Driving net-zero construction through evolutionary machine learning in Vietnam: a strategic framework for sustainable performance

Construction industry is expected to comprise 14.7% of the world's GDP by 2030 (Iqbal et al., 2022). However, it faces criticism for negative impacts on environment and society, including high energy consumption, high greenhouse gas emissions, poor working conditions and high rates of worker injuries (Craveiro et al., 2019; Zhang et al., 2025). Its dual goals of increasing business performance while maintaining carbon neutrality challenge sustainability initiatives of both organisations and governments. Technological integration in the construction sector has been found to exert a positive contribution to carbon emission reduction (Regona et al., 2024). Firms thereby target modernising existing construction sector systems and improving implementation of construction projects to more sustainable levels by integrating new technologies into conventional processes and products (Maqbool et al., 2023a, b).

Construction 4.0 (C4.0) is one technological application that incorporates the use of digital technologies to assist in achieving net-zero goals in the construction supply chain. These technologies comprise Industrialised Building Systems, Building Information Modeling, Augmented Reality, Virtual Reality, Internet of Things, Digital Twin, Cloud Computing, and Blockchain (Ohueri et al., 2025; Oyejobi et al., 2024). EML, a subset of artificial intelligence (AI) within evolutionary computation, has emerged as a potential alternative for optimising complex systems. EML is inspired by nature-based mechanisms and genetic algorithms of evolutionary computation to deliver optimal solutions for multiple objectives (Carvalho et al., 2024). Unlike traditional machine learning approaches, EML's iterative operational characteristics are highlighted due to its effectiveness in dealing with uncertainties and dynamism in construction activities (Boulesnane, 2024). It can overcome the complexity inherent in the construction sector by not relying on gradient-based optimisation approaches, which require differentiable objective functions and well-structured, labeled datasets (Boulesnane, 2024; Yuan et al., 2024). EML has been integrated with BIM projects to leverage both visual programming and evolutionary algorithms to create comprehensive optimisation frameworks that minimise weight, cost and environmental footprint simultaneously (Yavan et al., 2024) which cannot be performed by gradient-based approaches solving single-objective problems of traditional machine learning. Yet, the performance influence of EML on sustainability in the construction sector remains underexplored, leaving a gap in sustainability literature. Hence, greater focus is necessitated on achieving carbon footprint reduction with the facilitation of EML-enabled technologies.

However, the construction sector is notable for its low levels of advanced technology integration due to perceived high costs, excessive training requirements and risk of disruption to existing operations (Newman et al., 2021). Developing countries especially face barriers when incorporating AI into their sustainability strategies (Wang and Guo, 2022). Initially, firms need to invest substantially, but it takes a considerable length of time to obtain business returns from AI-enabled projects, thus placing great pressure on firms from top-to-bottom-line stakeholders (Huynh-Xuan et al., 2024; Luqman et al., 2024). The post-adoption phase when integrating AI-based solutions faces further significant challenges, including financial constraints (Pham et al., 2020), resistance to change (Huynh-Xuan et al., 2024) and complex legacy frameworks (Regona et al., 2024), which hinder the application of new technologies in building projects. Nevertheless, firms continue to place their hope in AI due to recognising its potential benefits in both finance and carbon reduction despite these challenges (Luqman et al., 2024). Consequently, it is essential to have a structural framework on how to effectively integrate EML-enabled technologies into construction projects to assist companies in sustainability transition progress. More research has been called for investigating how EML-enabled technologies can be adopted (Balasubramanian et al., 2024; Regona et al., 2024), especially in developing countries (Statsenko et al., 2023).

To bridge the research gap, this study employs Ecological Modernisation Theory (EMT), Adaptive Structurisation Theory (AST), and Diffusion of Innovation (DOI) to build a comprehensive framework on how to adopt EML-enabled technologies to improve carbon emission reduction and business performance in the construction sector with the concentration on post-adoption phase. This novel theoretical integration transcends the limitations of each individual theory, providing a multilevel analytical lens to examine how EML-enabled technologies transform construction operations toward carbon neutrality while maintaining business viability.

In this study, we argue that applications of EML-enabled technologies in the construction industry can improve carbon reduction initiatives through resource consumption optimisation, carbon footprint reduction, and supply chain resilience. They can further stimulate the business performance of organisations sustainably in return. Their adoption attributes, relative advantage, observability, compatibility, complexity and trialability, are initially vital for construction firms in determining how to adopt and integrate with existing operations to maximise the benefits of those solutions. This study aims to answer two research questions:

To address the research objectives, this study applies a quantitative approach by using a survey for data collection and Structural Equation Modeling (SEM) for data analysis processes. A cross-sectional survey design is particularly suitable for attaining a broad response in a time-efficient manner, especially ensuring adequate representation of construction firms (Bell et al., 2022; Wang and Cheng, 2020). Hence, the survey enables this research effectively to capture understandings about firm-level practices and outcomes for the research objectives. Besides, SEM is chosen as an analysis method because it is effective for testing and validating the theoretical framework (Hair et al., 2021). SEM enables the simultaneous examination of multiple linear relationships among latent constructs, making it well suited for assessing the effects of EML-enabled technology integration on carbon neutrality and business performance.

