Visual AI has significant potential to improve environmental monitoring, but scaling and sustaining such systems remains challenging. AquaWatch Solutions has addressed this by developing a robust, scalable visual AI framework to detect and manage illicit stormwater discharges across Auckland, in partnership with Auckland Council. This work demonstrates how complex AI can be operationalised into an accessible, reliable tool for surface water contaminant management. Unlike many visual AI solutions that remain confined to pilot projects or require intensive manual oversight, this framework prioritises adaptability, automation, and integration with regulatory workflows. A modular system architecture supports continuous model retraining, region-specific customisation, and efficient edge-to-cloud data processing, enabling deployment across diverse catchments with minimal resources. Emphasis on transparency, auditability, and user-focused interfaces ensures regulators and compliance officers can operate advanced camera networks without specialist AI expertise. A growing network of smart cameras across Auckland automatically detects, classifies, and logs discharge events in near real time, supporting rapid investigation and intervention by the Regional Council. The system has already reduced reliance on community complaints and chance discovery, enabling targeted compliance monitoring and proactive engagement that contributes to improved surface water quality. The presentation explores both technical and policy dimensions, highlighting lessons from scaling environmental visual AI, including co-design with regulators, intuitive interfaces, and reducing model training overhead.