Although research highlights the potential of AI-powered tools in language education, my work experience in an educational technology company suggests challenges when these tools are used without humans. This aligns with the research findings of Fryer et al. (2017), Song and Song (2023), Davis (2022) and Saleem et al. (2025), that standalone AI technologies can result in less sustained interest and reduced learning outcomes. Chan and Tsi (2024) and Satar (2021) further contend that AI cannot replicate human thinking, creativity, or emotional interaction, and its predictable responses may weaken long-term engagement. However, much of the existing literature emphasises AI’s technological capabilities or roles in language education rather than its integration with human teachers (i.e., De la Vall & Araya, 2023; Xu et al., 2021; Sun et al., 2024; Leat & Taylor, 2023). The pedagogical integration of AI remains under-researched (Muthmainnah et al., 2022; Godwin-Jones, 2022). Therefore, Chatfield (2025) and Moulieswaran and Prasantha (2023) call for active teacher involvement in both research and design processes to ensure AI applications align with learning objectives and to support the development of useful practices. AI research should focus not on replacing teachers but on enhancing teacher-student interaction (Asrifan & Dewi, 2024).

Grounded in sociocultural perspectives that conceptualise learning as socially mediated, this study investigates not only how AI-powered speaking tools can be integrated into teacher practice, but also the methodological challenges that arise when researching such integration in dynamic classroom environments. To achieve these aims, this research will address the following questions:

1. Which design features of AI-powered tools support student engagement and performance in speaking instruction?

2. How can AI-supported approaches be used to improve student engagement and performance?

3. What are the perceptions and experiences of students and teachers towards AI-teacher-student collaboration?

4. What are the challenges and barriers in AI-teacher-student collaboration?