Abstract: Systems and methods for using rule-based preprocessing to improve artificial intelligence prediction of next best actions are disclosed. Prescription information may be used to identify a parameter set for a machine learning model. The parameter set may first be processed through a rules engine having a first rule set configured to conditionally determine the next best action service call without processing through the machine learning model. The machine learning model may then selectively process the parameter set to determine the next best action service call if the first rule set did not meet a next best action condition in the rule set. The next best action service call may be initiated to execute the next best action. For example, in response to rejection from insurance adjudication, the next best action may change the prescription information for automatically resubmitting the prescription to insurance adjudication.