The majority of acute myeloid leukemia (AML) patients will respond to standard chemotherapy, however, resistance is a prevalent problem contributing to incomplete responses, refractory disease, and ultimately patient death. Therefore, choosing more sensitive and effective chemotherapy regimens is of key clinical importance. In order to explore this issue, we investigated and optimized PharmaFlow, an automated flow cytometry method for evaluating the sensitivity of leukemia cells to multiple chemotherapeutic drugs ex vivo. We examined bone marrow samples from 38 Chinese AML patients and incubated them for 48 or 72 h with a panel of 7 single drugs and 6 combinations with cytarabine at different concentrations. Leukemic cell depletion was assessed by PharmaFlow and drug response parameter, called PharmaFlow score, was estimated using population pharmacodynamic models. We identified that most chemotherapeutic drugs and combinations could effectively eliminate pathological cells ex vivo. Estimated drug activities strongly correlated with the patients' duration to achieve clinical remission and PharmaFlow chemosensitivity measured ex vivo was highly predictive of the clinical outcome after chemotherapy. Applying a classification model, we determined a PharmaFlow score of 89.4 as the threshold to predict response to chemotherapy. Using this threshold, we found that in 84.2% of cases patient's cell response ex vivo predicted the observed clinical response and performed similarly or better than prognostic subgroups determined by cytogenetic characteristics. PharmaFlow has the potential to predict chemosensitivity for de novo, secondary and relapsed AML patients prior to treatment and may guide clinicians to tailor treatments and improve patient outcome.