Jun 17, 2025

The global trade war tests the limits of predictive modeling, experts say. But there are other uses for AI. From supercharging employee productivity and streamlining processes to inspiring innovation, Microsoft ’s AI is designed to help you build the next big thing. No matter where you're starting, push what's possible and build your way with Azure's industry-leading AI. Check it out . When retailers plug into Altana’s supply chain management technology, they now have some timely AI-powered controls. They can game out models of how various tariffs could hit different parts of their logistics operations, and maybe renegotiate contracts or switch suppliers as a result, according to Amy Morgan, Altana’s VP of trade. What Altana doesn’t offer—and what no other AI tool can really promise right now—is a way to confidently predict the future. At a time when a scattershot trade war has thrown global trade into flux, fast-changing policies are testing the limits of AI’s predictive powers. Tech vendors say AI has a lot to offer when it comes to supply chain management, from scenario planning to agents that manage coordination. And retailers have been beefing up on this kind of tech, especially since the supply chain chaos of the pandemic. But there’s only so much that the historical data powering these algorithms can tell them amid unprecedented moves and teetering tariffs. “[You can say] ‘Here are my top categories, or maybe my top products, that I import and that I sell. I want to know how they’re going to be impacted if the tariffs go back up to 145%. Or what if they come down?” Morgan said. “Or what if I were to find a supplier in Mexico that qualifies for [the US-Mexico-Canada trade agreement,] which removes the tariffs altogether?” “That’s the sort of modeling we’re talking about,” Morgan said. “We’re not passing predictions on what’s going to happen.” When past isn’t prologue: Retailers are increasingly turning to AI for tasks like route mapping and diversifying their sourcing beyond a reliance on China, according to Sumeet Trehan, co-founder and CEO of supply chain tech startup Starboard. But when it comes to inventory planning—deciding how much of a given product to stock at a given time—predictive machine learning models might fall short in this new environment, he said. “Basically, the prediction is as good as your past data,” Trehan said. “And if it’s a black swan event, those data models are not really useful, and that’s why what we are seeing is still a combination of machine learning models and human gut instinct acting together.” Pando AI, another supply chain tech startup, aims to automate parts of supply chains for big retailers and manufacturers with consolidated data and analytics, as well as teams of agents. But the company stops short of offering predictions, according to CEO and founder Nitin Jayakrishnan. Rather, Jayakrishnan said he pushes “a fundamental shift” toward creating systems that are more dynamic and flexible in real time by, for example, more frequently revisiting procurement contracts. “Our view is you shouldn’t predict, because if you do predict, that prediction is going to be wrong anyway,” Jayakrishnan said. “The time and effort and dollars spent on predicting is, I think, better spent on sensing and pivoting.” Kurt Muehmel, head of AI strategy at data management platform Dataiku, said the platform’s tariff-exposed clients are thinking along similar lines. “The unpredictability and the rapid changes [are] creating a context where what they’re looking to do is not so much predict the future explicitly—because you can’t right now,” Muehmel said. “But rather, they’re looking to equip their teams who are doing pricing forecasting, who are doing all the supply chain analysis…demand forecast, all that—they’re looking for ways that they can equip them to basically move as fast as possible.” Price check: That’s not to say retailers should just throw up their hands on predictive modeling. There are certain sources of historical data that are useful no matter the present environment, as long as the dataset is big enough, according to Kaitlyn Glancy, a partner at Eclipse Ventures, which focuses on early-stage and early-growth startups in physical industry tech spaces (including Starboard). “What historical data does give you is…some sense of consumer behavior,” Glancy said. “Consumers aren’t going to stop buying toilet paper or key inputs for their home…Retailers are always experimenting with price elasticity. I actually think for bigger retailers with bigger datasets, some of that data can be helpful, because you can use that to extrapolate, like, ‘Hey, if, in fact, this product is going to be hit with a 20% tariff, and we expect to pass through half of that back to our customer, and therefore it’ll cost 10% more, based on historical trends of price increases, price elasticity, what could likely happen here?’” Pandemic lessons: The current global trade war isn’t the first time that AI predictive models have faced this problem. Many of the experts we talked with for this article discussed the lessons from the supply chain chaos during and following the Covid-19 pandemic. “The story then, which was true, was that essentially it broke every machine learning model,” Muehmel said. “Because there was nothing in the historical patterns that reflected, or at least in any way resembled, the Covid supply chain shocks. And so there’s a lot of those learnings which are fresh enough in everyone’s minds.” Justin Honaman, global head of worldwide retail, restaurants, and consumer goods business development at Amazon Web Services, said most retailers have been rethinking their supply chain tech in the past couple of years, whether that means economic resource planning, point of sale systems, or warehouse management. “Traditional forecasting techniques don’t often account for some of the things that we’ve experienced, not only in the last six months but the last couple of years, whether it be Covid or the barge getting stuck, or containers being misplaced in different locations,” Honaman said. “What we tell our customers is, let us help you get to a place where you have supply chain resiliency.” New normal: One thing that experts we spoke with also agreed on was that global trade volatility is likely here to stay for the foreseeable future. “The current president has made it very clear that he intends to use tariffs as a negotiating tactic,” Eclipse’s Glancy said. “Knowing that, I think most retailers are expecting volatility over the next three and a half years. And so what does that look like in practice? You’re probably diversifying your supply chain…That is a whole additional level of complexity and operational complexity that will exist in the ecosystem. And the question is, can you pull that off without adding a ton of cost structure?...So you’re going to have to leverage the latest and greatest AI tools.” Copy Keep up with the innovative tech transforming business Tech Brew keeps business leaders up-to-date on the latest innovations, automation advances, policy shifts, and more, so they can make informed decisions about tech. Subscribe