Digitization of buildings is concerned with converting buildings information into digital formats,and improving various processes from designing and construction to operation and maintenance,so as to achieve energy efficiency and operational flexibility. Much of the available metadata,however, exists in unstructured, non-machine-readable formats such as control cards, blueprints,and system diagrams. This thesis project aims to investigate the possibility of automating or semi-automating the extraction and mapping of building metadata using a combination of AI and BrickSchema. The work starts from evaluating various extraction methods, including optical characterrecognition (OCR), large language models (LLMs), and vision-based AI, and then comparingtheir performance across control cards from three different companies with varying formats andcomplexity. The evaluation results show that vision-based AI has the best performance and bestaccuracy, handling diverse and complex document structures, while OCR-based methods requiremanual intervention to achieve useful results. The proposed pipeline shows strong potential forscaling metadata extraction and mapping across buildings, leading to significantly reducing thetime and effort required for digital transformation in the building sector. After the evaluation,validation of the methods is conducted by comparing the automated brick schema with a manuallycreated one, the results of which show a similarity of the two schemas up to 84.25%. The future work of this thesis is to investigate different approaches of extracting information anddifferent methods of storing the buildings information.