Port Klang handled a record 15.14 million twenty-foot equivalent units in 2025, up from 14.64 million the year before, according to the Transport Minister. It was the highest throughput since the port began operating in 1973. The Port of Tanjung Pelepas grew faster still, up 14% to 14.03 million TEUs. Port Klang has entered the ranks of the world’s ten busiest container ports.
More capacity is on the way. Westports began dredging and reclamation in 2025 for eight new container terminals.
Cranes and berths are the visible constraint. The invisible one is documents.
A container is a stack of paperwork
A single shipment generates a bill of lading, a commercial invoice, a packing list, a customs declaration, permits, a delivery order and a proof of delivery. These pass between shipper, forwarder, haulier, warehouse, customs broker and consignee. At most handovers, someone reads a document and types it into their own system.
That retyping is where delays and errors enter. A container held because the weight on the packing list does not match the declaration. A delivery that cannot be invoiced because the signed proof is in a driver’s cab.
For the forwarders, hauliers, warehouses and distributors behind the ports, most of them small and medium businesses, this is the daily cost. And it is the part of logistics where AI is already reliable.
Three uses that work today
1. Reading shipping documents into your system. Language models are good at pulling structured fields out of messy documents: shipper, consignee, container number, weights, HS codes, incoterms. We use this technique in production for a different industry, turning PDFs, slides, spreadsheets and photographed pages into structured records, where it cut manual entry and review by 76%. The rule that makes it safe is the same in any sector: the model fills a strict schema, and a person confirms before the record is used.
2. Checking documents against each other. Does the invoice total match the packing list quantities? Does the declared weight match the bill of lading? This cross-checking is tedious for people and trivial for software once the fields are extracted. Mismatches are caught before submission instead of at the gate.
3. Recording handovers with proof. A QR code on a consignment, scanned at pick-up, at the warehouse and at delivery, with a photo and a timestamp. This is not AI. It is the foundation that makes everything else trustworthy. We built exactly this chain for a manufacturer: scan at the point of work, link to a delivery order, and issue a document a third party can verify by scanning.
What to leave for later
Route optimisation, demand forecasting and dynamic pricing all work, and all depend on clean historical data. If your shipment history is in spreadsheets and email, start with the three uses above. They create the data the advanced tools need.
What this means if you run a logistics business
- Find the retyping. Wherever a person copies from a document into a system, a model can do the first pass.
- Keep a person on the final check. Customs declarations carry legal liability. AI drafts, a qualified person submits.
- Give customers the answer before they ask. A portal with status, documents and proof of delivery removes most “where is my shipment” calls.
- Own your data. Tracking history is an asset. Keep it in a system and an account you control.
The ports are scaling with concrete and cranes. The businesses around them scale by no longer typing the same shipment five times.