The volume and variety of data from streams of sensor data to the unstructured text contained in emails have created a disjunction between when data is available and when it is utilized in timely and trustworthy ways. Knowledge that is not used is inert. This paper provides a closed-loop cognitive architecture for changing data into information, information into knowledge, knowledge into action, and action outcomes into better future knowledge and decision-making. It uses ontologies, knowledge graphs, temporal logic, probabilistic models, deep learning, anticipated utility, robust optimization, and reinforcement learning.