By addressing these questions, this study extends sustainability literature by centralising applications of EML-enabled optimisation in driving net-zero construction supply chains, filling the gap in prior research (Regona et al., 2024; Statsenko et al., 2023). Concurrently, by integrating multiple theories such as AST and DOI, this research delivers multi-level perspectives on how the interaction between technology and humans can drive successful performance rather than centralising the technological role while abandoning the human role, contributing to literature on technological adoption and integration in business (Balasubramanian et al., 2024; Regona et al., 2024). Meanwhile, in practical terms, it provides insights and a structured framework for companies in managing the technology integration process, supporting the transition, especially in developing countries where construction activities are still performed manually. In particular, by analysing the post-adoption stage, it highlights key factors that influence long-term success, helping early adopters plan operational adjustments and non-adopters plan better adoption management.

The paper is organised as follows: Section 2 reviews key literature and outlines our hypotheses developed. Section 3 covers the methodology of study. Section 4 and 5 then discuss the data analysis and findings. Section 6 presents implications for theory and practice. Section 7 concludes the article with a discussion of limitations of the present study and directions for future studies.

The circular economy has emerged as a development model for carbon neutrality by replace the traditional linear economy to promote resource efficiency, waste reduction, and renewable production systems through concepts such as industrial symbiosis, industrial ecology, cradle-to-cradle and reverse logistics (Geissdoerfer et al., 2017; Nobre and Tavares, 2021). In the construction industry, recent studies have increasingly positioned digitalisation as a key enabler of circular economy and net-zero transitions. Rather than focusing solely on operational efficiency, digital technologies such as BIM, Digital Twins, IoT, RFID, Blockchain, and AI support circular construction practices through material traceability, lifecycle assessment, construction waste reduction, material recovery and design-for-disassembly strategies (Banihashemi et al., 2024; Oyejobi et al., 2024). For example, decentralised and immutable natures of blockchain allow stakeholders to streamline products and materials of a building throughout their lifecycle, fostering transparency and sustainability in construction supply chain (Singh et al., 2023b). Long et al. (2025) further identified ten digital functions and fifteen circular construction strategies as core enablers of circular construction and demolition waste management to support circular outcomes and carbon emissions across the building life cycle.

Particularly, EML-enabled technologies become increasingly prominent in promoting circular transition to pursue carbon neutrality in construction industry. It is inspired by Evolutionary Computation (EC) and Machine Learning (ML), which operates on population-based approaches, iterative refinement and genetic operators to process optimal solutions for complex multi-objective problems (Carvalho et al., 2024). It iteratively learns from the population of solutions through models and algorithms, then generates optimal solutions meeting multi-objective requirements (Boulesnane, 2024). EML-enabled applications have also been proven to facilitate accurate resource allocation, effective energy consumption, proper waste management, and ensure real-time adherence to sustainability standards by overseeing life-cycle assessments (LCA) to support economic and environmental objectives of organisations. In terms of economic performance, Nikoukar and Tavakolan (2025) developed a simulation method based on genetic algorithms and discrete event simulation models to optimise resource allocation and logistics in construction projects. By considering project limitations, site layout characteristics, warehouse locations and production capacities, this method significantly reduced costs compared to traditional contractor scheduling methods. Sallam et al. (2019) demonstrated that the combination of evolutionary algorithms with Cuckoo search exhibited superior performance in solving resource-constrained project scheduling problems in construction by using genetic selection mechanisms based on solution quality and population diversity to achieve high-quality results. The integration of evolutionary algorithms with LCA tools in BIM projects has been also shown to promote sustainable construction by incorporating integrated energy calculations and material life cycle considerations to support long-term environmental performance (Yavan et al., 2024). Significantly, evolutionary algorithms help project planners attain circular economy by optimising the industrial networks and the location of organisations within the eco-industrial parks to promote industry symbiosis and create a more sustainable industrial system (Genc, 2025b; Genc and Kurt, 2024).

However, empirical study on digitalisation-enabled construction for carbon neutrality and sustainability remains limited in developing countries (Lindblad, 2026). In these countries, construction and demolition waste continues to increase, while waste management infrastructure and regulatory systems are still under development. Therefore, integrating EML-enabled technologies within this stream enables this study to address a recognised gap by demonstrating how intelligent technology can simultaneously advance carbon-neutrality targets and business performance for construction firms operating under such conditions. It then provides the guidelines for orgnisations on how to effectively operate and manage this integration and digitalisation process while pursuing carbon neutrality and sustainability.

The synthesis of DOI, AST, and EMT theories creates a multi-dimensional framework for analysing the adoption of EML and its implications for sustainability in construction, DOI supports this study by explaining how new technologies such as C4.0 technologies spread and diffuse in organisations through critical factors including relative advantage, compatibility, observability, complexity, and trialability (Rogers, 2003). However, DOI does not sufficiently address structural adaptation processes or the nuanced environmental impacts of such innovations (Oorschot et al., 2018). When integrating advanced technology into existing operations, technology does not easily work alone in delivering targets. AST highlights the interaction between technology, individuals, and organisations, assisting in creating diverse views on technology applications, which in turn leads to different organisational performance (Martin et al., 2022; Wang et al., 2023). This framework includes concepts such as faithfulness and consensus, revealing how employees and management navigate the changes associated with integrating new innovations (Poole and DeSanctis, 1990). Nevertheless, AST has limitations in explicitly incorporating environmental considerations, which can be addressed by incorporating complementary frameworks (Birken et al., 2012). EMT, complementarily, asserts that resource-efficient practices and an emphasis on environmental consciousness enable firms to achieve significant operational and environmental objectives, highlighting a crucial intersection with contemporary sustainability demands in corporate practices (Bugden, 2022). Consequently, in synthesising these three theories, this research contributes to a nuanced understanding of how EML adoption attributes influence the structural adaptations of construction firms and their subsequent performance related to carbon reduction and overall business effectiveness.

