Abstract
The swift integration of virtual conversational agents within public services has prompted inquiries regarding their quality and impact on user experience. This study examines taxpayers' perceptions/experiences of the South African Revenue Service’s AI-powered chatbot, Lwazi, with an emphasis on privacy awareness, perceived risks, usability, efficiency, and overall satisfaction. A quantitative, cross‑sectional survey was conducted involving 233 taxpayers visiting SARS branches in both metropolitan and non‑metropolitan areas. Utilising a snowball sampling method, respondents evaluated statements on five‑point Likert scales. Data analysis was performed using IBM SPSS, incorporating descriptive statistics, exploratory factor analysis, and reliability testing; Spearman’s ρ was employed to assess associations between constructs, and Mann‑Whitney U tests were used to compare regional experiences.
The findings reveal that users felt moderately informed about Lwazi’s privacy measures and data handling,yet expressed only neutral‑to‑slightly‑positive satisfaction. Language clarity was identified as the most significant ease-of-use attribute, whereas accessibility for users with varying digital literacy was the lowest. Efficiency factors averaged 3.27, with notable concerns regarding the timely resolution of queries. Privacy awareness exhibited a positive association with satisfaction, while risk concerns showed a negative correlation; ease of use and chatbot efficiency demonstrated strong positive correlations with satisfaction. Regional analysis indicated no significant differences in satisfaction between metropolitan and non‑metropolitan users. Overall, the study suggests that while chatbot Lwazi provides an acceptable user experience, improvements in privacy communication, responsiveness, and digital inclusivity could enhance user trust and satisfaction.