Current page

Current page

HTS Code Identification Assistant for US Customs

Conversational Assistant

We built an intelligent conversational assistant for US custom brokers to accurately identify HTS codes based on vague or incomplete product descriptions.

HTS code misclassification is a major fraud vector. It 
ranges from simple duty evasion to national security threats through mis-declared goods. 

While some cases are honest mistakes due to vagueness, many are strategic manipulations, making this a high impact area for AI, data, and compliance tech innovation. 

photo of cargo cru ship
photo of cargo cru ship

A two stage architecture using LLMs:

First clarify the user provided product descriptions and then retrieve the most relevant HTS codes.

Multi-query retrieval and RAG based reasoning used to refine ambiguous inputs and search for most contextually accurate codes

Dynamic follow-up questioning, where the bot asks for additional product details such as material usage to disambiguate entries.