Importers Bought AI to Widen the Vendor List. Nearly Half Are Cutting It.
Nearly half of US fashion importers plan to work with fewer suppliers by 2027 while running AI across their sourcing operations. The study's own numbers show what that software is used for, and finding a vendor is not on the list.
Neritus Vale
Nearly half of the US fashion companies surveyed this year plan to work with fewer suppliers by 2027, almost three times the share that said so in 2025. The same survey finds AI already running inside their sourcing operations. FashionUnited reported both findings together, as a matched pair of trends; Supply Chain Dive covered the consolidation number on its own. They are closer to a contradiction than a pair: the software was sold on its ability to surface vendors a buyer had not found, and the buyer has just decided to stop looking.
The collapse in expansion appetite is the clearest single movement in the 2026 Fashion Industry Benchmarking Study, which USFIA runs with Sheng Lu and Emilie Delaye of the University of Delaware. Only 26.3 percent of respondents plan to add vendors over the next two years, down from 41.2 percent a year earlier. Geographic appetite fell harder, and it matters more, because switching countries is the move a tariff is supposed to provoke. On that question the share planning to expand went from 58.8 percent to 21.1 percent in twelve months. Respondents explain themselves without euphemism, one giving the goal as a “lower number of vendors to stabilize lead time and cost.”
What the industry does with AI sits in the same document, and it is not searching. The most frequently cited application is demand forecasting and inventory planning. Half of respondents use AI for sourcing strategy and cost optimisation, a job that presumes the vendors have already been chosen. Another 43.8 percent point it at customs, trade compliance and tariff work, which only begins once a purchase order exists. Discovery of new suppliers does not appear on the list.
The software arrived to answer a question the sourcing committee had already closed.
Search was never the binding constraint on a buyer’s options. Qualification is. Under the Uyghur Forced Labor Prevention Act, 31 percent of respondents now report sourcing from fewer vendors or slowing approval of new ones, against 18 percent a year earlier. Extended producer responsibility rules have loaded the same gate with traceability files and supplier documentation. A platform that turns a tech pack into curated quotes from factories it has scanned, the hybrid sourcing model Forbes described in June, still hands its output to a compliance process the platform does not run. Cutting the cost of the cheap half of a two-part problem does not move the total.
The buyer’s answer to tariffs has already migrated off the supply base and onto the customs entry. Applying for refunds from US Customs is the most common mitigation at 79 percent of respondents, which is a legal claim rather than a sourcing decision. First-sale valuation ranks second, renegotiating supplier contracts third, and only then does diversifying sourcing appear. Use of duty-free sourcing from US trade agreement partners fell to 47 percent from more than 72 percent a year earlier, because most preference programmes carried the higher tariffs too in 2026. Substitution stopped paying in the same year the software for finding substitutes reached the sourcing floor.

The strongest objection is that the option has not been destroyed, only relocated. Country coverage rose while vendor counts are only scheduled to fall. Respondents sourced from 49 countries this year, against 46 in 2025; the vendor cuts are a 2027 plan, not yet realized in the data. That combination is possible only if the surviving vendors are themselves multi-country. Sixty-five percent of respondents with more than 1,000 employees still source from ten or more countries, a share that rose rather than fell, and one of them gives the plan as “consolidating to fewer vendors with production in multiple countries.” A buyer on that model keeps every geographic option it had and stops paying to maintain duplicate relationships to reach it.
The objection identifies a real change in structure and the wrong owner of the option. A buyer holding two vendors in two countries could use one quote as a price on the other, which is what made renegotiation credible. Move both plants under a single vendor and the transfer gets priced by the party performing it, turning a right into a request. The study’s own figures show the widest menus cut first: about 7 percent of large firms now source from more than 20 countries, down from roughly 20 to 23 percent across the two prior surveys. Consolidation also selects for the vendors holding the scarcest capability, since compliance and traceability strength is the stated criterion for survival. A supplier retained because few rivals can pass the same audit has little reason to absorb a tariff on the buyer’s behalf.
These are two-year commitments priced inside a single quarter of maximum legal noise. The study surveyed 30 companies between April and June 2026, a small sample weighted toward the firms large enough to place the orders that move trade data. In May the Court of International Trade struck down the emergency surcharge the administration imposed after the Supreme Court voided the IEEPA tariffs in February, granting relief only to the three importers who had sued. In June USTR proposed action across 60 Section 301 investigations that would set duty rates by whether a country enforces forced-labour import bans, with a separate mechanism for apparel and textiles. Compliance capability would then price the duty, and importers have just concentrated their orders on the vendors who hold most of it. If the 301 regime lands close to that proposal, the firms that shortened their lists this spring will negotiate with counterparties whose scarcity the tariff schedule itself has raised.
The useful version of this software works on approval, not discovery. An AI that clears a traceability file, pre-screens an audit, or holds a qualified but dormant vendor in reserve at low cost restores some of the optionality consolidation is removing. Aimed at search instead, it will keep producing candidates the compliance queue cannot absorb and the two-year plan has already excluded. Importers are going to learn what a short list costs. The software they bought will describe that list in more detail than any sourcing team has ever had, and it will not be able to lengthen it.