The 1990s to early 2000s period saw a shift toward hybrid intelligent systems that fuse learning-driven models with symbolic reasoning and real-time control. Researchers favored modular architectures and practical deployment, integrating neural networks, rule extraction, and agent-based planning to tackle time-critical decision making in domains such as robotics and energy systems. The approach emphasized interpretability, adaptability, and distributed operation, laying groundwork for later neuro-symbolic and explainable AI developments.