From Rules to Learning: Why AI-Native Platforms Are the Future of Customs Technology
Jean Gurunlian, Chairperson of Webb Fontaine and the founder and architect of the ASYCUDA system, has issued a stark warning on the future of Customs technology, arguing that legacy platforms are no longer viable in an era shaped by large language models (LLMs).
Speaking at the WCO Technology Conference 2026 in Abu Dhabi, Gurunlian said that the emergence of LLMs has fundamentally altered the technological requirements of Customs administrations worldwide.
“No Customs system that has not been built on LLM will survive,” Gurunlian stated. “The first real outcome of LLMs is that they have made all existing Customs systems obsolete.”
Having designed ASYCUDA and overseen its deployment in more than 100 countries, Gurunlian highlighted how the pace, scale, and unpredictability of regulatory change have outgrown traditional system architectures. He noted that many Customs platforms remain dependent on static rule sets, manual updates, and lengthy development cycles—approaches he said are incompatible with today’s trade environment.
“Customs systems that cannot adapt to new laws, regulations, or operational requirements within very short timeframes simply cannot survive,” he said. “If a system takes years to adjust, it is already irrelevant.”
According to Gurunlian, LLMs have exposed a critical structural weakness in legacy systems that rely on predefined logic rather than continuous learning. Many platforms, he explained, still require months or even years to implement legislative amendments, tariff revisions, or new non-tariff measures.
“Tariffs and non-tariff barriers have increasingly become political instruments,” he said. “They can change overnight, sometimes without warning. With LLM-enabled systems, those changes can be interpreted, applied, and operationalised in seconds.”
Gurunlian stressed that adaptability is no longer a competitive advantage but a baseline requirement. Customs technology, he argued, must now be designed for continuous improvement, contextual understanding, and rapid learning—capabilities that only AI-native architectures can deliver.
“The systems, including those I created, are destined to become obsolete,” Gurunlian acknowledged. “If a system cannot be improved while in production, it should not be deployed at all.”
He concluded by warning that governments and Customs administrations face a pivotal decision. Continued investment in static, rule-based platforms risks locking institutions into technology incapable of responding to geopolitical shocks, regulatory volatility, and the growing complexity of global trade.
