Achieving Trustworthy Real-Time Decision Support Systems with Low-Latency Interpretable AI Models