AI-driven intrusion detection systems are widely used in enterprise cybersecurity, but most existing studies emphasize detection accuracy while giving limited attention to real-time visualization for operational response. This paper proposes Neural-Heatmap, a predictive cybersecurity visualization technique that combines a BiLSTM-based intrusion classifier with a heatmap-based risk monitoring interface. The proposed system processes enterprise network traffic, estimates anomaly probability, and maps divisional risk levels into a real-time visual dashboard for faster compromised-zone identification. Experiments on the CIC-IDS2017 dataset show that Neural-Heatmap achieves 98.86% average accuracy, 98.82% precision, 98.76% recall, and 98.79% F1-score under 10-fold cross-validation. It also reduces compromised-zone tracking time by 89.82% compared with log-based inspection. These results indicate that Neural-Heatmap can support faster cybersecurity monitoring and response in multi-division enterprise environments.