From Version 2 to Version 31: What 2025 Did to Global Trade

In a normal year, the US Harmonized Tariff Schedule gets one new version, sometimes two. By the end of 2025, it had reached version 31.

That number, surfaced by Gaia Dynamics CSCO Tom Gould on the first episode of the Trade and Tech podcast, captures why the year felt unlike anything industry veterans had seen before. Average US tariff rates climbed from 2.5% to 18.6%. Decisions announced on Friday could take effect by Monday morning. Small importers who had never thought about a customs broker were suddenly on the phone with one.

On the episode, Gould and Gaia CEO Emil Stefanutti sat down with Bojan Strbanovic, global director of trade at Logistics Plus, and Steve Bozicevic, who leads three cross-border logistics and trade-tech companies. Together, they mapped what 2025 actually did to global trade and where the people inside it think 2026 is headed next.

What Made 2025 Unlike Any Year in Modern Trade?

Strbanovic remembers when HTS books arrived twice a year.

“January, we would get brand new books and we would put them on our desks. They would be nice and you can smell the new paper. Then comes June, July, there will be inserts that were blue color.”

That old cadence was manageable. Changes came in batches. People could breathe between them. In 2025, that rhythm collapsed. A regulator published 31 distinct schedule revisions in 12 months, which is the sort of number that does not merely signal change. It announces a new climate.

The real disruption was speed. Gould described a new tempo where tariff actions were announced on Friday and live by Monday, often with no Federal Register notice in between. Bozicevic confirmed the pressure from the supply side.

“It’s like the tariff gets announced on this day, but doesn’t take effect until this day if you’ve already shipped it. So you’re kind of in the clear and then everyone’s like, well, how do I clear this in bond? Because I don’t know what’s going to happen next week.”

The financial impact moved just as fast. Bozicevic cited a widely circulated late-2025 analysis, though exact numbers vary by methodology and product mix, showing the average effective US tariff rate rose from 2.5% in 2024 to 18.6% in 2025. That kind of jump does not stay abstract for long. It shows up in renegotiated supplier contracts, in landed cost calculations, and in hard conversations about who absorbs the hit.

As Bozicevic put it, the money has to come from somewhere. It comes from profits, tougher supplier negotiations, or higher prices.

Strbanovic captured the human side of the year.

“I’ve seen customers, really, lack of better term, freaking out. They’re getting hit by different opinions on what needs to be done. They’re surprised, and they’re certainly not ready.”

That mix of pace, magnitude, and policy delivered through tweet-length announcements made 2025 historically unusual. For trade compliance teams, the year did not unfold. It ambushed them.

For the AI tools meant to help them keep up, the bar moved from useful to load-bearing infrastructure almost overnight.

Why Customs Brokerage Stopped Being Invisible

For decades, customs brokerage was invisible work. Most people outside the trade world did not notice it unless something went wrong.

Bozicevic put it bluntly.

“A year ago, most people didn’t know what a customs broker was or what they did. Most US receivers weren’t the importer of record. They didn’t know how customs work.”

Strbanovic told a story from his UPS years that made the point sting. In a global meeting, one of UPS’s biggest customers complained about how the carrier handled its international business. The customer framed the issue as a massive fee problem, but the numbers revealed a deeper misunderstanding. The CFO was treating duties and services as one bucket. The brokerage and compliance work that shaped the duty outcome was effectively invisible in the P and L.

For some industries, that invisibility lasted because brokerage looked commoditized. Brokers were competing on price, while the strategic value of classification, country of origin, and tariff planning rarely reached the executive table.

That changed in 2025.

Bozicevic said US importers suddenly understood that the real questions were not only about supplier pricing or freight rates. They were also asking whether there was a better country to make a product in, or better materials to use. Strbanovic agreed that the year finally exposed the value of the work brokers and trade consultants had been doing quietly for decades.

The spotlight came with strain attached. Brokers were flooded with calls from clients who suddenly needed to understand work they had outsourced for years.

Which is why automation moved from a nice-to-have to a survival tool.

Who Absorbed the Hit, and Who Did Not?

The pain did not land evenly.

Importers with paperwork in order

Large importers with USMCA or CUSMA certificates already in shape had a better year than the headlines suggested. Bozicevic described the moment plainly.

“When I get called into the principal’s office by customs, I can have the proof points in my supply chain and my rules of origin. I was fine.”

The first few months still brought panic, especially from February through May. But once free-trade-eligible importers checked the paperwork and saw it held up, the picture stabilized.

SMBs and de minimis sellers

Small and mid-sized businesses had the roughest ride. They had the least leverage with suppliers, the least internal expertise, and the least attention from brokers under pressure.

“Who am I going to focus on, the million-dollar client or the ten-thousand-dollar client?” Bozicevic said.

The end of de minimis treatment for low-value shipments hurt a whole layer of cross-border ecommerce sellers on both sides of the US-Canada border. Some businesses folded. Others rebuilt around higher-value bundling or moved sourcing.

There were winners too. Canadian manufacturers benefited from steeper tariffs on Asian-origin goods, which suddenly made their products more competitive in the US market.

The pattern was uneven, but it was not random. Those who adapted early and kept documentation discipline came through better. Those who had not built a customs function before 2025 found themselves assembling one in real time, under pressure, with margins shrinking around them.

