The integration of artificial intelligence (AI) in language learning has created new opportunities to enhance students’ communicative competence and cognitive skills. This study investigated the effectiveness of integrating a flipped classroom approach with a conversational agent (Replika-AI) in improving students’ willingness to communicate (WTC), self-regulation (SR), and higher-order thinking skills (HOTS) in English-speaking learning. This research employed a mixed-method approach with a quasi-experimental design involving two groups of students from the English Language Education Study Program at Universitas PGRI Semarang. The control group implemented a traditional flipped classroom model, whereas the experimental group used a flipped classroom integrated with Replika-AI. Quantitative data were collected using Likert-scale questionnaires measuring WTC, SR, and HOTS and were analyzed using analysis of variance (ANOVA) through SPSS. Qualitative data on students’ speaking performance were analyzed using Brown’s speaking assessment framework and processed with MAXQDA software. The results indicated that students in the experimental group achieved significantly higher scores in all three variables compared to those in the control group. The integration of Replika-AI had a significant impact on students’ willingness to communicate, self-regulation, and higher-order thinking skills. Qualitative findings also revealed improvements in speaking components, including comprehension, vocabulary, fluency, pronunciation, and grammar. Overall, the findings showed that combining a flipped classroom with an AI conversational agent created a more interactive and independent learning environment, which contributed to the improvement in students’ English speaking skills.