Abstract
The World Trade Organization (WTO) maintains extensive trade concern records; however, extracting actionable insights from these largely unstruc-tured datasets remains challenging. Even though they are several formal information systems available like the Sanitary and phytosanitary Informa-tion Management (SPS IMS), Trade Concerns Database (TCD), and the Technical Barriers to Trade Information Management System (TBT IMS) which help with these issues but the challenges of retrieving structured in-sights from unstructured trade concerns descriptions, linking these textual descriptions to their specific Harmonised System codes, and tracking of mem-ber states who raised, supported or responded to specific concerns continue to persist. To this end, this study designs and implements a Retrieval Aug-mented Generation-enabled AI assistant grounded on WTO trade concern records, HS classifications, and institutional trade data. This system com-bines semantic search, document retrieval, classification, summarisation, and basic trend analysis with HS code product mapping to provide evidence-based support and automated response to user questions. This research combines prototype development and quantitative evaluation, which was implemented using Microsoft Copilot Studio and evaluated across five analytical task cate-gories using representative WTO trade concern cases including classification, summarisation, identification of member states, HS code mapping, and trend analysis, the result shows an overall accuracy of 96%. By grounding responses in retrieved WTO documents, the proposed RAG framework reduces the risk of hallucinated outputs associated with standalone large language models. This work contributes both practically and theoretically, showing how conversational AI engineer-ing and RAG can be used alongside ethical governance principles to support transparent data-driven decision-making even within the global trade con-text.