Vopak is looking for a Head of Customs.
Let’s get straight to the point. Vopak is not a garage startup. It’s a global titan in storing everything that flows, burns, and evaporates: oil, chemicals, gases. A company with vast infrastructure, incredibly complex processes, and a reputation built over decades. Now, like all giants, they are looking to the future: hydrogen, ammonia, green energy. Very solid.
And this giant needs a person. One person for the “Head of Customs” position, who must become a living bridge between the dry letter of the law, roaring logistical operations, and the silent code of IT systems. They must develop and control customs policy across all of Belgium, be the chief negotiator with the stern Belgian customs authorities, train teams, manage declarants, and, the cherry on top, lay the tracks for customs clearance of future products, like hydrogen, which no one has really traveled yet. Essentially, Vopak isn’t just looking for an employee. They are looking for a walking supercomputer with diplomatic skills and the gift of foresight. The pain they are trying to alleviate is colossal risk. The risk of errors in declarations, multi-million euro fines, loss of AEO certificates, and, consequently, business downtime. They want to plug this hole with one very experienced and very expensive person.
Now, let’s pour ourselves some coffee and talk like adults from IT. What if I told you that 90% of this job is not for a human, but for a neural network? Instead of searching for a “unicorn” who remembers all articles of the EU Customs Code and can also charm an inspector, we could build a system that does it more reliably, faster, and 24/7.
Imagine a dialogue in a Vopak meeting room. On one side – a hypothetical Petrovich, Director of Operations, old guard. On the other – Max, IT Director.
Petrovich: Max, this is customs! This isn’t like redesigning a website. This requires experience, intuition, personal contacts! Your machine won’t call Inspector Janssens and resolve an issue humanly!
Max: Petrovich, I agree. For calling Janssens, we’ll keep one high-class specialist. But tell me, what percentage of time will your ideal “Head of Customs” spend calling Janssens, and how much will they spend sifting through thousands of pages of laws, checking hundreds of declarations a day, and trying to predict how a new EU emissions regulation will affect ammonia imports in three years? 99% of their job is data analysis. Artificial intelligence was created for this. We’re not firing people; we’re giving them an exoskeleton.
How would this look in practice?
1. Creation of a “Unified Customs Brain.” This isn’t science fiction; it’s Retrieval-Augmented Generation (RAG) technology. We take a large language model (e.g., a fine-tuned Llama 3 or GPT-4) and “feed” it all relevant information: the complete EU Customs Code (DWU), all Belgian laws and regulations, Vopak’s internal regulations, their entire history of customs declarations for the past 10 years, including all errors and disputed cases. This system becomes the single source of truth.
2. Proactive monitoring system. An AI agent 24/7 scans all official government and European portals for legislative updates. As soon as a new document appears, the system doesn’t just report “a new law has been released.” It analyzes it, compares it with Vopak’s current operations, and provides a summary: “Attention, as of July 1st, the excise duty on product X changes. This will affect our deliveries at terminals A and B. It is recommended to update declaration template Z. Projected cost increase – 2%.” Your Head of Customs is still asleep, and the system has already prepared an action plan.
3. Interactive advisor for everyone. Instead of a sales manager running to the Head of Customs with the question “What documents are needed to import a new batch of chemicals from Norway?”, they simply type into the corporate chatbot: “We plan to import 100 tons of product Y, HS code [such-and-such], from Norway to Antwerp. Provide a complete list of required licenses, certificates, and describe the customs clearance procedure.” The AI instantly provides an exhaustive checklist with links to specific articles of law. The Head of Customs ceases to be a “bottleneck.”
4. Automatic audit and risk management. Most importantly. Before sending any declaration, the human declarant uploads it to the system. The AI instantly checks it against thousands of rules, finds inconsistencies, potentially risky formulations, or suboptimal codes. It doesn’t just say “error”; it explains: “Product code 12345 is incorrect. Based on the product composition, code 12346 would be more accurate, which would also reduce the duty by 0.5%. Source: EU Regulation No…” This reduces human error almost to zero and guarantees the preservation of AEO certificates.
How to overcome distrust? Petrovich is right; people don’t trust a black box. Therefore, implementation should be phased.
Step 1: “Shadow” mode. We run the system in parallel with human work. It analyzes everything but does nothing, only provides recommendations. At the end of the month, we compare: how many errors that humans missed did the AI find? How much time would the AI have saved?
Step 2: “Assistant” mode. AI becomes a mandatory tool for double-checking. No declaration leaves without its “approval.” People still make the final decision, but based on data from the AI.
Step 3: “Autopilot” mode. Routine, low-risk operations are fully automated under human supervision.
How to validate the result? It’s simple. AI is not magic; it’s mathematics.
Firstly, traceability. Every recommendation the AI makes must be supported by a reference to the source – a specific article of law, a point in a regulation, or a precedent from a past declaration. No “I feel like this.”
Secondly, A/B testing on historical data. We take declarations from last year where errors were made and fines were incurred. We “run” them through the system. Did it find these errors? Did it suggest the correct solution?
Thirdly, key performance indicators. Reduction in the number of errors in declarations by X%. Acceleration of the clearance process by Y%. Savings on fines and duties of Z euros. Numbers don’t lie.
Vopak is looking for a person who should work like an ideal algorithm. But perhaps, in 2024, it’s time to hire the algorithm itself and leave humans to do what they do best: make strategic decisions, conduct complex negotiations, and manage relationships. And delegate the routine, superhumanly complex data analysis work to what was created for it.
Источник: https://www.linkedin.com/jobs/view/4407863573/
