QUIET pressure is building across our profession. Customs brokers are being pushed, with increasing urgency, toward automation, digital platforms and AI-driven workflows. We are fewer in number, we are ageing as a workforce and the case for efficiency grows louder every year.
That case is not without merit. AI has a genuine role in processing, data handling and reducing operational load. Nobody is arguing otherwise. But efficiency is not the only value that matters in this line of work and this is the point being lost in the rush toward automation.
The responsibilities of a customs broker have not changed because the technology around us has. The legal framework within which we operate is the same as it was before anyone had heard of large language models and within that framework, the standards are clear.
In tariff classification, we are still required to apply the legal notes and work through the interpretative rules. These are not administrative to be set aside, rather they are the foundation of a defensible classification decision. An AI system that generates a tariff heading does not apply those rules in any legally meaningful sense. What it does is pattern-match, predict and produce a plausible output. That is not the same thing.
It was suggested to me recently that industry and government would simply adopt AI classification as the truth. I found that troubling and it deserves a direct response.
Australia operates under the Westminster system of government, which provides for the separation of powers, including the independence of the judiciary. It is the courts and only the courts, that can authoritatively interpret legislation. And it is therefore the courts alone that can give a deciding classification in a disputed matter.
No AI system sits above that. No consensus among industry participants changes that. A classification that cannot withstand scrutiny before a court is not ‘the truth’, it is simply an error that has not yet been challenged. When a classification is disputed by the Australian Border Force, what matters is whether a defensible, evidence-based decision was made by a qualified professional who understood the rules and applied them to the facts.
“AI did it” is not a defence. This may seem obvious, but it bears repeating, because the logic of efficiency can obscure it. When a broker submits a classification, that classification carries their name and their licence. The accountability does not transfer to a software vendor.
AI systems, including the most sophisticated currently available, hallucinate. They produce confident, well-structured, entirely fabricated answers when they encounter gaps in their training data. They are optimised to give you something rather than nothing and they are not optimised to be correct. In a dispute with the ABF, a confidently wrong AI-generated answer is not a mitigating factor. It is a liability.
Experience cannot be shortcut. Judgement cannot be downloaded. The capacity to look at a set of facts, apply rules, weigh competing arguments and arrive at a defensible position comes from years of practice and accumulated knowledge and experience. That means from being present, from making mistakes and learning, from understanding how the system works at the operational level, not just in theory.
None of this is an argument against technology. It is an argument for perspective. AI can be a powerful tool, accelerating research, flagging inconsistencies, managing routine processing and freeing up cognitive capacity for the work that genuinely requires human judgement.
But it must be used as a tool, not deferred to as an authority. The professional remains responsible for the outcome. The professional remains accountable for the decision. And the professional must remain engaged enough with the substance of the work to know when the tool is right and when it is not.
Doing our job right matters. Meeting timeframes matters. Experience matters and bringing genuine expertise to every transaction matters. As our profession navigates a period of real change, these things are worth saying clearly and often.
The technology will continue to evolve, but our obligations will not.