Auditing can be a time-consuming task due to unintentional or deliberate errors made by the domain users. Computational methods, including machine learning and blockchain, can address these issues. Our work presents a two-step approach for a transparent automated logistics audit classification system using a combination of blockchain and machine learning. The first stage employs unsupervised learning techniques to identify distinct transaction subtypes, which are subsequently categorized manually by expert auditors. The next stage involves training a classifier, potentially using artificial neural networks, on the tagged records to enable real-time transaction classification. This enables the system to automatically detect any anomalous activity, thereby enhancing the transparency and efficiency of logistics audits, presenting a conceptual framework for a continuous and compliant audit system.