The experimental study used machine learning (ML) procedures to classify the intensity of flow based on physiological signals and yielded multiple insights. Firstly, evidence was found that flow-classifiers can be created based on peripheral nervous system features alone. Secondly, it became apparent that cardiac features (HRV-LF, HRV-HF, LF/HF ratio) played an important role in classifying the intensity of flow as the prediction accuracy with these features was 72.3% (with Random Forest ML-procedure). Thirdly, the study showed that flow can also be induced with a real-world working task (SAP HANA invoice matching task) in a laboratory environment. This insight can help to extend the body of knowledge of how to induce flow in experimental settings. Finally, our findings can serve as a cornerstone for further research efforts to build flow-aware IT-systems (called Flow Computing) capable of automatically assessing flow in real-time during task execution based on human physiological data. For instance, in the future Flow Computing system could be developed to prevent an individual from being interrupted in the middle of an ongoing task by ensuring that no e-mails or notifications are forwarded, as long as the system is “sensing” flow.