DOI argues that innovation diffusion can be evaluated through five dimensions, including relative advantage, compatibility, observability, complexity and trialability (Rogers, 2003). In terms of technology attributes, relative advantage describes the superior performance of adopting new technology compared to existing practices, while observability refers to the ability to perceive the expected benefits of the technology. Meanwhile, focusing on the interaction between technology and people, compatibility refers to the consistency between emerging technology and existing norms, values and organisational structures. Additionally, complexity refers to the level of difficulty users perceive in understanding and using a new technology, and trialability refers to pre-evaluation and step-by-step testing before full adoption.

Construction firms can firstly take advantage of EML in data analytics and complicated algorithms to deliver actionable insights, make timely adjustments and reshape the building process for sustainability adherence (Regona et al., 2024). It has also been proven for its optimisation benefits to enhance the effectiveness of resource allocation to reduce carbon footprints (Mamun, 2018; Regona et al., 2024). EML has been indicated to help construction projects use materials and resources more effectively and reduce waste compared to conventional approaches to achieve more environmentally friendly performance (Hussain et al., 2024; Nikoukar and Tavakolan, 2025). By capturing the relative advantage in EML-enabled solutions in mitigating damaging factors towards the environment, construction companies are more inclined to replace existing operations with EML into their projects, supporting the transition to net-zero emissions supply chains.

Second, the core idea of EML is training machines with large datasets to assist the anticipation of tasks with a more data-based foundation (Genc, 2025b). Evolutionary algorithms have demonstrated the value of uncovering hidden gaps through genetic operators on the population of solutions to attain optimal solutions for multi-objective problems such as balancing resource, energy and waste management with quality (Carvalho et al., 2024; Sallam et al., 2019; Yavan et al., 2024). Following that, the iterative operation of EML further strengthens operations and minimises shortcomings of manual observations due to bias in conventional management methods to attain environmental performance. It can enhance project outcomes through better decision-making process by delivering preliminary insights and recommendations (Genc, 2025b; Regona et al., 2024) and optimise energy usage and carbon footprints during the building phase (Mamun, 2018). Observing these gains, organisations widely diffuse EML-enabled technologies to fully embrace carbon neutrality and enhance construction supply chain sustainability.

On the other hand, to ensure successful integration, AI-powered solutions must be aligned with business requirements and defined sustainable objectives, coordinated with operating governance mechanisms, and updated regularly to adapt to changing environments (Singh et al., 2023a). Especially, synthesised interaction between AI and human expertise is highlighted as a critical determinant of integration success and value delivery (Genc, 2025a, b). Yet, the construction sector is notorious for its low-level technological integration and low-skilled employees (Newman et al., 2021; Turner et al., 2021), especially in the constructing works sector. Hence, established operation procedures and processes mainly serve existing manual work, which may be unable to support projects with EML-enabled integration. It has been noted that integrating AI into existing processes becomes more difficult when firms fail to realise the alignment of AI initiatives despite resources, time, and energy investment (Saleem et al., 2024). Therefore, organisations need to develop regular assessments and adjust procedures for seamless AI-enabled operations to assist carbon mitigation objectives.

Although EML harnesses great benefits for net-zero construction advancement, the perceived complexity from integration arising from stakeholders may prolong adoption. Inconsistent legal systems and an insufficiently skilled workforce mainly induce challenges (Turner et al., 2021). For instance, organisations must deal with the challenges of ensuring the alignment of AI outputs with regulatory standards when integrating AI into mapping the ecosystems to support mapping ecosystems for the industry symbiosis (Genc, 2025b). Hence, they should know how to integrate regulatory constraints into algorithms to ensure compliance with standards and requirements and avoid undesirable violations. Construction firms also deal with employees' skepticism and resistance due to fear of potential replacement by AI (Singh et al., 2023a). Stakeholders' inability to interpret the benefits and usage of AI lowers their confidence in adopting the transformation (Statsenko et al., 2023). Hence, to accelerate successful integration, coordinated efforts and collaboration between firm strategies, AI experts, stakeholder support, and employee engagement are crucial to facilitate positive AI integration for reducing carbon footprints (Genc, 2025b; Shaik et al., 2024). Understanding the complexity of AI integration, firms can actively lower the barriers to innovation embracement, which then develops more efficient building processes for neutralising carbon emissions.

Finally, construction firms may face challenges when starting to employ AI solutions, including uncertainty regarding processing and the functions of AI as well as the exploitation of breaches resulting in operational process breakdown (Turner et al., 2021). The risks associated with construction projects vary according to different stages of the project life cycle (Zhao et al., 2010). Technology application is hence discussed to be flexible and gradually adjusted at different stages to ensure viable integration (Zhang and Li, 2022). In this context, through trials, firms can estimate the risks of implementation and foresee carbon reduction estimations thanks to the preliminary insights processed from large amounts of available data (Genc, 2025a). This further promotes comprehensive functionality of EML in capturing footprint reduction performance when adopted on a large scale.

