This pilot study investigates the implementation of AI spoken agents in oral language acquisition. While Large Language Models have garnered significant attention in recent years, most research has focused on text-based applications, with little emphasis on oral language. Furthermore, AI-spoken agents remain an under-researched area despite their increasing prevalence. This study examines the implementation of the Doubao Agent (Kyle) in facilitating oral Chinese acquisition and explores how the spoken agent contributes to oral Chinese learning. Also, it examines learners' perceptions of such an application. The participants were four intermediate L2 Chinese learners in China. Data on oral communication was collected weekly over a month, followed by an interview. The results showed that learners demonstrated improved oral communication, suggesting the effectiveness of the spoken agent in stimulating oral learning. However, oral productions were primarily simple, highlighting the need for strategies to stimulate complex language. Interviews showed interesting results, particularly in the dominant role of the spoken agent in communication. Such findings call for more research on strategies to take advantage of spoken agents in oral language learning.