How would you decide whether to use a heuristics-based rule engine or a machine learning model for fraud detection at Rippling?