Moreover, technological innovation directly promotes the business performance of organisations by focusing on two pillars: economic and social. Technological innovation has been indicated to assist win-win strategies for firms in balancing economic productivity and environmental protection (Agyekum and Ali, 2025). Prior researchers have indicated that C4.0 supports driving sustainable values for individual firms and whole construction supply chains. With the use of EML, firms can reduce stock surplus costs by mitigating the bullwhip effect through effective resource allocation (Nikoukar and Tavakolan, 2025) and attain optimal economic value for market competitiveness (Genc and Kurt, 2024; Hussain et al., 2024). Regarding the social pillar, genetic algorithms can analyse various factors within the construction site, including safety risk levels, psychological expectations, mental health and morality, to map out various scenarios to support safety performance for employees at construction sites (Peng et al., 2023). The benefits of real-time tracking and EML also help companies create safe working conditions for employees at construction sites by having repetitive employee movement data to predict risks, improving progress productivity. Hence, integration of EML-enabled technologies innovates the operation process and elevates business performance. Consequently, witnessing relative advantages and observability of EML-enabled solutions adoption in promoting business performance, firms further expand integration on a larger scale.

EML-enabled technologies further interact with social structures in existing operations to create synergy for business performance improvement. Compatibility between emerging technologies and operating systems/objectives can deliver harmony to help firms achieve greater return values (Özbek et al., 2022; Singh et al., 2023a). Construction companies thus achieve labor productivity and efficient use of resources to be competitive in the market. Second, when dealing with arising complexities such as new technological knowledge, firms invest effort in training to upskill employees (Momade et al., 2024) and share knowledge with stakeholders in the supply chain (Zhang and Li, 2022). Consequently, the competencies of employees and business partners are improved to operate more efficiently. Firms can further develop strategic partnerships with key stakeholders to attain long-term financial returns (Mukherjee et al., 2023). Finally, the trialability of emerging technologies such as EML enables firms to avoid risks inherent in new operating procedures (Mamun, 2018). Trials thereby enable firms to make timely adjustments to operate smoothly and attain productivity in the late adoption stage. AST also indicates the importance of changes in resources to adapt to new structures (DeSanctis and Poole, 1994). Hence, compatibility, complexity and trialability attributes of EML adoption specifically contribute to competencies of workers and business partners and the operation process of construction firms, in turn, promoting their competitiveness in the market.

Moreover, carbon reduction initiatives have been proven to play a key role in driving business performance. The external environment can mediate construction firm strategies on how to implement operational practices. Firstly, companies are coercively urged by stakeholders such as clients and the government to initiate green practices to meet business objectives for themselves and overall net-zero economy goals for society (Mukherjee et al., 2023). In the environment-oriented context, firms delaying their moves in achieving carbon neutrality may create opportunities for competitors to gain market share in emerging markets (Özbek et al., 2022). Construction companies can observe the environmental focus of their clients, such as building owners, and capture carbon-concerning criteria from government policies. They then provide more innovative implementation practices such as carbon-reduced building progress, addressing carbon concerns to meet their stakeholders' demands. Acknowledging the advantages of environmental performance, management aspires to implement net-zero carbon emission practices with the aim of attaining competitive advantage and promoting long-term economic performance (Mukherjee et al., 2023). Hence, emission reduction initiatives act as drivers for overall business performance improvement by advancing competitiveness, stakeholder relationships, and policy adherence. The final hypothesis is proposed accordingly.

In summary, the hypothesis model is schematically depicted in Figure 1:

Vietnam's economy is growing robustly with a GDP growth rate around 7.1% in 2024 (IMF, 2025) with 2.7% contribution from industrial and construction sector (National Statistics Office of Vietnam, 2025a). The government thereby places greater concentration on the industrial and construction sector, leading to growth of 7.32% within the first quarter of 2025 (National Statistics Office of Vietnam, 2025b). Despite industry growth, Vietnamese construction companies still find it difficult to adopt digital technologies to pursue sustainability. Pham et al. (2020) pointed out that financial constraints, high investment costs, existing skill gaps among employees, and unavailability of resources have increased barriers for companies in the transition to a sustainable construction supply chain. Due to its unfavorable influence on the environment and the existing challenges, Vietnam's government has recently promulgated decisions concentrated on sustainable practices in the construction industry to respond to climate change issues:

1. Decision No.280/QĐ-TTg issued in 2019 concentrates on energy saving and efficient energy use execution in the 2019–2030 period. It especially emphasises the integration of information technology to monitor energy use (The Prime Minister, 2019).

2. Decision No.1266/QD-TTg issued in 2020 is concerned with the importance of building materials in greenhouse gas emissions mitigation and natural resources conservation with a target by 2050 (The Prime Minister, 2020).

3. Decision No. 258/QD-TTg issued in 2023 orders BIM integration in construction nationally, urging companies to adopt digital technologies (Ngoc and Xuan, 2024).

This active promulgation is a driver for digital development to promote a sustainable construction sector in Vietnam (Pham et al., 2020). With opportunities empowered by government regulations, it is timely to evaluate how EML-enabled solutions can improve carbon footprint targets while securing the economic aspect of building projects through their iterative refining solutions in the construction sector in Vietnam. Hence, samples used for data collection were construction firms employing EML and operating in Vietnam. Based on information provided by the government information page–General Statistics Office of Vietnam, 862 companies applying EML solutions were listed and contacted.

