How do our brains transform perception into action? In this talk, I will present a series of MEG studies that reveal how sensorimotor decisions unfold dynamically across the human cortex. We show that motor cortex activity, long thought to reflect only action execution, predicts upcoming responses several seconds in advance. These early fluctuations are shaped by prior actions and drive systematic biases in behavior. Critically, we also identify abstract neural representations of perceptual choices, independent of motor responses, that are preserved across task contexts. These signals predict decision confidence and accuracy, suggesting they reflect an internal decision variable. Complementary recordings in non-human primates reveal converging evidence for distributed cortical interactions and opposing flows of sensory and related-related information. Together, these findings reveal a dynamic and distributed cortical architecture in which abstract and action-related signals jointly shape sensorimotor decisions.