Answers are grounded in this paper and cite their sources.
Purpose The integration of artificial intelligence with Lean construction has been shown to promote added value and drive growth in the built environment. However, understanding of their interaction across multiple levels of abstraction remains limited, especially beyond technical optimisation. This article systematically reviews the interaction between AI and Lean construction to identify current interactions, major applications and research gaps. Design/methodology/approach Following the PRISMA protocol, 63 peer-reviewed studies were identified and analysed qualitatively. AI techniques were mapped against Lean construction principles, waste categories and Lean methods to enable a multi-level comparative synthesis. Findings The results show a growing interest over the past four years, with a focus on AI techniques such as machine learning, optimisation and computer vision used to enhance workflow and to reduce cycle times, particularly within the Last Planner System. Nonetheless, principles centred on value creation, collaboration and partner development remain largely underexplored. Similarly, a co-adaptive relationship was observed to be emerging, in which AI improves Lean's technical efficiency while Lean principles guide AI's development towards ethical and value-driven outcomes. Research limitations/implications The findings offer a foundation for advancing Lean Construction 4.0 through socio-technical integration and human-centred innovation. Practitioners can start by applying mature AI techniques to improve measurable Lean outcomes such as workflow reliability, constraint prediction, visual management and waste reduction, then extend GenAI and NLP tools to collaboration, requirement capture, knowledge reuse and cross-project learning. Originality/value The study proposes a dual-focus model that separates AI's current role as a technical enabler from its future potential as a holistic partner.
No takes yet. Share an insight, caveat, or question.
Akande et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: