Optimization of Decision-Making in Multi-Stakeholder Architectural, Engineering, and Construction (AEC) Projects: A Novel Data-Driven Framework

Abstract

Abstract : The complexity and dynamic nature of regulatory compliance in the Architectural, Engineering, and Construction (AEC) industry create significant challenges for stakeholders, often leading to delays, inefficiencies, and costly errors. Variations in interpretation, reliance on tacit knowledge, and evolving regulations further complicate decision-making. This study addresses these challenges by developing a custom GPT model, identified as the best fit for small datasets and specialized Q&A applications, to provide instant, accurate, and context-specific answers to regulatory queries. The study is applied within the Kingdom of Bahrain, utilizing Chapter 14 of the Bahraini Unified Guidebook of Building Permit Regulations as its primary data source. The study compares multiple LLMs to determine the most effective model for handling a small, domain-specific dataset related to regulatory compliance. The Custom GPT model achieved 100% across all performance metrics, while the base model, ChatGPT-4.0, recorded an accuracy score of 98.8%. These results demonstrate that CustomGPT outperforms other models by delivering more precise and contextually relevant answers, making it the most suitable choice for developing an AI-driven regulatory compliance assistant. This study introduces and employs a systematic 10-step methodology that integrates human-AI collaboration, expert feedback, and domain-specific fine-tuning to develop a scalable compliance framework. The study is grounded in Knowledge Management (KM) Theory, facilitating the transformation of tacit expert knowledge into structured, actionable insights Expert interviews were conducted with four out of six domain specialists from the authority overseeing Chapter (14), representing a sampling rate of 66.7%. The findings confirm that Custom GPT significantly improves decision accuracy, accelerates access to regulatory knowledge, and fosters collaboration among AEC professionals. The model serves as a centralized knowledge hub, bridging the gap between complex regulations and practical implementation. Its scalability allows for potential integration across additional regulatory chapters and broader industry domains, supporting digital transformation in compliance management. Despite its effectiveness, challenges remain in data availability, generalizability across jurisdictions, and handling complex interpretive queries. Future research should explore multilingual capabilities, adaptive AI frameworks, and integration with Building Information Modeling (BIM) systems to enhance real-world applicability. This study contributes to AI-driven compliance solutions in the AEC industry, demonstrating the transformative potential of data-driven methodologies in regulatory adherence, decision-making optimization, and stakeholder collaboration