. Using this nonlinear method, we identified a subpopulation of dendritic regions driving network reorganization during learning. Surprisingly, most cells were identified as a whole, with all their recorded dendritic regions either driving the dynamics or not. Furthermore, our results show that these cells are mostly responsible for the transformation in network dynamics towards an expert configuration in terms of ensemble activity, functional connectivity, and behavior encoding. Our findings provide new insights into the hierarchical nature of plasticity in motor learning, showing how the activity of different tuft dendrites gradually reorganizes within and across neurons while mice acquire new motor skills. By elucidating the dynamics of dendritic activity during learning, this work advances our understanding of the neural basis of motor skill development.