AI tariff classification for footwear importers
A defensible tariff code, with the reasoning on the record.
TariffWise interviews you the way a customs expert would. It asks only the questions that change the code, suggests a ten digit HTS classification, explains why, and stores the record for audit.
Anatomy of a classification
Footwear
Textile uppers
Sports footwear
Duty line
Each segment narrows the classification, and each turns on a fact about the product. Upper material, ankle height, and toe construction decide where a shoe lands. TariffWise captures those facts one question at a time.
Live demo
This is the working prototype. Enter a footwear product below and answer the interview. The engine walks verified 2026 USITC Chapter 64 data and the AI layer extracts facts from your answers.
Recorded walkthrough
A complete interview session, from product description to stored audit record. The recording shows the full flow even when the live Space is asleep.
How it works
Five steps move the user from rough input to a documented, exportable record. The user supplies what they know. The AI gathers what the code requires.
Describe the product
A short form takes the product name and a plain description. No technical detail is required to start.
Answer the interview
The AI asks only the questions that change the code. A live panel shows the working classification narrow with each answer.
Review the result
The suggested code arrives with a numbered rationale tied to your answers. A chat box lets you question it.
Keep the history
Every classification is stored with its date, status, and reasoning. Past decisions stay reviewable.
Export the record
The full audit record holds the facts, the rationale, and the steps taken. It exports as evidence for customs review.
Design decisions
Classification capability is common across competing tools. TariffWise competes on how clearly it reaches a defensible code and documents the reasoning. Each interface choice below serves that goal and is grounded in a specific design principle.
The intake form stays short
The form asks only what the importer already knows. The AI gathers the technical facts in the interview. Good discovery reduces the effort a user must spend to get value (Cagan, 2018).
The rationale forms in view
A live panel shows the working code update after every answer, so the user watches the reasoning take shape. Showing why an output was generated helps users judge whether to accept it (Weisz et al., 2024).
Limits are stated before use
The app declares what it can and cannot do on the first screen, including that it suggests and explains but does not file. Upfront explanation of capabilities and limitations calibrates user trust (Weisz et al., 2024).
The result can be interrogated
The suggested code carries a numbered rationale, and a chat box lets the user challenge it in plain language. Trust built through the interface shapes the intention to use a tool, at times more than usability does (Zieglmeier & Lehene, 2022).
Every control signals its action
Buttons name what they do, and each answer produces a visible change in the working classification. Signifiers show where action is possible and feedback confirms what the action did (Norman, 2011).
The AI assists, the importer decides
The AI extracts facts and documents reasoning. The importer reviews, questions, and accepts the final code. AI serves users best when it augments their judgment rather than replaces it (Moran & Nielsen, 2023).
TariffWise is decision support. It suggests a classification and explains the reasoning. It does not file entries and it is not legal advice. The importer remains responsible for the final code.
Sources
Cagan, M. (2018). Inspired: How to create tech products customers love (2nd ed.). Wiley.
Moran, K., & Nielsen, J. (2023, November 3). AI for UX: Getting started. Nielsen Norman Group. https://www.nngroup.com/articles/ai-ux-getting-started/
Norman, D. A. (2011). The design of everyday things. Tantor Media.
Weisz, J. D., He, J., Muller, M., Hoefer, G., Miles, R., & Geyer, W. (2024). Design principles for generative AI applications [Preprint]. arXiv. https://arxiv.org/abs/2401.14484
Zieglmeier, V., & Lehene, A. M. (2022). Designing trustworthy user interfaces [Preprint]. arXiv. https://arxiv.org/abs/2202.12915