The online questionnaire survey was carried out to observe the impact of integrating EML-enabled solutions to improve carbon objectives in the construction sector. We concentrated on assessing responses from construction company managers based in Vietnam's key economic hub with excessive growth in infrastructure, the Mekong Delta region. The region's strategic development in both infrastructure and sustainability makes it relevant for studying EML-enabled transformation and its effect on sustainable business performance. Additionally, the measurement items were in Vietnamese to facilitate local participants' ease of understanding. To avoid misinterpretation and ensure comprehension, the survey then used a back-translation method by translating from Vietnamese to English and was minorly adjusted for clarity. Furthermore, to ensure that businesses actively participate in using EML-supported technology, the participants were asked a screening question if the EML-enabled technology remained adopted before answering the main questionnaire section.

An initial pilot test was conducted randomly with 100 firms to revise items as necessary for validity and reliability assurance. The modified survey design was then emailed to 862 executives working at targeted construction firms in Mekong Delta locations from June to October 2024. The survey was followed up with two rounds of reminder emails to improve response rates. We received 279 responses from participants with a 32.4% response rate. Finally, 213 data-cleaned responses were employed for further analysis after removing 66 incomplete responses. The finalised response met the required sample size for SEM analysis of 200 responses (Boomsma and Hoogland, 2001; Kline, 2023).

The employed measurement items were referred to previous relevant research, including Chinda (2017), Ozorhon and Karahan (2017), Olatunde et al. (2023), Wang and Guo (2022), Famakin et al. (2022) and Srimathi et al. (2017) as illustrated in Table 1. The survey was also designed and pre-tested with support from 10 academic experts and 11 managers of construction firms in Vietnam. The survey was separated into two sections. In the first section, the questionnaire assessed the perceived attributes that impact EML-enabled solutions adoption behavior of participants to assist their building projects within the last five years. It was measured with a five-point Likert scale ranging from 1 (Strongly Disagree) to 5 (Strongly Agree). The perceived attributes, including relative advantage, observability, compatibility, complexity and trialability, were measured through 26 items to study EML solution adoption behavior. In the second section, the questionnaire assessed sustainable performance of participating companies compared to other large companies in the same industry. It was measured with a five-point Likert scale ranging from 1 (Significantly Worse) to 5 (Significantly Better). Sustainability performance was measured using an 8-item scale focusing on three dimensions of sustainability performance: economic, social and environmental.

Covariance-based Structural Equation Modeling (CB-SEM) was chosen to study both direct and indirect relationships between constructs for its advantages in theory assessment and theory confirmation (Hair et al., 2011). The SPSS software packages, including SPSS Statistics 26 and SPSS AMOS 20, were then used to assist the data analysis process, including descriptive demographic analysis, measurement model analysis, and structural model analysis.

This paper concentrates on how EML adoption attributes influence carbon reduction initiatives and promote carbon footprint and business performance of construction companies. Specifically, by analysing DOI attributes (relative advantage, observability, complexity, compatibility, and trialability) and its influence on carbon neutrality, this paper can offer dual perspectives on both the integration process of EML-enabled technologies and its effectiveness reflected on the organisational performance. This following section provides the profile of the participant firms and results of the relationships between constructs.

Table 2 illustrates the descriptive information from 213 responding firms, of which 46% were private companies, 28.2% were state-owned companies, and the remainder were joint ventures. Nearly half of the responding companies (47.4%) were active in the design sector, followed by 22.1% in the construction works sector and 14.1% in the zoning sector in the construction supply chain. Two-thirds of participating companies had been in operation for 10–30 years, with 29.1% having been in business for 10–20 years, 19.2% for 20–30 years, and 18.3% for 5–10 years. The majority, 70% of companies, were small and medium-sized with fewer than 200 employees; in particular, 49.8% had fewer than 50 employees. Besides, respondents from participating companies came from both managerial and executive levels, 56.3% and 43.7% respectively, providing comprehensive insights into management and implementation practices when integrating EML. They mainly came from the design department (34.0%), project management department (28.2%), and management boards (20.6%).

To assess the potential for non-response bias, a t-test was performed to compare responses between early and late participants, following the method proposed by Armstrong and Overton (1977). The analysis revealed no statistically significant differences in the mean scores of all measured items, with a confidence level of 99%, suggesting that non-response bias was unlikely to be a concern in this study.

Furthermore, since both independent and dependent variables were gathered from the same respondent within each organisation, there was potential risk of common method variance (CMV). To examine this, Harman's single-factor test was employed (Podsakoff et al., 2003). An unrotated exploratory factor analysis including all measurement items was conducted. The emergence of a single dominant factor would suggest the presence of CMV; however, the results identified six distinct factors, indicating minimal risk. Still, recognising the limitations of this test when numerous items are involved (Podsakoff et al., 2003), a more focused analysis was conducted. Items from each independent construct related to EML adoption were combined with those of the dependent construct, sustainability initiatives, and subjected to factor analysis. In each case, two or more factors emerged, providing further evidence against significant CMV.

The results of the measurement model were assessed through reliability and validity tests, as illustrated in Table 3. An exploratory factor analysis (EFA) was conducted to measure reliability. Items were initially selected according to their overall association across the factor (item-total correlation). Accordingly, items with lower than 0.35 item-total correlations were removed. The remaining items were analysed using principal component analysis. All the factors defined by EFA had variances extracted from 60.9% to 86.6%. All constructs were shown to pass the reliability test by meeting the requirements of Cronbach's alpha (CA) and Composite Reliability (CR). Following Fornell and Larcker (1981), CA of all variables ranged from 0.844 to 0.908, above the minimum satisfactory level of internal validity of 0.7. The results indicated internal consistency among constructs since CR ranged from 0.850 to 0.886, higher than the required value of 0.7 (Hair et al., 2022).

