Updated publication reference for PubMed record(s): 34857927. Protein phosphorylation dynamically integrates environmental and cellular information to control virtually all biological processes. Identifying functional phosphorylation amongst the thousands of phosphosites regulated by a perturbation at a global scale is a major challenge. Here, we introduce personalised phosphoproteomics, an experimental and computational framework to link signalling with biological function by exploiting human phenotypic variance. This is achieved by measuring individual subject phosphoproteome responses to interventions with corresponding phenotypes measured in parallel. We employed this method to investigate how exercise potentiates insulin sensitivity in human skeletal muscle. Our approach extracted signalling highly enriched in known functional sites, as well as many previously unknown sites on proteins intimately involved in glucose metabolism. This included a new co-operative relationship between mTOR and AMPK, for which we found a role in metabolic regulation. These results establish personalised phosphoproteomics as a general approach to investigate the signal transduction underlying complex biology.