Electricity shortages and load-shedding create challenges for reliable electricity management when available capacity is insufficient to supply all connected loads. This paper presents a low-cost, proof-of-concept Internet of Things (IoT) system for priority-aware electrical load management that integrates real-time sensing, lightweight Artificial Intelligence (AI)-based demand prediction, and deterministic priority-based load control. The system estimates sector-level electricity demand using linear regression based on real-time consumption, time of day, sector type, and historical load patterns, while predefined priority rules allocate available capacity between industrial and residential loads. During a four-hour test, the system achieved an average network response time of approximately 48 ms, 99.6% data-transmission success, and 99.8% reported system uptime. Under emulated capacity constraints ranging from 100 W to 400 W, sensor readings remained within ±2% of expected values. The retained AI results comprised more than 1,000 data points, with a training time of approximately 2.5 s, an MSE of 0.021, and an R² of 0.94. The prototype reduced connected load by 25% during peak operation and 50% under an emergency condition.