Abstract: Systems and methods are provided for automated modification of delivery parameters. Particularly, computing model that is trained to determine a probability that a delivery defect is likely to occur for a given delivery or set of deliveries. Based on the probability, various limitations associated with the deliveries may be activated or deactivated on a mobile device application used by a delivery driver to perform the deliveries. The systems and methods reduce the number of delivery defects that occur while simultaneously reducing the use of unnecessary guardrails for low-risk deliveries. The model may be queried in real-time such that guardrails for a delivery itinerary may be optimized prior to the delivery driver beginning the delivery route.