A confirmatory factor analysis (CFA) was conducted to analyse validity among constructs. It is firstly important to evaluate model fit to supervise overall fit between data and the measurement model. To assess the fit, the result should satisfy the following criteria: particularly, χ²/df is less than 3 (Kline, 2023), root mean square error of approximation (RMSEA) is less than 0.08, and comparative fit index (CFI) is greater than 0.8 (Hair et al., 2009). The overall fit of the measurement model was satisfactory, where χ²/df was 2.11, RMSEA was 0.07, and CFI was 0.937.

Then, convergent validity assessment considers standardised regression weights, standard errors, and average variance extracted (AVE) of variables. As shown in Table 4, the results indicated that seven items from observability, compatibility, and sustainable performance were eliminated due to failing to meet the condition of standardised regression weights being above 0.5 and being at least two times the following standard errors (Hair et al., 2009). Moreover, AVE of factors ranged from 0.533 to 0.733, exceeding the threshold value of 0.5 (Hair et al., 2009), confirming convergent validity assessment. We then conducted a discriminant validity test to ensure that each construct in the measurement model was empirically significantly different from the others. We followed the procedures established by Fornell and Larcker (1981), Hair et al. (2022), Henseler et al. (2015), examining cross-loadings between constructs. Discriminant validity is confirmed when each indicator has a higher loading on its intended construct than any other related indicator (criterion A > B). As presented in the Appendix, the results showed that almost all indicators met this criterion, as each clearly indicated a distinct construct in the measurement model. For example, in the Relative Advantage construct, the Relative02 indicator had a primary loading of 0.873, while its highest cross-loading on another construct was 0.752. Only Observability01 deviated slightly (loading = 0.777; highest cross-loading = 0.812), suggesting slight conceptual overlap between Observability and Trialability. This slight deviation likely reflected semantic or perceptual similarity but does not significantly affected overall validity. Ultimately, the consistently high loadings between related constructs indicated good discriminant validity and convergent validity, providing a solid basis for evaluating the structural model.

Both reliability and validity of elements were satisfied. Furthermore, the measurement items of EML-enabled technologies adoption attributes and sustainability initiatives had R² greater than 0.3 (Hair et al., 2009), indicating considerable explanatory power for its corresponding endogenous variable.

Illustrated in Figure 2, the result indicated relationship assessment regarding the influence of EML-enabled solutions on carbon neutrality and business performance of construction firms. Only Observability (0.842, p < 0.01), Compatibility (0.156, p < 0.1), and Complexity (0.307, p < 0.01) were shown to positively impact carbon neutrality. The observable benefit of EML-enabled integration was the strongest factor driving net-zero initiatives. Meanwhile, the relative advantage of EML-enabled solutions did not significantly reduce carbon emissions (−0.025, p > 0.1), and trialability of EML-enabled solutions showed a significantly negative impact on carbon neutrality (−0.46, p < 0.01). Consequently, the findings supported only H1b, H1c, and H1d while not supporting H1a and H1e.

Conversely, business performance was revealed to be supported by Observability (0.378, p < 0.01), Complexity (0.144, p < 0.05), and Trialability (0.689, p < 0.01) of EML-enabled solution integration. The findings, however, indicated that Relative Advantage (−0.436, p < 0.01) and Compatibility (−0.503, p < 0.01) have significantly negative influences on business performance. The findings thereby supported hypotheses H2b, H2d, and H2e while rejecting hypotheses H2a and H2c. Moreover, net-zero performance was shown to enhance business performance (0.364, p < 0.01), providing evidence to confirm H3. Finally, these values indicated that the model explained 68% of the variance in carbon neutrality initiatives and 72% of the variance in business performance, suggesting moderate to strong explanatory power.

The results indicate how EML adoption attributes influence carbon reduction initiatives and promote carbon footprint and business performance of construction companies. The findings support that observability and complexity attributes of EML integration positively reduce carbon emissions and enhance business performance of construction firms, consistent with prior research (Mamun, 2018; Regona et al., 2024; Wang and Guo, 2022). Among these two attributes, observability exerts a greater effect on the improvement of carbon and firm performance. EML-enabled solutions for multivariate time series can help forecast energy consumption and provide real-time activity monitoring for greenhouse gas reduction to serve carbon reduction targets. In turn, these solutions improve stakeholders' satisfaction by achieving shared social values and delivering long-term economic feasibility. Nevertheless, if firms secure collaborative support from leadership, conduct training programmes, and promulgate well-structured policies, they can overcome adoption complexities to better achieve carbon-reducing operations and feasible economic values (Ebekozien et al., 2024; Zhong et al., 2025). Furthermore, this integration can have a greater positive impact on carbon neutrality goals than on business performance when businesses can successfully address the complexity and leverage the power of EML-enabled technologies. Therefore, these insights further confirm that these technological solutions greatly support the transformation of the construction industry towards sustainable development.

