Artificial Intelligence (AI) Driven Proactive Customer Service Excellence Platform in e-commerce Industry

Abstract

In this paper, quantitative research was conducted using AI models to classify and rank customer service messages in e-commerce. Accuracy, efficiency, and severity ranking Five model types, namely RNN, LSTM, CNN-BERT hybrid, DistilBERT, and BERT were tested. The findings demonstrate that transformer models are better than traditional deep learning models in all key metrics. DistilBERT is an algorithm that offers the most appropriate tradeoff between speed, accuracy and memory consumption, thus suitable in real-time application. Transformer models were also more stable in prediction of severity. The results justify the use of current NLP systems in the context of faster, more precise, and responsive automation of customer service.