Machine Learning (ML) has demonstrated significant potential in facilitating automation through its scalable predictive capabilities. Smart Industry represent transformative phase in industrial processes, characterized by smart connectivity and automation. Given the complexity and scale of industrial challenges, ML presents a promising solution; however, gaining a comprehensive understanding of its role in smart industry requires synthesizing an extensive body of research. This systematic review analyzes 45,863 papers from Scopus and Web of Science using BERTopic to identify major research themes and ML methodologies. Additionally, a manual review of 10 consulting white papers provides an industry perspective. Results highlight security and predictive maintenance as dominant topics, with convolutional neural networks as the most used ML method. Notably, smart industry stakeholders prioritize the successful adoption and integration of ML technologies over the development of novel ML models. While existing research themes remain highly relevant, future work should emphasize technologies that facilitate ML adoption and deployment to bridge the gap between academic advancements and real-world industrial implementation.