Large Language Models in Consumer Electronic Retail Industry: An AI Product Advisor

Abstract

This paper explores the development of an AI product advisor utilizing large language models (LLMs). Firstly, we discuss the needs and the current problems faced by industry. Subsequently, we reviewed past works by various scholars regarding AI Assistant in Retail and Other Industries, and LLM Models and techniques used in generative AI chatbot for sales and service activity related works. Next, we assessed the performance of various models including Llama2B, Falcon-7B, and Mistral-7B, in conjunction with advanced response generation techniques such as Retrieval Augmented Generation (RAG), fine-tuning through QLora and LLM chaining. Our experimental findings reveal that the combination of Mistral-7B with the RAG and LLM chaining technique enhances both efficiency and the quality of model responses. Among the models evaluated, Mistral-7B consistently delivered satisfactory outcomes. We deployed a prototype system using Streamlit, creating a chatbot-like interface that allows users to interact with the AI advisor. This prototype could potentially increase the productivity of frontliners in the retail space and provide benefits for the industry.