Multitenant IP networks, such as those found in data centers and service provider environments, face recurring challenges in IP address allocation due to overlapping address spaces among tenants. Virtual Routing and Forwarding (VRF) provides logical network segmentation but does not inherently prevent IP address conflicts. This paper proposes a hybrid AI-based framework that integrates unsupervised learning, clustering techniques, and heuristic expert systems to predict, detect, and resolve IP addressing conflicts in VRF-enabled multitenant infrastructures. The framework includes a predictive overlap detection model, a recommendation engine based on allocation heuristics, and an automation interface for real-time policy enforcement. A proof-of-concept implementation demonstrates over 95% reduction in address conflicts and up to 98% accuracy in conflict prediction, validating the feasibility and effectiveness of the approach for proactive network management.