What Supply Chain AI Did When the Tariffs Hit

There's a useful thought experiment for evaluating any planning tool: how does it perform when the scenario you didn't model actually happens?

The tariff landscape of late 2025 and early 2026 has been an extended live test of that question for supply chain AI. The announcements came in waves: new tariff structures on goods from multiple origin markets, timelines that shifted, policy uncertainty that made traditional planning models effectively useless because the core input variables kept changing. McKinsey's December 2025 supply chain risk survey found that 82% of respondents said their supply chains were affected by new tariffs.

And 95% of retail executives surveyed by Deloitte were anticipating rising costs as a direct consequence of shifting trade policy. That's not a niche concern. It's essentially the whole sector.

So what actually happened in organisations that had invested in supply chain AI, versus those that hadn't?

The Scenario Modelling Difference

The honest answer is that AI didn't make tariff disruption easy. It didn't give anyone a crystal ball or automate their way out of a genuinely difficult trade policy environment. What it did, for the organisations that had built it properly, was compress the time between "new tariff structure announced" and "we understand our exposure and our options."

Traditional supply chain planning processes for something like a major tariff change would typically involve pulling data from multiple systems, running it through spreadsheet models built for a different set of assumptions, coordinating across procurement, logistics, and finance, and arriving at a scenario analysis two or three weeks later. By which time the situation had often moved again.

AI-powered scenario modelling (systems that have continuous access to live cost data, real-time freight rates, supplier location mapping, and demand forecasts) can run thousands of what-if calculations across sourcing combinations in hours, not weeks. Landed cost calculations that factor in tariff rates, transit costs, and geopolitical risk simultaneously. Sensitivity analyses showing which scenarios hold up under a range of policy outcomes, not just the base case.

The difference in response time is the thing. In an environment where tariff policy was changing month-to-month, the ability to rerun the analysis quickly enough to actually use the output was the competitive advantage, not the sophistication of the model itself.

What "AI Supply Chain" Actually Means in Practice

It's worth being specific here because "AI supply chain" covers a lot of ground, from inventory optimisation algorithms that have been running for a decade to genuinely novel multi-agent planning systems.

For most mid-to-large retailers in 2025, the relevant AI supply chain tools fell into a few categories. Demand forecasting models that incorporate a broader range of signals than traditional statistical forecasts: marketing calendars, competitor activity, external datasets, weather patterns. Inventory allocation systems that continuously reoptimise stock distribution across a distribution network based on current demand signals. Supplier risk platforms that aggregate news, financial, and geopolitical data to give early warning on supplier reliability concerns. (For a deeper look at AI demand forecasting specifically, the demand forecasting piece is worth reading alongside this one.)

The tariff scenario is an interesting stress test because it's not primarily a demand-signal problem. It's a cost and sourcing structure problem. The AI systems that performed best in the tariff crunch were the ones with good integration into procurement data (actual supplier contracts, lead times, alternative sources) rather than just demand-side data. A lot of supply chain AI investment has been heavily demand-side, and those systems were less useful when the primary variable being disrupted was the cost and origin of supply. VentureBeat's coverage at the time put it plainly: tariff turbulence exposed costly blind spots in supply chains and AI.

Worth noting that this is not a 2026 problem with a 2026 solution. The demand-versus-procurement integration gap has been a structural weakness in supply chain software for years. Tariffs just made it urgent.

The Nearshoring Acceleration

One downstream consequence of the tariff environment that AI is actively shaping is sourcing diversification and nearshoring decisions. Companies are fundamentally rethinking where and how they manufacture and source goods. Deloitte's 2026 Retail Industry Outlook found that 66% of retail executives planned to restructure their supply chains through onshoring, nearshoring, or supplier diversification if input costs rose. The data infrastructure required to evaluate a nearshoring decision, comparing supplier options across quality, cost, lead time, risk profile, and tariff exposure under multiple policy scenarios, is exactly the kind of multi-dimensional analysis where AI systems add genuine value over spreadsheet modelling.

The UK context here is specific. Post-Brexit, many UK retailers had already done some degree of supply chain reconfiguration for EU trade. That process, however painful, built some operational familiarity with structural supply chain change. Whether that's useful preparation for navigating the current US trade policy volatility, or whether it just means UK supply chains are already tired from one round of reconfiguration when another arrives, probably varies by retailer. My sense, talking to people in the sector, is that the retailers who came through Brexit with their supply chain data in better shape are the ones who found the tariff modelling less overwhelming. The muscle memory wasn't in the decisions. It was in knowing where the data lived.

The Honest Assessment

AI didn't save anyone from tariff impact. Costs went up. Margin got squeezed. Some product categories became significantly harder to price competitively. The retailers who had invested in supply chain AI didn't escape that; they navigated it faster and with more visibility into what was happening and why.

That's actually a meaningful difference in a fast-moving environment. Knowing that your exposure is primarily in a specific category, from a specific origin market, with three alternative sourcing options that have been pre-evaluated, and knowing this three weeks before your competitor, matters.

What the tariff environment has also done is surface the weak points in supply chain AI investments. Systems that were good at demand forecasting but didn't integrate well with procurement data weren't useful here. Systems that required significant human effort to set up new scenarios couldn't keep pace with how quickly the policy environment was shifting. Data quality problems in supplier and product origin data (the kind that sit quietly causing no problems for years) became urgent when you needed to accurately assess tariff exposure at SKU level.

Deloitte's survey found that only 30% of retailers currently use AI for supply chain visibility, with that figure expected to reach 41% within the year. The gap between those two numbers is where the competitive divergence is happening right now. The warehouse automation piece on autonomous warehousing and what it actually delivers covers the adjacent infrastructure question.

In that sense, the crunch has been clarifying. Supply chain AI that actually helps under pressure is built differently from supply chain AI that looks impressive in a demo. More companies have now learned which kind they have.

---

Data Sources and Further Reading

• McKinsey Supply Chain Risk Survey (cited via Supply Chain Strategy)
• Deloitte 2026 Retail Industry Global Outlook
• Tackling Tariff Turbulence with AI — SupplyChainBrain
• Tariff Turbulence Exposes Blind Spots — VentureBeat
• Tariffs, AI and Centralisation: Key Supply Chain Trends in 2026 — The Supply Chain Report