Meanwhile, the relative advantage of EML-enabled technologies is shown to have no impact on carbon performance initiatives and a negative impact on business performance. Concurrently, EML-enabled integration is moderately compatible with existing structures to accelerate carbon reduction in building projects but fails to enhance business performance of construction companies. Findings are not in line with Regona et al. (2024). EML normally requires acquiring a large amount of data, including extensive, consistent, and accurate details about the whole construction process. Meanwhile, the construction industry still heavily relies on manual activities (Momade et al., 2024) including in Vietnam, creating a fragmented and inconsistent dataset. The integration of EML thereby may be burdensome and beyond the handling capabilities of construction firms due to lacking professionally skilled workers, especially in Vietnam (Momade et al., 2024; Turner et al., 2021). Moreover, since participating firms in this research are relatively small-sized, they may face a lack of infrastructure resources and experience in integrating and implementing EML-enabled technologies effectively compared to existing systems. The current system of construction companies has low compatibility with the new system when introducing EML-enabled technology into operation. Subsequently, they cannot effectively capture optimisation benefits in carbon reduction and economic value from EML-enabled technologies compared to other operating systems due to moderate-to-low compatibility.

Moreover, the results indicate trialability of EML solutions before full adoption negatively impacts carbon neutrality initiatives but is beneficial to firm performance. Especially, among constructs, trialability holds the greatest influence on business performance of construction firms. This is inconsistent with the findings of Wang and Guo (2022) and Mamun (2018), which indicate EML-enabled technology trials can positively impact firm carbon and business performance. In the short term, trials of EML-enabled technologies can harm carbon footprint performance due to their extensive use of energy and high emissions in trial processes. Moreover, as EML-enabled technologies have only recently been introduced in Vietnam construction in 2020 due to government support policy and started to accelerate in 2023 (Dao, 2023), firms still fail to realise carbon reduction from the complicated operation process. However, the risks of construction projects vary according to different stages of the project life cycle (Zhao et al., 2010). Hence, in the long term, trials can still effectively help firms estimate risk and make prior restructuring before full adoption to ensure sustainable operations for business performance improvement.

Finally, carbon performance achievement is found to lay a strong foundation for business performance improvement of construction firms, aligning with previous studies (Meng et al., 2024; Mukherjee et al., 2023). By targeting net-zero objectives, construction firms successfully disclose their social responsibility in assisting government regulatory requirements for building a net-zero economy, thereby satisfying stakeholder demands. This concurrently helps companies enhance their reputation towards existing and potential stakeholders such as clients, subsequently having long-term economic value and competitive advantages through sustainable partnerships and support (Mukherjee et al., 2023). Hence, construction firms should concentrate more on incorporating innovative solutions for net-zero carbon operations achievement to capture sustainable business performance enhancement.

First, the findings empirically enrich the literature on carbon neutrality and firm performance by creating a comprehensive framework on how firms can balance green initiatives and competitiveness in the construction industry. They contribute to understanding how firms can simultaneously achieve business performance and net-zero performance. Drawing on EMT, AST, and DOI, firms can adopt more technical advancements such as EML-enabled technologies to optimise energy consumption, emission reduction and operational efficiency of building projects. Subsequently, firms can build a more positive image and reputation among stakeholders by contributing to the net-zero emissions economy for society's benefit. Firms can anticipate new green market demands and achieve long-term returns. Therefore, companies are encouraged to innovate their operations and systems to adopt more sustainable practices, which are beneficial not only for society and the environment but also for their organisational performance.

Second, this research addresses the gap by providing evidence regarding the implementation of EML-enabled technologies in the construction sector in a developing economy. Focusing on EML-enabled adoption in the post-adoption phase illustrates how organisations can function and streamline an AI-enabled system in construction projects to advance the net-zero economy. The findings indicate that the introduction of EML-enabled solutions is still an emerging concept for the construction industry in Vietnam. The integration of these technologies is nevertheless challenging due to incompatibility with conventional operating structures, resulting in low efficiency and productivity. The construction industry remains largely managed through manual activities in Vietnam. It is more costly for construction firms to operate EML-enabled technologies in Vietnam despite their observable benefits. The relative advantage of EML-enabled systems is not as apparent as conventional operating systems.

Our findings can provide valuable implications for construction companies seeking to gradually increase the adoption of EML-enabled technologies in their projects by focusing on the post-adoption phase. Companies will face more disruptions in project management and supply chains as the construction industry in Vietnam becomes more globalised (Pham et al., 2023). To mitigate risks, companies should develop comprehensive action plans to adopt EML-enabled technologies into their sustainability plans and initiatives. For construction organisations aiming at enlarging EML adoption, a structured implementation approach is paramount. The following roadmap synthesises findings from contemporary sources to propose practical steps for effective integration, as shown in Figure 3.

• Step 1: Enhancing Workforce Capabilities

To prepare the workforce for technological advancements, construction firms must invest in training programmes tailored to develop employee skills in utilising EML technologies. Such investments can enhance workforce productivity, contribute to a decentralised organisational environment and alleviate relationship conflicts among team members (Tariq and Rehman, 2020). Moreover, leveraging a higher rate of workforce engagement through improved training and rewarding mechanisms can significantly increase labor motivation and satisfaction, leading to better performance outcomes (Tam et al., 2022). This enhancement of workforce capabilities addresses findings related to complexity by ensuring that employees are equipped to manage arising challenges effectively, thereby advancing both carbon and business performance. Moreover, progress can be evaluated through metrics such as employee training participation rates, digital skills assessment scores, and reduction in project rework, providing a clear indication of workforce readiness for EML adoption.

