AI Voice Avatar as an ESP Speaking Partner: An Autoethnographic Exploration of Business English Vocabulary Acquisition

Abstract

This autoethnographic study investigates how an AI voice avatar functions as an English for Specific Purposes (ESP) speaking partner for Business English vocabulary acquisition. Specifically, this study used Sesame AI, a web-based AI voice-avatar application, as the speaking partner for simulated professional conversations. It aims to examine the researcher-participant’s lived experience of using an AI voice avatar for professional speaking practice, to understand how AI-supported interaction facilitates contextual vocabulary learning, and to identify the business-related vocabulary that became more usable through repeated practice. The study employed a self-reflection design grounded in autoethnography. Data were generated across six speaking sessions over four weeks using a treatment log, vocabulary learning log, reflective journal, and critical incident reflection. The data were analyzed thematically by tracing patterns of vocabulary noticing, repetition, contextual use, confidence, autonomy, and technological constraints. The findings show that Sesame AI supported retention of Business English lexical chunks, encouraged autonomous rehearsal, and strengthened speaking self-efficacy by providing a low-pressure conversational environment. However, the interaction also revealed pragmatic limitations, including occasional generic responses, imperfect speech recognition, and limited emotional sensitivity. The study contributes a reflective learner-centered perspective to ESP and CALL by showing how web-based AI voice avatars may serve as supplementary speaking partners for self-directed professional language learning when supported by critical reflection and human pedagogical judgment.