1. Overview (Project Background)

This project is an AICX counselor chatbot PoC,
and was pursued to verify functional feasibility of whether generative AI can improve counselors' work efficiency.

To enable faster and more accurate responses to repeatedly occurring inquiries in international flight consultation,
we built a generative AI chatbot exclusively for counselors
and applied RAG (Retrieval-Augmented Generation) structure based on international flight counselor processing guides.

Additionally, by linking practical functions such as consultation history inquiry and airline ticket refund fee calculation through APIs,
we aimed to create an environment where the entire consultation process can be handled through a single interface.

Before project initiation,
and consultation guides were managed in Excel file format, resulting in low searchability and usability.
In particular, due to the structure requiring manual calculation of airline ticket refund fees,

2. Solution (Resolution Approach)

To resolve these issues,
we configured a Vector DB-based RAG pipeline specialized for counselor tasks.

We loaded counselor processing guide documents and airline-specific regulation documents
into an OpenSearch-based Vector DB,
and designed the system so that generative AI accurately understands counselors' query intent through semantic search
and generates responses based on relevant documents.

Through this, we implemented a consultation support structure that enables context-based search and consistent answer provision,

3. Result (Achievements)

Through this PoC, we were able to verify the effectiveness of generative AI-based consultation support.

Counselors can now quickly search for airline-specific regulations and processing standards
through natural language queries alone,
and have secured an environment where consistent answers can be provided for identical questions.

Additionally, by automating airline ticket refund fee calculations,
we reduced the time required and error potential from manual calculations,
and confirmed that overall consultation processing speed and work efficiency can be improved.