Acute myeloid leukaemia (AML) is a blood malignancy with poor prognosis and a limited number of first-line treatments, one of which is midostaurin combined with intensive chemotherapy (MIC). MIC is approved for FLT3 mutation-positive (FLT3-MP) AML, yet many eligible patients are refractory or experience an early relapse. Development of a MIC stratification method that outperforms FLT3 mutational status would thus benefit a large number of patients with AML. We employed phosphoproteomics, a mass spectrometry technique that quantifies thousands of cellular signalling events in a single sample, to analyse 71 diagnosis samples of 47 FLT3-MP patients with AML who subsequently received MIC. We then used machine learning to identify biomarkers of response to MIC, and validated the resulting predictive model in two independent validation cohorts. We identified three distinct phosphoproteomic subtypes amongst long-term AML survivors. The subtypes showed similar duration of MIC response, but different modulation of AML-implicated pathways, and exhibited distinct, highly-predictive biomarkers of MIC response. Using these biomarkers, we built a phosphoproteomics-based predictive model of MIC response, called MPhos. When applied to two retrospective real-world patient test cohorts (n=20), MPhos predicted MIC response with 100% sensitivity and 87.5% specificity (log-rank p<3.0e-4, HR=0.04 [95% CI:0-0.39]). In validation, MPhos outperformed the currently-used FLT3-based stratification method. Thus, it has the potential to transform clinical decision-making, with important implications for the role of phosphoproteomics in precision medicine beyond AML.