The integration of artificial intelligence (AI) into Building Information Modeling (BIM) is crucial to accelerate the digital transformation of the construction sector. Scan-to-BIM is a process that integrates 3D laser scanning with BIM to produce data-rich digital representations of physical environments. The captured spatial data points are assembled into point clouds, which are then converted into As-Built BIM models. Despite significant advancements, challenges persist, particularly in the classification and identification of architectural, structural and engineering elements and in the management of irregular geometries, non-standard industrial components and incomplete scan data. Our methodology combines a systematic literature review with an in-depth factory case study to examine current Scan-to-BIM challenges. The case study involves the renovation of a factory, in particular its external facades, using Scan-to-BIM process. Through this case study, we examined the difficulties involved in identifying and classifying architectural, structural and MEP elements. The finding of the literature review indicate that AI can significantly accelerate Scan-to-BIM workflows by automating point cloud segmentation, object recognition and semantic classification. However, AI performance is limited when faced with complex geometries, incomplete data and atypical industrial components. Consequently, this study advocates a hybrid, Human-in-the-loop approach, in which human-machine collaboration promotes an active synergy between the computational capabilities of AI and human intelligence. AI operates as an intelligent assistant capable of managing repetitive, large-scale, or computationally intensive tasks, while human professionals provide interpretive reasoning, domain-specific expertise, creativity, and ethical oversight. This hybrid framework with AI-assisted automation strategies provides practical guidance for Scan-to-BIM workflows.