AI in Construction Budgeting: Review of Trends, Tools and Limitations

Abstract

Construction budgeting is an essential aspect of planning a project. The traditional methods of estimation come with a host of challenges, including inaccuracies, outdated manual procedures, and inability to adapt to shifting market conditions. This paper investigates the development and present state of Artificial Intelligence (AI) in construction budgeting, using empirical and review literature published from the year 2000 to 2023. Critical AI methodologies, including support vector machines (SVM), artificial neural networks (ANN), ensemble methods such as XGBoost, and natural language processing (NLP) alongside hybrid models, are discussed with regard to their AI-driven automation and adaptability capabilities as well as their predictive accuracy. Ensemble and hybrid models are increasingly used to enhance predictive accuracy in construction budgeting. The results show that modern ensemble and hybrid models, particularly XGBoost, outshine regression-based models in predictive accuracy and adaptability, with high R² coefficients (sometimes exceeding 0.95) and low mean absolute percentage errors around 9%. Moreover, NLP tools excel in parsing contract documents and cost item organization, while automated quantity takeoff is provided by computer-vision and BIM-integrated systems. However, data quality, lack of trust in AI algorithms (due to their opaque nature), prohibitive costs of implementation, and a lack of transparency in automated processes remains challenging. Practitioner feedback underscores that, while AI offers efficiency gains, it must function as a human-assisted “copilot” rather than a full substitute for expert judgment. The paper concludes with recommendations to promote data governance, hybrid workflows, pilot deployment, transparency, and stakeholder skills development—paving the way for responsible and scalable AI adoption in construction budgeting.