• Step 2: Stakeholder Engagement

Encouraging high-power stakeholders' engagement is crucial for transitioning towards sustainability within the construction sector. Constructive stakeholder participation not only stimulates collaborative environments but actively helps organisations identify and leverage sustainability initiatives through machine learning (Shaik et al., 2024). Regular vision-sharing sessions align stakeholders around sustainability goals, enhancing the efficacy of integrated efforts for achieving environmental responsibility (Shaik et al., 2024; Zhong et al., 2025). The positive impact of collective engagement on performance outcomes underscores its importance in sustainable practices. Organisations can monitor stakeholder engagement through indicators such as the number of stakeholders participating in partnership meetings, stakeholder satisfaction ratings, and the number of sustainability collaborations or initiatives launched through EML-based initiatives.

• Step 3: Gradual Technology Adoption

A phased approach to adopting EML-enabled technologies, starting from small-scale trials, allows organisations to mitigate risks while progressively integrating new systems. Research highlights that the trialability of new technologies contributes positively to business performance as firms can adjust based on collected feedback and operational insights (Li et al., 2021). This incremental adaptation makes the transition smoother and less disruptive, ensuring that firms can reduce potential resistance and enhance overall acceptance of new systems (Zhang and Li, 2022). Organisations can assess adoption progress by tracking the number and success rate of pilots, user feedback scores, and the time taken to evolve an EML solution from pilot to full deployment. This ensures that the integration process is continuously monitored and timely adjusted if any issues arise.

• Step 4: Long-term Strategic Planning

Positioning EML integration as part of a long-term strategic vision is essential. Unlike conventional initiatives, the benefits from EML technologies may take time to manifest, particularly in achieving carbon neutrality and improving financial returns (Huynh-Xuan et al., 2024; Luqman et al., 2024). Organisations must not only prepare strategically but also financially for the adaptive processes linked to integrating emerging technologies. They should cultivate an organisational culture that embraces continuous improvement and adaptation, ensuring that the benefits of innovation are recognised over time (Momade et al., 2024). Furthermore, organisations can develop metrics such as investment levels in EML initiatives and short- and long-term economic and carbon targets to ensure EML is effectively integrated into corporate strategies and successfully implemented in phases.

• Step 5: Continuous System Optimisation

Establishing robust project governance is key to maintaining environmental accountability alongside business continuity. Regular assessments and iterative adjustments are necessary to facilitate seamless operations within AI-enabled frameworks and align them with carbon objectives. Firms should also emphasise regulatory compliance and collaborative mechanisms that support continuous enhancement of environmental and operational performance (Saputra et al., 2023).

Implementing EML technologies in construction necessitates a strategic and structured approach focusing on workforce enhancement, stakeholder engagement, gradual technology adoption, long-term planning, and continuous optimisation to achieve desired outcomes in emissions mitigation and business performance.

This study explores how EML adoption affects carbon and organisational performance in the construction sector in Vietnam. Using data from 213 companies in the Mekong Delta region, the study found that only observability and complexity had a positive impact on carbon footprint reduction and business performance. In contrast, relative advantage had no impact on carbon performance and a negative impact on business performance. Moreover, despite empirical evidence demonstrating that EML can improve business outcomes, companies struggle to integrate these technologies into their existing operations to achieve their carbon emission targets. However, environmental initiatives can help maintain performance. Overall, these findings highlight that sustainability is a long-term transformation process that requires companies to integrate technological innovation and adaptive business systems to achieve net-zero operations and business performance resilience.

By focusing on the discourse of EML integration, this study offers clear theoretical and practical takeaways. Theoretically, by integrating multidimensional theories, the study moves the research literature beyond technological focus by explaining how post-adoption attributes of EML-enabled technology shape both carbon and organisational outcomes in the construction industry. By emphasising the post-adoption phase, the study clarifies how interaction between AI-assisted systems and human expertise can support carbon neutrality and performance resilience over time. The study also challenges the assumption that perceived relative advantage automatically improves performance in the context of developing economies; instead, actual value depends more on the ability to effectively observe indirect functional benefits and manage complexity. In practice, companies should not adopt EML solely based on the claim of “better technology,” but should prioritise workforce capabilities, system transparency, and phrased pilot programs to minimise disruption and build adoption. Besides, policymakers can accelerate adoption by providing training, establishing standards and offering incentives to reduce integration costs and barriers. These actions help align digital investment with measurable emissions reductions while maintaining project performance and delivery quality.

This study nonetheless has some limitations despite its contributions to theoretical literature and practical implications. As previously mentioned, the complexity of adopting EML-enabled technologies also occurs during the process, and its impact on sustainable performance will ultimately occur over the long term. As this study used a cross-sectional research design, it limited the view on firm transition, progress, and achievement. Firms may make incremental adaptations as they progress, while the net-zero transition is reliant on continuous optimisation of AI-driven decision-making processes. Hence, longitudinal studies are needed to track the influence of EML on construction projects in future research. This will allow for greater detail on how AI-enabled technologies address carbon emissions and how organisations adapt to emerging structures with emerging AI applications. In addition, this study examined only the adoption of EML and organisational-level sustainability outcomes. Meanwhile, sustainability transitions entail a broader range of policy frameworks, supply chain ecosystems, and global initiatives. Therefore, future research could build upon the literature on AI-enabled sustainability at a macro level by studying the effect of EML on supply chain decarbonisation, circular economy transition, and regulatory frameworks on carbon neutrality.