This research aims to solve a critical problem in the construction sector, particularly in infrastructure projects, namely the efficient extraction of information from a myriad of dispersed data sources, such as PDFs, CSV files, Excel sheets and CAD models, which are integral to the sector’s operations. The importance of this research lies in its potential to streamline data access in an industry that is both information-sensitive and characterized by significant data management challenges. The main objective is to develop a question-and-answer (QA) system for infrastructure elements, such as bridges, enabling fast and efficient extraction of information from a collection of online and offline data sources in different formats, with an interface integrated with the digital twin web platform. A key aspect of this project is the consideration of data security, ensuring that the system allows different data access permissions depending on the role of the user (e.g. manager, inspector, external company and project manager) involved in the projects. This is a crucial requirement in the construction sector, particularly for sensitive projects. The proposed methodology involves evaluating and selecting the machine learning models best suited to this specific task, taking into account both their information retrieval performance and their ability to incorporate robust safety measures. This research not only fills a gap in current applications of quality assurance systems in the construction sector, but also contributes to the wider field of AI in data-sensitive industries, paving the way for safer, more efficient and industry-specific information management solutions.