Optimizing Conversational Commerce Involving Multilingual Consumers Through Large Language Models’ Natural Language Understanding Abilities

Document Type

Conference Proceeding

Publication Date

1-1-2024

Abstract

Due to the emergence of natural language processing (NLP) interfaces, there has been growing intent to use conversational channels for commerce. Beyond customer service, NLP-enabled AI agents are being integrated into various steps of the order-to-cash (OTC) process. Social media and messaging platforms such as Facebook Messenger have become pivotal for businesses, especially during and after the COVID-19 pandemic, but adoption has been limited. In addition, attitudes towards fully-automated conversational agents (CA) have been mixed, and there is room for human involvement in transactional conversations. A distinguishing contribution of this research is leveraging the inherent capabilities of Large Language Models (LLMs) in handling multilingual conversations and extracting transactional details through named entity recognition (NER). The study describes a hybrid human-AI setup augmenting agents with an auto-agent leveraging LLMs’ natural language understanding (NLU) capabilities, designed using the OTC process pattern applied to conversational UX frameworks. A prototype of the setup aims to streamline operations and reduce errors by enhancing the user experience during key OTC steps through improved conversational design. Recognizing the irreplaceable essence of human interaction, the hybrid human-in-the-loop approach was chosen, mitigating the impersonal nature of full automation. A prototype handling customers and humans augmented by LLMs for NER handling of transaction, customer, and product information was built. Sample synthetic bilingual conversations between customers and sales agents were generated using ChatGPT and fed into the system for evaluation.

Link to Full Text

Share

COinS