Airwaybills and the volume of documents and data that air freight produces is well-known. Digitalisation and technological capabilities are shaping the way that customs declarations and these clearances are passed, from going paperless to deploying automation to speed up the process.

“A lot of time is still lost on manual work: searching for the right data across invoices, packing lists and air waybills, copying it into systems and structuring it into a usable declaration. Documents arrive as PDFs, scans and emails, each in a different layout, so every shipment starts with the same manual work. And in airfreight there is little time to spare, especially now that ICS2 requires good data before loading,” Wim Van Emelen, CCO at Customaite explained.

The question is, how is artificial intelligence (AI) now entering the ring, and how is it affecting the current automative systems that are already deployed across custom declarants’ operations?

The next step

Data is defined by its structure, and hence the majority of tools automation are able to extrapolate this in the format it has been provided. “Traditional automation works well when data is structured and predictable: fixed templates, EDI messages, clear rules. The problem in airfreight is that most input isn’t like that. Every shipper sends a different invoice, descriptions are vague, and information is spread across several documents,” Van Emelen continued.

The news coming from providers such as Customaite is, however, that the deployment of AI can bolster automation processes by standardising the data, interpreting the data regardless of language and other formatting barriers and flag when documents such as when the airwaybill and invoice don’t match.

“It changes the role of the people involved. Instead of searching, copying and typing, they review and decide. In practice we see declarations that used to take hours, now prepared in a few minutes, with the expert staying in control,” Van Emelen continued.

This argumentation for AI to adopt a greater role within these processes boils down to a widely debated topic, of customs professionals can adapt to the greater pressure of volumes generated by eCommerce and other air freight. It revolves around safe and precise exception management as hours are saved from time intensive and laborious data entry or duplication.

Bridging the gap

“What AI shouldn’t do is hide the gaps. If a goods description is too vague to classify, or the weight on the invoice doesn’t match the air waybill, the system should flag it, not guess. That’s where customs knowledge matters: knowing which data is critical for a correct declaration, and which inconsistencies actually carry risk. What AI doesn’t replace is accountability. The declarant still signs off. Good AI makes that job faster and more reliable, and shows its reasoning so people can trust the result.” Van Emelen stated.

Knowing AI’s role in the process and training teams to understand this is just as critical as correctly introducing it into the workflow. Customaite are clear that it cannot operate in isolation, but it can benefit those using it within their operations. It also consolidates information for professionals who face ‘messy’ or incomplete data that can be detrimental to their output.

“AI doesn’t remove the data quality problem, but it changes how you deal with it. Instead of people checking everything line by line, they get a clear view of what is missing or doesn’t add up. Over time that also helps improve data quality at the source, because you can show shippers exactly where their information falls short,” Van Emelen said.

“The biggest difference comes from connecting the steps. Many tools focus on one part: document processing, or classification, or a compliance check. These steps depend on each other. A tariff code is only as good as the goods description it’s based on, and a compliance check only works if the data behind it is correct and complete.”

Even though the ONErecord system has sought to standardise data across the air freight supply chain, complete adoption and compliance within this framework is currently not an operational reality.

Barriers to cooperation

“All of those play a role, but one barrier is often underestimated: the reluctance to share data. Air cargo runs on cooperation between shippers, forwarders, airlines, handlers and authorities, yet many parties still treat their data as something to protect rather than something to connect,” Van Emelen commented.

Data is of course commercially sensitive. Whether it is in products, shipments, or competitors milestone checks in cargo handling, there is often something that can be gleaned by competitors. There is also the desire from companies to maintain sole control of their data, and avoid it leaving their systems where possible.

“The result is that every party keeps its own version of the truth, and the same data is retyped and checked again at every step. That’s exactly the inefficiency digitalisation should remove. Breaking through this requires trust and clear rules. Companies need to know where their data is stored, who can access it and what it is used for,” Van Emelen answered.

“Technology providers have to be transparent about that and offer guarantees, such as data staying within the EU, certifications like ISO 27001 and no use of customer data for other purposes without consent. Sharing data shouldn’t mean giving it away. It should mean sharing the right data, with the right party, for a clear purpose.”

Outlook for AI in airfreight

What does this look like in real terms for a customs declarant? The eventual goal would appear to be managing by exception, the removal of the routine and exhaustive by ensuring customs experts’ time is efficiently deployed working on the dilemmas that necessitate greater care and attention.

“With AI, routine shipments can be prepared and checked automatically, so people only step in where something doesn’t add up: a missing certificate, an unclear goods description, a value that doesn’t match. That’s managing by exception, and it fits airfreight well because it puts expert time where it matters and keeps routine cargo moving,” explained Van Emelen.

“Two things are needed to make it work. First, the system has to be good at knowing what it doesn’t know. An exception model only works if the right cases are flagged and nothing risky slips through as “routine”. Second, people need to trust it. That comes from transparency: showing why something was approved or flagged, and letting teams build confidence step by step.”
Declarants do not disappear in this new proposed system, but the role would evolve from processing to supervising and deciding.

The post Bridging the gap between automation and interpretation appeared first on Air Cargo Week.