Motivation: This research investigates the transformative impact of integrating AI into mathematics education, aiming to enhance students' conceptual understanding and self-efficacy. It addresses the crucial need for innovative teaching methods in response to contemporary challenges in education and aims to fill gaps in understanding the potential drawbacks and benefits of AI implementation. Objectives: The primary goal of this study is to investigate the effects of AI tools, such as GeoGebra and ChatGPT, on students' conceptual understanding and self-efficacy in a mathematics classroom setting, with a focus on collaborative learning, teacher-led discussions, and problem-solving. Methods: Employing a mixed-methods approach, the research involves pre- and post-implementation surveys, supplemented by qualitative data from interviews, focus groups, and classroom observations. The study emphasises the utilisation of AI in collaborative learning environments and explores the challenges and opportunities associated with its integration. Results and Conclusions: The study reveals a noteworthy enhancement in students' conceptual understanding and self-efficacy belief when AI tools are incorporated into mathematics education. However, it highlights concerns such as technological challenges, potential teacher replacement fears, biases in AI systems, and difficulties related to pacing and social interaction. To optimise the use of AI, the paper suggests strategies like adaptive learning platforms, real-world simulations, intelligent tutoring systems, and data-driven instruction, positioning AI as a complement rather than a substitute for traditional teaching methods. The study recommends addressing challenges through additional support resources and transparent communication, emphasising the need for a thoughtful integration of AI in education.