All professions

Business and operations

Logistics and Supply Chain

The problem

A declaration prepared with an assistant looks complete, but has a VAT number in the wrong format, a four-digit commodity code instead of six and a goods description that is too short. The import control system sends it back flagged.

30 courses · 5 specific to this profession

A documented failure pattern

The arrival date: the tool's, not the supplier's

  1. What the tool doesA status notice for an import shipment, which an AI tool drafts from the supplier's email for a logistics employee, who will send it to the client.
  2. Where the tool goes wrongThe notice fills in arrival and clearance dates that the supplier's email never gives.
  3. The verification stepWrite each date's source next to it; mark a date without one “not confirmed”.
Source: Scientific Reports, 2025, and 2 more.
More: why it is easy to miss and the full check

Where the tool goes wrong, and why it is easy to miss

The notice template has three date fields (loading, arrival, clearance) and the tool fills all three, although the email gives only one date and says outright that the arrival is not yet known. The added dates do not look invented: they fall after the loading date, at intervals that resemble an ordinary truck and ferry trip, and they are written in the same format as the real date. Anyone who reads only the notice no longer sees the sentence "the carrier will confirm", so the client plans unloading and sales around a date nobody has given.

The check that catches it

  1. Before sending the notice, write next to each date where it comes from: the supplier's email, the carrier's confirmation or the ferry booking.
  2. Count the dates in the email and the dates in the notice. Every extra date comes from the tool, not the supplier: find its source or remove it.
  3. Mark every field without a source "not confirmed" and say who gives the date and when, for example "the carrier, after booking the ferry".
  4. In your request to the tool, say that any field the document does not cover must stay empty, and check in the draft that it did.

Where the failure pattern is documented: A study in Scientific Reports on extracting data from logistics documents (August 2025); a study on arXiv by OpenAI and Georgia Tech researchers (September 2025); a JPMorgan AI Research study on arXiv (October 2024).Sources: Scientific Reports, 2025; arXiv, 2025; JPMorgan AI Research, on arXiv, 2024.The case above is ours, written for work in Albania; the laws it names were checked against the official text.

What it looks like

The supplier's emailThe pallets are loaded onto the truck on 14 October in Verona. The carrier will confirm the arrival date in Durrës once it has booked the ferry.

The AI tool's draft notice to the clientShipment status: loaded on 14 October in Verona. Expected arrival in Durrës: 18 October. Customs clearance: 20 October.

Translated from the Albanian original.

Specialist courses for this profession

What you leave with

The declaration trail, completed by the second declarant before submission: each field compared with the source document, each code with the heading in the official nomenclature, each rule the tool added removed, and a named confirmation. A client or an auditor finds there which declaration each code was used in. The programme teaches you to prepare the carrier's document pack and cross-border correspondence with AI, and to verify them before every submission.

Courses that apply to every profession

25 more courses are the same for every field: the foundations of AI, checking its answers, agents, and what to do when you get stuck.

Logistics and Supply Chain

The catalogue lists all 30 together, with filters for format and time.