Fusion of machine learning to enhance the adaptability of lean construction maturity models (lcmms)

Purpose – Lean construction maturity models (LCMMs) offer a structured approach to guide the implementation of lean practices in construction organisations. However, in developing countries such as Sri Lanka, their adaptability is constrained by contextual barriers and limited technological integration. Although numerous emerging technologies are available, the integration of machine learning (ML) into LCMMs remains largely unexplored. Thus, this study aims to investigate how ML can be fused with LCMMs to improve their adaptability in the Sri Lankan construction context. Design/methodology/approach – Rooted in pragmatism, the research adopted a qualitative choice, using the Delphi technique. Three rounds of semi-structured interviews were conducted in eight phases to collect data from 25 experts. Data were validated via two cases and analysed using code-based content analysis. Findings – The study identified a seven-stage LCMM, with the seventh level being the "Lean Ideal Level". In total, 25 barriers were mapped across these seven stages. To address these, ten ML attributes were identified, and their suitability was assessed across each level. A comprehensive integration framework was developed, outlining relevant ML tools, techniques and fusion methods. Organisations at early maturity levels face more barriers, which decrease with progression, while ML attribute suitability shows the opposite trend, with fewer applicable attributes in early stages and more at advanced levels. Originality/value – To the best of the authors' knowledge, this is the first study to bridge ML and LCMMs for developing countries, offering a structured, evidence-based framework that aligns ML capabilities with maturity stages. It advances theoretical understanding of LCMM adaptability while introducing a novel integration pathway tailored for resource-constrained contexts.