Designing an AI-based personal sales advisor for fashion retail : a human-AI task allocation framework

Abstract

The use of artificial intelligence technologies in retail is growing. It has created many new ways to improve the quality of the sales process for customers. However, retailers do not have clear recommendations on how to divide responsibilities between AI systems and human sales assistants in physical fashion retail stores. This thesis tries to fill this gap by developing and evaluating a Human-AI Task Allocation Framework for in-store fashion retail. After conducting a literature review on the topics AI in retail, collaboration of people and AI in retail stores, and technology acceptance, thirteen main fashion retail sales tasks were identified as well as six task characteristics. These are used to develop a three-layer framework that takes each task and based on decision-rules assigns it into one of three categories: AI execution, Human execution, Hybrid execution. The framework was reviewed by industry experts, who provided their gradings of each task across six task characteristics. The framework was also evaluated by the means of quantitative customer survey, in which 50 responses were collected and then analysed. The findings showed that the allocation results are accepted by fashion customers and that the idea of dividing tasks between actors based on characteristics provides meaningful results and can serve as a basis for designing Human-AI collaboration.