This study addresses limitations in traditional Problem-Based Learning (PBL) for object-oriented programming (OOP) education, which focuses on isolated programming knowledge and fails to personalize learning. We propose an AI-driven framework where students collaborate with Agentic AI to tackle authentic challenges. The AI assists by decomposing projects into Minimum Viable Solutions (MVS), co-designing personalized pathways, and providing real-time scaffolding. Results show students transition from syntax learners to system architects, deeply engaging with OOP principles like encapsulation and polymorphism. Teachers shift to design mentors while AI handles adaptive support. This synergy cultivates adaptable problem-solvers for evolving technological landscapes.