AI Is Thinning the Interface, Thickening the Experience in Global Trade

The Interface Is Thinning, the Experience Is Thickening

For decades, import/export software meant dense dashboards, endless forms, and a maze of menus. You had to learn the system before you could move a container. But that’s changing fast. Large language models and automated agents are rewriting how trade professionals interact with digital tools. Instead of clicking through screens, you might just type, “Check if my shipment from Shanghai cleared customs.”

The interface—the buttons, the tabs, the data-entry grids—is getting thinner. AI can generate its own UI on the fly or skip the UI entirely. But the experience, the actual feeling of control and confidence, is getting thicker. Because when AI starts acting on your behalf, the design challenge shifts from “where do I click?” to “what is it doing, and can I trust it?”

From Human-Finds-Function to AI-Understands-Intent

Old software had a basic assumption: the human must understand the system before using it. You needed to know that the “Bill of Lading” button lived under the “Shipping” tab. You memorized the path. That’s why so much UX work went into reducing clicks, flattening hierarchies, and making navigation feel intuitive.

AI flips that. Now the system tries to understand you first. You don’t need to know the exact name of a customs broker API or the right form for a certificate of origin. You just say what you want. But this creates a new kind of friction: the cost of being misunderstood. If the AI thinks “handle this shipment” means “file the export declaration” when you actually meant “book the vessel,” you’ve got a problem.

That’s why experience design in trade software now includes what some call intent design. It’s not just about flows anymore. It’s about making sure the machine’s interpretation matches your intention—and making that interpretation visible early, so you can correct it before it’s too late.

Fewer Pages, More Rules

Here’s a trap: when AI shrinks the number of screens, it’s tempting to think the design work is done. But most of the experience just moves from visible pixels to invisible system behavior. For example, a user says, “Process these invoices.” Does the AI just suggest a payment schedule? Or does it actually send money to a supplier in Vietnam, update the ERP, and trigger a bank transfer? That’s a huge difference.

The real experience now depends on questions like: When should the AI act without asking? When should it pause and confirm? What can it decide on its own? What must it never touch without explicit approval? How does it tell you what it just did? Can you undo it? When should it stop and hand control back to a human?

These aren’t static-page questions. They’re behavioral rules. So while the interface gets thinner, the rulebook gets thicker. That’s what “experience thickening” means in practice.

From Usable to Delegable

For years, the gold standard was usability. Can the user find the function? Is the flow smooth? Can they complete the task without getting lost? That still matters. But when AI starts acting on your behalf, a bigger question emerges: Would I actually trust this thing to do the job?

Call it delegability. An AI can be brilliant—fast, accurate, efficient—yet still not worth handing your shipment to. You worry: Does it really understand my Incoterms? Will it make a risky choice on its own? Can I see what it did? If it messes up, is there a way back?

In import/export, the stakes are high. A wrong tariff code or a missed document can cost thousands or delay a cargo. So the experience design goal shifts from “making users able” to “making users willing.” Intelligence determines how far the AI could go. Experience design determines how far the user will let it go.

Sometimes the Right Move Is to Ask One More Question

Classic UX says shorten the path. Remove the extra click, the confirmation dialog, the redundant step. Efficiency is king. But in AI-driven trade software, that principle breaks down.

Imagine a user types, “Delete these old shipment records.” If the AI immediately deletes them, it’s technically efficient. But it might also be dangerous. Maybe those records are needed for an audit. Maybe “these” referred to a different folder. In such cases, the better experience isn’t fewer steps—it’s the right pause.

That’s boundary design. It’s not about what the AI can do, but what it should do, and where it must stop. As models get more capable, the hard part isn’t “can it?” but “should it?”—and that’s a design decision, not an engineering one.

Designing AI Behavior Like a Director

Think of the old interface design as building a stage: you arrange the entrances, the paths, the lighting. But AI experience design is more like directing an actor. The AI has to know when to speak, when to stay silent, when to make a suggestion, when to act, when to check in, when to admit uncertainty, and when to step aside and let a human decide.

This is often called AI behavior design. In trade software, it might mean the AI proactively flags a potential customs delay—but doesn’t reroute the shipment without asking. It might suggest a cheaper freight option, but only after you’ve approved the budget. The design target isn’t a page; it’s a performance. And the script is the behavioral rules that govern how the AI behaves in each scenario.

Setting Expectations Before Action

Traditional software is predictable. Click “Download,” and you know what happens. Click “Submit,” and you know the system will process the data. With AI, you often can’t predict: Is it just suggesting, or about to act? Will it do one step or ten? Will it access other data? Will it change a system state?

That’s why expectation design matters. A good AI experience doesn’t need to explain itself constantly. It just needs to give you a mental model before action—like “I’m about to file three documents and notify the broker.” And after action, it should confirm what it did. When the system’s behavior gets complex, building the right expectations is a core design skill.

Reversibility Might Matter More Than Intelligence

Why are people hesitant to let AI handle their trade operations? Often not because it’s dumb, but because they can’t see a way back if it goes wrong. That’s why reversibility is a crucial metric.

Can you undo a generated document? Can you restore a modified record? Can you confirm before sending an email to customs? Can you see an audit trail of every automated action? Can you stop a multi-step workflow mid-way? Can you hand back to a human if something looks off?

These aren’t flashy AI features. But they’re the difference between a tool people use and a tool people trust. A trustworthy AI doesn’t just do things—it lets you change your mind.

From UI Standards to Experience Governance

In the past, enterprise UX focused on interface consistency: same colors, same components, same interactions across products. That’s still important. But as AI spreads into trade operations, a new kind of consistency is needed.

Do all AI features use the same confirmation mechanism? Do different skills have clear permission boundaries? Is there a uniform risk warning for sensitive actions? When AI fails, is there a standard human takeover procedure? Are results verifiable, traceable, and undoable?

This is experience governance. Instead of standardizing how the interface looks, you’re standardizing how the intelligence behaves. In import/export, that might mean: every AI action that sends money requires a two-step approval. Every change to a shipment’s routing must be logged. Every failure offers a “human override” button. These rules live beyond any UI kit.

Design Value Isn’t Disappearing—It’s Moving

AI will absolutely reduce some traditional design work. Standard pages, repetitive visuals, basic prototypes—those are getting easier to automate. That’s fine. The real question isn’t “how many jobs will disappear?” but “where do new experience problems appear?”

From what I see, the shift is clear: from pages to intent, from operations to behavior, from efficiency to boundaries, from usability to delegability, and from interface consistency to behavioral consistency. The organization that thrives isn’t the one with the most UI designers. It’s the one that can turn increasingly powerful AI into experiences that are coherent, understandable, controllable, and worthy of trust.

Shaping Trustworthy Intelligence

If you think design is about making things look nice, AI is indeed eating the job. But if you think of experience design as deliberately shaping the relationship between people and systems, then AI isn’t killing the practice—it’s expanding the problem space.

We used to design how people operate software. Now we design how software understands people. Next, we’ll design how people and AI work together. The real deliverables aren’t buttons or screens. They’re understanding, expectation, boundary, action, feedback, reversibility, and trust.

In import/export, where a single mistake can ripple across borders, that trust is everything. The future of experience design isn’t just simplifying complex tools. It’s shaping increasingly powerful intelligence into something people can understand, control, and—when it earns it—delegate to.