Daniel Schüler, AI Strategist at AEB, explains why AI alone does not solve structural problems and why data quality, process clarity and strategic automation are crucial.
AI can speed up processes and assist with tasks such as classification. Several analyses have now revealed efficiency gains, most recently a report by the WCO on customs administration. But it does not solve structural problems. The conditions have to be right.
Firstly, it is important to be clear on the fact that AI is not the same as automation. It does not replace clean data, clear processes, and system integration. Alongside established machine learning, generative AI is a building block within an automation architecture. Companies that use AI in isolation create new dependencies and increase their compliance risk. AI reinforces existing structures, both good and bad. That is why an automation strategy is necessary; tailored automation if you will.
That very much depends on the company, the initial data, and the particular use case. A precise automation strategy combines several levels. It asks: What technology do I need for what application? How do tools work together? How are they integrated? In the area of classification, we use a combination of rule-based automation for clear-cut cases, machine learning for pattern recognition, and generative AI for complex individual cases and justifications. When classifying simple goods, for example, a customer can achieve a degree of automation of up to 100 percent. And in the other areas, the classification process is accelerated considerably.
Yes, where it is sold as a magic bullet. AI cannot make up for inconsistent data. It cannot replace a lack of process structure. And it cannot establish governance. Yet, at the same time, its strategic potential is underestimated when it is deeply integrated into systems and interacts with them. “AI alone” is no longer sufficient in 2026. The crucial factor is how precisely it is integrated into processes, software, and data structures.
That it's often not good enough. Many companies overestimate the quality of their data. In reality, customs data and material master data are rarely consistent. Different languages, variable levels of detail in descriptions, organically grown structures from different ERP generations, different storage locations, and historical inconsistencies prevent stable automation. An AEB analysis on the automation of classification shows that the biggest weakness in AI and automation projects in the customs environment is the data base.
There are two crucial factors: 1. operational process optimization and, secondly, strategic data sovereignty. Process optimization means clear responsibilities, clean workflows, and automation of recurring tasks. Data sovereignty goes deeper. Companies that structure and control their data centrally retain control over automation, compliance, and speed, even during ERP transformations. This becomes a competitive advantage.
Independent and central master data management reduces dependencies on changes in IT, for example in ERP systems, and increases the stability of customs processes. Many ERP systems are not designed to handle master data relating to global trade. Managing this data exclusively within such systems leads to structural bottlenecks.
Manual or semi-automated processes work so long as volumes and markets remain stable. As soon as new markets are added, regulatory changes increase, or product variants grow, classification becomes a bottleneck and customs processes become a holdup within the company. This is because such changes trigger a cascade of adjustments in a wide variety of areas and systems. As a result, the risk of errors and compliance violations also increases.
Human expertise remains indispensable. For complex machines, modular systems, and new technologies with little history, artificial intelligence prepares and thus accelerates classification decisions. Standardized, recurring, yet time-consuming classification cases, on the other hand, can be fully automated.
Data. AI is becoming a commodity and will continue to develop dynamically. Costs are falling. Competition remains intense. The value of AI lies in the application, the specialization, and in how it is embedded. The competitive advantage comes from the quality, structure, and availability of the data. And from the ability to embed AI in a stable architecture.
Product Classification is a reliable, intuitive tool for finding just the right import and export commodity code and export control number – in a process that can be manual, automated, or customized to your precise needs.
