There’s a particular kind of cognitive dissonance that sets in when you’re staring at two very different stories in the same spreadsheet.
The first story, as told by freight rate indices and capacity reports heading into 2025’s peak season, suggests that everything is normalizing. Major import lanes have loosened up, surcharges are easing, and volume growth is modest at best. The supply chain, at least according to the aggregate data, is exhaling.
The second story lives in your P&L, and it’s considerably less relaxing. Supply chain costs as a percentage of sales continue to climb into the teens, sometimes reaching the low twenties, well above the single-digit benchmarks that were once considered gospel.
So while the market says things are cooling off, your margins say otherwise. You might be operating in a “normal” demand environment now, but your cost baseline never reverted. Unless you address those structural cost drivers, you’ll watch your margins erode even as everyone assures you the worst is over.
Pre-pandemic, most companies ran logistics at around 6% of revenue. Today, some firms are seeing 12-20%, and analysts think it could climb to 25%. This isn’t temporary. The floor moved.
Then there are the costs that sneak up on you. Detention and demurrage charges hit $10 billion globally in 2024. At some ports, you’re paying over $300 per container per day once you blow past the first 10 days. You can’t negotiate these away because they’re tied to how congested the network is, not how good your carrier relationships are.
Container volumes on the Asia-to-US trade lane have dropped back to something resembling normal. Free time windows haven’t expanded to match. Drayage capacity is still tight. The operational bottlenecks that generate those penalty charges are still there, even though you’re moving less freight than you were in 2021.
You might be booking fewer TEUs than you were in 2021, but you still need drivers, warehouse staff, planners, brokers, depot coordinators. Your workforce doesn’t scale down as smoothly as your shipment count. Add wage inflation and a shrinking driver pool, and you’ve got fixed costs that don’t care about softer demand.
Then there’s what you might call the complexity tax. Global sourcing means juggling multiple entry ports, inland transit routes, drayage appointments, customs holds, free-time windows, container repositioning. Each node in that chain adds another handoff, another system, another opportunity for manual intervention. Manual intervention tends to cost more per unit. When volumes decline, you can’t spread those fixed costs as efficiently, so the per-unit economics get worse.
Finally, there are the costs you don’t see until they’ve already happened. Free time expires while you’re waiting on a drayage appointment. Detention piles up because a container sat too long at a congested depot. Customs pulls a shipment for inspection. These charges don’t shrink in a soft market. They get more expensive because you’re squeezing margin everywhere else and nobody has slack in the system to absorb them.
This is where most companies reach for AI. More dashboards, better predictions, smarter algorithms. Some of that helps. But if you think better visibility alone is going to fix a structural cost problem, you’re about to be disappointed.
By now, you’ve probably sat through at least a dozen vendor pitches promising that AI will solve all of your supply chain problems. While it all sounds great, and some of it even is great, the problem is that most of what’s being sold as “AI for supply chain” is point solutions. They optimize one piece of the puzzle (forecasting, networking optimization, inventory positioning) without fundamentally changing how decisions flow through your network. They make individual processes faster, but they don’t necessarily make your overall operation smarter or more coordinated.
Imagine a large electronics importer that recently implemented an AI-powered routing engine for their Asia-US shipments. It worked beautifully: shaved two days off transit times, improved arrival forecasts. But downstream, nothing changed. They still maintained buffer inventory. Still dual-sourced to hedge risk. Still deferred container returns when depots became congested, which meant free time expired more often and demurrage charges climbed. The AI delivered on its promise (faster, more accurate routing), but the cost per unit moved went up because the system around the AI didn’t change.
While insights are getting better, the problem is that insight without orchestration is just expensive trivia.
If your AI sits on top of fragmented data, disconnected systems, and manual handoffs, it’s going to optimize locally while running into the same execution problems you had before.
To truly root out costs, the system’s architecture, not the algorithm, is where you must start. You need real-time visibility across your entire network, and you need it connected to actual execution systems (carriers, warehouses, customs brokers, 3PLs) so that when the AI spots a problem, it can do something about it. And you need exception management that doesn’t require three emails, two Slack threads, and a phone call to resolve.
This is where the concept of an intelligent control tower, when built correctly, can make a difference. Not because it has fancier algorithms than the point solutions (though it might), but because it’s designed from the ground up to orchestrate across the entire network. It sees the whole picture, connects to the systems that can actually change outcomes, and resolves exceptions without requiring constant human intervention.
To drive down logistics costs you need to unify data, automate execution, and orchestrate decisions across every node in the network. The AI matters, but only when it’s operating on a foundation that can act on what it knows.
We’ve moved past the crisis phase of global logistics into something that looks, on the surface, like normalcy. But the cost structure hasn’t normalized. It’s reset at a higher baseline. If you’re still approaching your supply chain like it’s 2019 (negotiating freight rates, managing carrier relationships, hoping volume returns), then you’re optimizing for the wrong variables.
The real challenge has moved from visibility to orchestration — connecting what you see to what you can do, and doing it without requiring your team to manually intervene on every exception. The AI matters, but only when it’s sitting on infrastructure that can execute across your entire network.
That means unified, real-time data across the full spectrum of logistics: ocean, air, parcel, LTL, FTL. It means connecting that visibility to execution through integrated carrier APIs and automated workflow engines. It means exception management that doesn’t live in email threads and Slack channels. When a system can spot that a container is going to miss its free-time window and automatically rebook drayage, notify your network, and adjust your downstream receiving schedule, you’re no longer just faster at firefighting. You’re preventing fires.
Platforms built this way – FourKites being one example – matter in ways that traditional visibility tools don’t. Not because they have shinier dashboards, but because they’re designed to orchestrate, not just observe. The companies bending their cost curves right now are doing it with better infrastructure that lets their AI analyze and take action.
Ready to take action?
