Unveiling Drivers and Barriers to the Application of Large Language Models in the Construction Industry: A Case Study of China

Abstract

The rapid advancement of artificial intelligence, particularly Large Language Models (LLMs), presents transformative opportunities and challenges across industries. In construction, LLMs hold significant potential to optimize project management through intelligent solutions. Yet barriers such as model hallucination, data privacy risks, and high implementation costs impede widespread adoption. This study systematically investigates the drivers and barriers influencing LLM adoption in construction, while assessing current adoption levels and industry willingness. Through a structured literature review, we identify 17 drivers (Company, Value Creation, Technology, Safety and regulations, and Service levels) and 19 barriers (Domain-Specific (construction industry), Technology, Adoption, and Ethical levels). Subsequent questionnaire deployment and quantitative analysis yielded significant findings: the three most influential drivers originate at the Company level, underscoring organizational leadership's critical role in LLMs implementation. In contrast, the predominant barriers emerge from the Domain-specific (construction industry) level, revealing sector-unique challenges as primary adoption constraints. While survey respondents predominantly express optimism regarding LLMs applications in construction, practical implementation remains at a nascent developmental phase. The study concludes by proposing tailored implementation strategies for construction enterprises and policy stakeholders, while contributing actionable insights for subsequent academic investigation and industry adoption.

Keywords: Large Language Models (LLMs), Construction Industry, Drivers, Barriers, Entropy weight method