In risk management, critical controls are processes that are put in place to prevent or mitigate the effects of material unwanted events (MUEs). One of the least effective categories of critical controls is administrative controls. This category of critical control encompasses manual, policy-based, and procedural controls. These manual inspections are ineffective as they are subjective and only provide point-in-time assurance. The application of computer vision, IoT, and predictive AI can help automate this type of controls and significantly reduce safety risks. In this research study, two applied experiments are performed in an industrial nickel refining plant to validate the effectiveness of this technology. For the first experiment, a system is developed to monitor the flow of molted material during granulation and detect any buildup that could potentially cause a risk-event. For the second experiment, a system was developed to continuously monitor the ambient brightness of the reduction processing area and use the collected data to predict potential risk-events before they happen.