<![CDATA[This paper addresses package type selection in oil and gas inbound logistics, a routine manual decision that directly determines the freight cost invoiced by the third- party carrier delivering shipments from suppliers and from the organization's own warehouses. Each selection maps to a package indicator carrying a contractually agreed carrier rate, so an inaccurate selection produces an inaccurate invoice. In a large oil and gas producing organization handling over one million packages annually, this decision depends on human judgment and is inherently error-exposed at scale, while manual verification covers less than five percent of delivered volume, leaving most selections unvalidated. Drawing on operational data covering millions of packages delivered across six years, this paper makes two contributions. It establishes package type selection as a distinct and previously unaddressed source of recurring cost leakage, a decision that no existing material classification standard governs and that the logistics literature has not treated as a problem. It then proposes a process automation framework that closes this gap without depending on labelled training data, beginning with a rule-based validation that checks each selection against the material's recorded dimensions and specifications and automates the cost calculation before shipment delivery. The framework is designed to evolve through machine-learning and master- data-driven phases toward fully autonomous, cost-accurate package selection, offering a transferable approach for any inbound logistics operation in which package type drives freight cost.]]>
