After People, Cargo — and After Cargo, People Again
In the last installment, we talked about resume matching and onboarding automation. Once you've sorted out where people come into the company, the next question that naturally follows is, "What about the cargo those people used to move — how is it flowing?" The fact that logistics is the first place the cry of "we don't have enough people" goes up — in Korea and in Japan alike — barely needs statistics to prove.
That said, talk of "transforming logistics wholesale with AI" doesn't really belong here. Somewhere along the pickup → tracking → customs pipeline, things break down — and almost always in the same spot. What you slot into that broken spot, and how, is what decides whether your automation takes root or quietly disappears a year later.
Why You Should Start with "Where It Breaks"
According to industry reports, Japan is projected to see a 27% drop in drivers by 2030, with as much as 36% of all freight potentially going unmoved. 36% sounds small, but it means one in three jobs simply stops. Few would confidently bet that Korea won't follow the same curve five to ten years later.
What's interesting is that this pressure doesn't translate straight into "let's automate the entire pipeline." Last month I spent nearly two hours with the CEO of a 3PL operator, and the first thing he said as he sat down was, "We already have a routing solution in place. The problem is what comes after." On his monitor, a single tracking number was frozen in red. The truck had left, but no one had filled in who was supposed to update the ETA for the shipper.
I initially assumed "just plug in some AI routing and you're done." But once I dug in, I realized routing is already installed at most decent-sized companies. The breaks happen between routing systems — between systems and people, between a Korean office and a Japanese customs counter, in the "margins." That's why the industry keeps reiterating that "before pouring money into reckless automation equipment, you should first audit and optimize your warehouse operations."
Slicing the Pipeline into Three Sections
On the ground, we usually break the pipeline into three sections to find where it breaks: pickup, tracking, and customs/paperwork.
1) Pickup — Are Handwritten Notes Still on the Dispatch Board?
As SMB-oriented last-mile tools like Onfleet and Hemut have become commonplace, handling routing, driver assignment, and customer notifications from a single screen has become the standard. But "standard" and "actually adopted" aren't the same thing. Plenty of companies still have a stack of paper memos sitting next to the dispatcher's keyboard.
The reason these notes don't go away is simple. A shipper drops "Can I send one more in 30 minutes?" into KakaoTalk, and there's nowhere in the system for that input to land. The heart of pickup automation isn't the routing algorithm — it's a small adapter that funnels these informal-channel inputs into the system, even as a "stub order."
2) Tracking — The Blank Space Is "Who" the Alert Should Reach
How far large platforms like Hyundai Glovis's ORCA and GVIS have pushed real-time visibility has been covered enough in the news. They automatically calculate alternative shipping routes for shocks like weather or geopolitical disruption. Small and mid-sized shippers can't lift these systems wholesale, but they can borrow the thinking.
"Where is the cargo?" has been more or less solved over the past decade. What hasn't been solved is "who should this information go to, and in what tone?" Even a 30-minute delay calls for a short note to a regular shipper, while a new shipper needs an apology paired with a revised ETA. The global trend is to move this judgment out of human hands and into "agentic" automation. Where simple RPA handled "rule-defined tasks at speed," this layer also decides "for this shipment, who gets what message, and when." That difference started showing up on SMB floors last year.
3) Customs and Paperwork — So a Human Doesn't Have to Check Every Time
Customs is the most nerve-wracking stretch for the trading companies and forwarders operating between Korea and Japan. Different regulations apply by lane and by SKU, and one blank field on a form can leave a container stuck for days. Anyone who has watched the overnight customs queue knows exactly how heavy that cost is.
The meaningful change of the past year or two is that attempts to slot LLMs into customs ETA prediction, AP matching, and document review have moved beyond the "demo" stage and into operation. The point isn't that the LLM replaces a customs broker's judgment — it's that the human no longer has to manually recall how the same lane has been handled over the past six months. In the automation projects we've delivered at 5years+, customs work almost always enters as "review assistance" rather than "full automation." Adoption rates are overwhelmingly higher that way.
So Where Should You Actually Start?
Reports that Amazon's Chiba Minato fulfillment center ships 600,000 items in two to three hours per day have been cited often since last year. It's impressive, but to be candid, the number doesn't help SMB decision-making much in Korea or Japan. It's a useful "this is what fully automated looks like" reference point, but the starting line is too different.
In practice, three questions are usually enough. First, where in your company is "the same person sitting in front of the same screen at the same time every week"? Second, where are messages between shippers, warehouses, and carriers scattered across KakaoTalk, phone calls, and email? Third, what judgment calls grind to a halt the moment someone steps away from their desk? Companies that can answer these three quickly are already half-finished before they even pick an automation tool.
For context, the cleanup of "where people come in" we covered in the previous installment on HR automation and the cleanup of "where cargo flows" in this one are not separate efforts inside the same company. The permissions and onboarding data the HR side has organized need to flow straight into the access rights on the dispatch screen and the user accounts on the customs system — only then do small overnight decisions, like who updates an ETA memo at 2am, keep getting handed off cleanly as shifts change.
Closing Thoughts
Logistics automation doesn't end at "we adopted AI routing once." It's the work of filling, square by square, the blank margins between pickup, tracking, and customs — and the best tool for finding those blanks is still asking "who got stuck where last week?"
At 5years+, we've worked with small and mid-sized businesses across Korea and Japan to roll out automation starting from small PoC units. If you'd like to map out where your company's pipeline is breaking down, feel free to reach out via our Free consultation. If you need prior case studies to review, we'll prepare those alongside.
Next time, we'll talk about the flow of money that follows once the cargo gets moving. From invoice processing to closing the books, we'll map out how far the human hand can step back across tax and accounting.