Scaling AI Across Retail: Leveraging NLP and Computer Vision to Enhance Customer Experience and Reduce Returns

Abstract

Product returns in fashion e-commerce are a persistent and costly challenge, often signaling deeper misalignments in customer expectations, product representation, and operational processes. This thesis explores how Artificial Intelligence (AI), specifically Natural Language Processing (NLP) and Computer Vision (CV), can be applied to reduce returns and improve the customer experience across digital retail channels. Through a systems-based approach, the thesis evaluates technical applications for NLP and CV, and provides strategic frameworks for deploying AI in real-world contexts. Using fashion e-commerce as a representative use case, the research analyzes applications such as virtual try-on, size and fit prediction, visual search, defect detection, and customer sentiment analysis. The thesis introduces a strategic AI adoption roadmap tailored to omnichannel fashion retailers, incorporating AI maturity assessments, change management frameworks, and KPIs to guide successful implementation. These approaches can support retailers to assess readiness, prioritize use cases, and scale responsibly. The thesis integrates qualitative insights and quantitative modeling to estimate the range in potential financial impact, highlighting that AI-enabled interventions can both reduce costs and increase sales from brand loyalty. The findings offer a practical framework for fashion retailers seeking to embed AI into customer-facing systems and decision-making processes for an enterprise-wide AI transformation.