Construction projects frequently face delays, presenting significant challenges to effective project management. Identifying the causes of these delays and determining which stakeholders are responsible for addressing them is crucial for effective mitigation. While progress reports from construction sites offer valuable insights into project status, they are primarily used for contract management. Consequently, despite containing crucial data on project delays, they are often underutilised for delay management. Moreover, manual comprehension and analysis of these reports further complicate the identification of delay causes, hindering thorough analysis. To address this issue, this research proposes an automated delay inference method using natural language processing (NLP) techniques. By leveraging large language models (LLMs), the aim is to extract critical project information, including delay causes, from 44 weekly progress reports. These causes are then categorised based on their location and nature. Finally, a flowline diagram is generated to visualise the planned and actual construction programmes. The gaps between them are explained by the identified delay causes and their impact on project completion, providing instant insights for effective decision-making. This approach aims to enhance the accuracy and efficiency of delay analysis in construction projects, ultimately improving project management.