Understanding how the brain flexibly adapts to its environment requires moving beyond isolated regions to consider system-wide interactions. In this talk, I will present our work developing NeuroAI models of brain-wide learning and dynamics. I will highlight the critical interplay between cortical, subcortical, and neuromodulatory systems and how this shapes adaptive behaviour. Our findings suggest that cortical networks may be less plastic than traditionally thought, with subcortical structures such as the cerebellum rapidly learning goal-driven predictions that guide cortical dynamics. At the same time, our work suggests that cholinergic neuromodulation adaptively regulates cortical learning to support resilient task representations. Together, these results point to learning as an emergent property of multiple interacting systems, whose continuous interplay underlies robust and flexible behaviour.