Where AI Fit in 2025, and Where It Did Not

The AI conversation in trade split into two questions. The first was what regulators would allow. The second was what practitioners actually needed.

The human in the loop still matters

On the regulator side, Gould was direct.

“We’re still a ways away from the point in time when customs will allow customs brokers to do the customs business that they’ve been doing traditionally without a human in the loop somewhere.”

That is not a temporary inconvenience. In Gould’s view, it is structural. CBP wants accountability tied to a licensed broker, and AI does not carry that liability.

The right question, he argued, is not replacement. It is how much efficiency the human can gain from artificial intelligence.

Where the speed gain is real

Gould gave a concrete example from Gaia’s HTS classification work.

“If a client came to me and they wanted to classify a hundred products, I might look at that and say it’s going to be 20 minutes per product, half an hour per product to classify it. But using the AI tool, I can do those in 20 seconds, 30 seconds per product.”

That is a dramatic productivity jump, but it does not remove the broker from the final call. It simply moves the work from slow to fast.

For trade teams living through the pace of 2025, that difference matters. It can turn a CFO’s question about tariff exposure on a SKU list from a weeklong scramble into an afternoon answer.

Why specialized AI beats general models

Bozicevic said usage of AI for trade questions was up a hundredfold year over year. The demand was obvious. Supply chain managers were suddenly trying to understand HS codes, HTS codes, USMCA certificates, landed cost math, and de minimis rules they had never cared about before.

General-purpose LLMs hit a wall quickly.

“I know general AI is pretty good and getting better fast,” Bozicevic said, “but the specificity of tools like Gaia, the prompting, the follow-up questions. Is this a woven or knit fabric? No other system asks that kind of question.”

That is the heart of the difference. A general model that does not ask the right follow-up can sound confident and still be wrong. A purpose-built classification tool is built to ask.

Strbanovic added another use case that general AI does not naturally cover. Trade leaders also need a signal. They need tools that help them see industry rumors, policy shifts, and emerging pressure before the market does.

That is where trade-specific AI starts to look less like a nice add-on and more like a competitive layer.

Three Views of 2026

Stefanutti closed the episode by asking each guest to look ahead. The answers did not match, but they did rhyme.

1. Bozicevic sees more stability

Bozicevic thinks 2026 should be better than 2025. More stable. More predictable.

His reason is simple. Tariffs were being used not only as trade policy but as a negotiation tool. Once those negotiations settle, the daily churn should ease.

He also expects to see where tariff revenue goes. If reshoring is the goal, he thinks 2026 should show signs of manufacturing investment, which would drive demand for raw materials and near-finished goods imports.

2. Gould sees more complexity

Gould was more cautious.

He sketched a scenario where the Supreme Court overturns IEEPA tariffs as unconstitutional, and the administration responds by layering in other authorities, including Sections 122, 338, 232, and 301. In that version of events, the short-term tariffs get replaced by long-term ones, but the complexity stays alive.

There is also the question of refunds. Companies would still have to figure out whether to protest after liquidation or pursue a pre-liquidation court action to stop liquidation in the first place.

His conclusion was stark. Even an importer-favorable ruling could still create more work, not less.

3. Strbanovic sees a middle ground, powered by technology

Strbanovic thinks the most aggressive bilateral negotiations will fade and agreements will settle in.

Still, he pointed out that lobbying in Washington remains intense and exclusions are already arriving. If IEEPA falls, the administration has other authorities available, and using them would still keep pressure high.

His optimism came from a different place. Technology has changed the response speed.

“If this was happening a few years ago, I don’t know that trade would be able to recover and react. The amount of work required is just insane. With technology, even if they make a ruling on Friday night, being active and live on Monday morning, technology helped us through all that.”

That may be the clearest lesson of the whole conversation. 2025 did not just increase complexity. It exposed which trade teams had the systems to move with it.

What Trade Teams Should Take Into 2026

The practical takeaway is less about prediction and more about readiness.

Keep free-trade documentation airtight. Build or rent classification expertise instead of depending on suppliers. Treat AI as an efficiency multiplier for human decision-making, not as a replacement for it.

The teams that handled 2025 best were the ones that had those habits before the year began.

Explore Gaia Dynamics

If your team is trying to keep pace with shifting tariffs, classification pressure, and cross-border compliance work, explore Gaia Dynamics.

Want to Hear the Full Conversation?

Listen to Episode 1 of the Trade and Tech podcast for the full version, including the Train the AI segment where Bozicevic and Strbanovic name the trade AI capabilities they want Gaia to build next.

Frequently Asked Questions

Why does specialized AI outperform general LLMs for HTS classification?

Because trade classification depends on product-specific follow-up questions and regulatory context. General models often miss those details. Specialized tools are built to ask them.

What is the de minimis change, and why did it hurt SMBs disproportionately?

De minimis rules historically allowed low-value shipments to enter with less friction. When those rules tightened, small sellers lost a cost structure they had built their business around. Larger importers had more room to absorb the change.

What is the practical takeaway for importers heading in 2026?

Keep your documentation tight, build classification capability, and use AI to speed up expert work. The companies that did those three things in 2025 were far better prepared for the churn.