Within construction, we have become increasingly accustomed to relying on the benefits of digital technologies, such as Building Information Modelling, to improve the performance and productivity of projects. We have, however, overlooked the problems that technology is unable to redress. One such problem is rework, which has become so embedded in practice that technology adoption alone can not resolve the issue without fundamental changes in how information is managed for decision-making. Hence, the motivation of this paper is to bring to the fore the challenges of classifying and creating an ontology for rework that can be used to understand its patterns of occurrence and risks and provide a much-needed structure for decision-making in transport mega-projects. Using an exploratory case study approach, we examine ‘how’ rework information is currently being managed by an alliance that contributes significantly to delivering a multi-billion dollar mega-transport project. We reveal the challenges around location, format, structure, granularity and redundancy hindering the alliance’s ability to classify and manage rework data. We use the generative machine learning technique of Correlation Explanation to illustrate how we can make headway toward classifying and then creating an ontology for rework. We propose a theoretical framework utilising a smart data approach to generate an ontology that can effectively use business analytics (i.e., descriptive, predictive and prescriptive) to manage rework risks.