Tariff exposure cannot wait for the next spreadsheet

When 78% of large organizations surveyed reported higher COGS in their latest quarter, and 51% reported margin declines, the problem was not a lack of opinion about tariffs. The cost landed first in the product, then in the margin, and only later in the spreadsheet that explained what happened.

According to KPMG, 34% of surveyed organizations were already passing more than half of tariff costs to customers, up from 13% in May 2025, while 55% planned additional price increases of up to 15% within six months. It is a planning-speed problem that hits pricing, sourcing, inventory, and cash at the same time.

The common failure is familiar to anyone who has had to reconcile a bill of materials against customs records and a contract folder. Companies usually know their direct suppliers, but not the sub-tier parts, shared dependencies, country-of-origin evidence, or the customer margins exposed when one input changes. By the time trade, procurement, engineering, finance, and sales each finish their own review, the opportunity to act has narrowed or disappeared.

Exposure hides in the handoffs, not the headline tariff

Tariff shocks persist because the enterprise has split the work into pieces that do not line up. Procurement owns supplier choice, trade compliance owns classification, finance owns margin, operations owns continuity, and sales owns pricing. None of them can answer the full question alone: which finished products, parts, suppliers, plants, and customers are exposed if a duty changes or a source fails?

That gap matters because the market is already moving. According to KPMG, 65% of global-trade respondents cited changing sourcing patterns as a tariff-mitigation strategy, and 57% planned to renegotiate supplier contracts. According to the U.S. Chamber of Commerce, 35% of surveyed SMBs had changed suppliers because of tariffs in the prior 12 months, usually because of higher costs, country-of-origin risk, reliability problems, and tariff-driven lead-time changes.

The old process cannot keep up with that pace. It depends on periodic spreadsheets, manual supplier outreach, and disconnected ERP, procurement, logistics, and finance data. It is slow because the records needed to trace exposure live in different systems under different identifiers, and the exception work is still done by hand.

The usable system is a graph, a retrieval layer, and a margin engine

The practical answer is a reference design that connects ERP, procurement, PLM, logistics, customs, finance, supplier portals, and contract repositories into one supply-chain graph, then uses retrieval and calculation to show exposure by part, supplier, country of origin, product, and customer margin.

That starts with entity resolution. Part numbers, supplier names, plant locations, product hierarchies, and shipment records must be normalized so the system can trace one duty change through the finished good, the bill of materials, the purchase order, the invoice, the customs record, and the customer price. Without that, the model is only reciting the data fragmentation everyone already knows.

The second layer is evidence-linked retrieval. The system should pull the supplier declaration, the origin certificate, the customs ruling, the contract clause, or the engineering note that supports a classification, and it should keep the source, date, and confidence attached. MIT researchers describe a planning workflow in which AI agents onboard company data and analyze supplier contracts for constraints such as minimum order quantities and lead times.

The third layer is the part people actually buy it for: exposure calculation and margin simulation. If a tariff rises, a supplier is late, or an alternate source is more expensive, the system should propagate that change through landed cost, standard cost, gross margin, contribution margin, and customer pricing. If the system cannot show the cost bridge and the assumptions behind it, finance will not trust the answer and operations will not use it.

A controlled workflow makes the system useful instead of theatrical. A high-risk part can trigger a retrieval of supporting documents, a scenario run, a task for alternate-source qualification, a pricing review, and a finance alert on the affected margin. It is to compress the time between an event and the decision that protects margin or continuity.

The architecture only works if the system is built around review, not around guesswork. The diagram below shows the stages and controls that matter.

Speed only matters if the answer can be defended

There is a reason many companies still hesitate to centralize tariff and supplier intelligence. A recommendation that cannot show source documents, assumptions, and confidence will be ignored the moment it collides with a qualification risk or a customer promise. That is why the system has to separate retrieved facts from inferred values and route uncertain classifications to trade specialists.

The weakest implementations also fail on incentives. Buyers are often rewarded for purchase-price variance even when the enterprise absorbs tariff, expedite, stockout, and lost-sales costs. Yet one survey found only 16% of respondents prioritized the lowest-cost provider, while nearly 70% prioritized reliability or strategic alignment. The decision criterion has already changed. The operating model has not caught up.

There is a cost to doing this properly. Data integration, master-data cleanup, document normalization, and governance are not lightweight work, especially when bills of materials are incomplete and supplier provenance is buried in PDFs or email. If the company cannot establish ownership for tariff classification, origin data, pricing, and approval rules, the system will still be stuck waiting on the same handoffs that slowed the process before.

The right question is not whether AI can predict every disruption or replace the trade team. It is whether the company can see exposure by part, supplier, and margin fast enough to act before the quarter closes. The organizations that do this well will not just change suppliers after the fact. They will know which products they can hold, which ones they should reprice, and which contracts need review while there is still time to matter.

If you want to see how that workflow can be built into your own environment, QueryNow will take the workflow you want gone, build it in two weeks, and run it from $1,200 a month, with the final setup payment due only after it meets the acceptance criteria you signed off on. Start there at Build Your AI.

For teams that want the broader operating model behind this, the free paper The mid-market AI advantage is a useful next read, especially if the current process still depends on monthly review decks and manual reconciliation.