This session explores the technology and pedagogy behind MiMi, an AI-powered chatbot introduced in over 50 universities. We examine the accuracy and effectiveness of ChatGPT and Google Gemini models, the pedagogical principles driving their impact, and their role in boosting student motivation and language assessment. The presentation includes key insights from deployments with over 15,000 students in Japan this past year, analyzing motivation trends and alignment with CEFR-based ‘CAN-DO’ benchmarks.