Smart grids (SGs) revolutionize existing power grids by using a wide range of developing disruptive technologies to generate clean, efficient, and predictable energy. Our study uses an action research method and focuses solely on the first two stages of the action research process, diagnosis and action planning, to evaluate ways to adopt artificial intelligence (AI) applications in SGs for predictive analytics in practice. The diagnosis stage of the study entails conducting a systematic literature review on AI applications in SGs, highlighting four areas of potential for predictive analytics: power outage prediction, demand response, control and coordination, and AI-enabled security to optimize decision-making, diagnose faults, and improve grid stability and security. The action planning step included a document analysis to devise methods to enable the practical implementation of AI in smart grids for predictive analytics. Finally, we address practical ways for implementing transparent AI for predictive analytics, followed by a conclusion and future research direction. The study’s key conclusion is that more research is needed to complete the action taking (implementing the solution), evaluation (assessing the results), and learning (reflecting on lessons learned) phases of the action research cycle.