Evaluating the efficacy of risk management practices for generative AI integration in sustainable construction projects: a fuzzy set theory approach

Abstract

Purpose:

This study assesses the significance of risks associated with integrating Generative Artificial Intelligence (GenAI) into risk management for sustainable construction projects (SCPs). Given the early stage of GenAI adoption in the construction sector, the study adopts an exploratory approach to evaluate how industry professionals perceive the efficacy and preparedness of existing risk management (RM) practices in addressing emerging GenAI-related risks. It also proposes response strategies to mitigate their potential impacts.

Methodology:

The research followed a five-stage methodology. First, a systematic literature review (SLR) identified and classified GenAI-related risks. Second, a multi-criteria assessment model was developed to evaluate risk significance and the perceived efficacy of RM practices in addressing these emerging risks. Third, a structured survey involving 80 construction experts provided input data for the developed assessment model. Fourth, a fuzzy-based model was developed to quantify the perceived level of RM practice efficacy in relation to GenAI-related risks. Finally, a semi-structured expert survey identified best-practice response strategies for each risk category.

Findings:

The study identified 30 risk factors grouped into five categories: input quality, technological adaptability, ethical and governance, information integrity, and financial risks. The results indicate a generally low-to-medium level of perceived preparedness of current RM practices in addressing GenAI-related risks. This finding suggests that existing RM structures may not yet be fully equipped to manage the complexities introduced by GenAI, particularly within the fragmented, multi-stakeholder, and sustainability-driven context of SCPs. Best-practice strategies were also identified for each risk category.

Implications:

The study presents an integrated RM assessment model positioned as an early-warning tool for evaluating organisational preparedness for GenAI integration in SCPs. It identifies gaps in perceived RM efficacy, highlighting the need for targeted mitigation strategies. The proposed response strategies offer practical guidance for improving resilience and readiness in SCPs while considering sector-specific challenges such as temporary project organisations, regulatory demands, and lifecycle sustainability requirements.

Originality/Value:

This study is among the first to assess GenAI-related risks in RM for SCPs using fuzzy logic. It adopts an anticipatory and perception-based approach suited to early-stage adoption rather than evaluating mature implementation. The findings highlight gaps in perceived RM preparedness and provide sector-specific insights that support more effective and responsible GenAI integration in sustainable construction.

1 Introduction

Sustainable Construction Projects (SCPs) have become a central component of modern construction practices due to their role in supporting long-term environmental, social, and economic objectives (Dobrovolskiene and Tamosiuniene, 2015). These projects aim to minimise negative impacts while maximising lifecycle value through practices such as reducing waste and carbon emissions (Hom, 2024), improving energy efficiency (Dobrovolskiene and Tamosiuniene, 2015; Iluyomade and Okwandu, 2024), and adopting durable, resource-efficient materials that reduce the need for frequent replacement (Gan et al., 2015). In addition, SCPs incorporate sustainable design principles that consider ecological systems and promote responsible resource use (Behrooz et al., 2023). Common strategies include the use of environmentally friendly materials, water conservation measures, and design approaches that enhance occupant wellbeing while preserving biodiversity (Farida et al., 2019; Liudmila and Balkiz, 2019). Through prioritising long-term efficiency and environmental stewardship, SCPs play a critical role in advancing sustainability within the construction sector (Datta et al., 2023).

Despite these advantages, SCPs are associated with increased levels of risk and uncertainty. The need to simultaneously address environmental, social, and economic objectives expands project scope and introduces complex interdependencies among project variables (Dobrovolskiene and Tamosiuniene, 2015; Hauke et al., 2016; Hossny et al., 2021). These projects often involve a broader range of stakeholders, including regulatory authorities and environmental organisations, which complicates coordination and alignment of interests (Bal et al., 2013). Furthermore, the adoption of innovative technologies and sustainable materials introduces uncertainties related to long-term performance and reliability (Iluyomade and Okwandu, 2024). Additional challenges arise from stringent environmental regulations, certification requirements, and the difficulty of predicting long-term sustainability outcomes (Farida et al., 2019; Nishant et al., 2020). These factors collectively increase the complexity of risk management in SCPs.

The distinctive characteristics of SCPs further intensify these challenges. Project delivery is often fragmented across multiple temporary organisations, requiring coordination among diverse stakeholders with differing priorities and responsibilities. In addition, sustainability objectives introduce strict regulatory requirements, lifecycle performance expectations, and the need for transparent and traceable data. These conditions increase the consequences of inaccurate risk assessments, as decisions may affect not only cost and schedule but also environmental compliance, certification outcomes, and long-term asset performance. As a result, risk management in SCPs requires approaches that can address both technical complexity and multi-dimensional accountability.

The growing complexity of SCPs highlights the need for advanced tools capable of enhancing risk management efficacy. Generative Artificial Intelligence (GenAI) has emerged as a promising solution due to its ability to analyse large datasets, identify patterns, and support data-driven decision-making (Yazdi et al., 2024; Hadid et al., 2024). Conventional risk management approaches are often constrained by limited data processing capacity and an inability to capture hidden relationships within complex project environments (Al-Mhdawi et al., 2024a). In contrast, GenAI enables real-time data analysis and predictive capabilities that support proactive risk identification and response (Raj et al., 2025; Mohamed et al., 2025a). Its application within construction management is expanding as organisations seek to improve project performance, enhance responsiveness, and manage uncertainty more effectively (Pan and Zhang, 2021; Afzal et al., 2020).

However, the integration of GenAI into risk management introduces a new set of risks that must be carefully examined. Input quality risks arise when models rely on inaccurate, incomplete, or biased data, which may lead to unreliable outputs and flawed decision-making (Afzal et al., 2020). Technological adaptability risks are associated with difficulties in integrating GenAI systems into existing organisational infrastructures (Mohamed et al., 2025b; Rane, 2023). Ethical and governance risks, including lack of transparency, weak accountability, and data privacy exposure, add further complexity (Regona et al., 2024). Information integrity risks also emerge due to the potential for misleading or inconsistent outputs generated by AI systems (Rane, 2023). In addition, financial risks related to implementation and operational costs may affect the feasibility of adoption (Pillai and Matus, 2020). These risks are particularly critical in SCPs, where regulatory requirements, data sensitivity, and sustainability objectives increase the consequences of inadequate risk control (Mohamed et al., 2025b; Tettey et al., 2025).

Despite the growing interest in GenAI, its application in construction risk management remains at an early stage and is not yet embedded in routine practice (Mohamed et al., 2026a). This situation raises an important question regarding how existing RM practices may perform when exposed to GenAI-related risks that are not yet fully realised in practice. Direct evaluation of implemented performance is therefore limited. However, exploratory assessment based on informed professional judgement remains valuable, as it provides early indications of potential gaps in preparedness before large-scale adoption occurs.

Although existing studies highlight the benefits of GenAI in construction, limited research has examined the risks associated with its integration into risk management for SCPs. Prior work has primarily focused on advantages such as improved predictive accuracy, automation of processes, enhanced stakeholder communication, and early risk detection (Pan and Zhang, 2021; Salih and El-Adaway, 2024). While these contributions demonstrate the potential of GenAI, they do not provide a comprehensive assessment of the risks introduced by its adoption. Furthermore, the literature lacks studies that systematically identify and categorise GenAI-related risks and explore whether existing RM practices are adequately prepared to address these risks within SCPs. This gap is particularly significant given the increasing reliance on complex data and the need for rigorous risk governance in sustainable construction. Given these conditions, this study is positioned as an exploratory and early-warning assessment rather than an evaluation of mature implementation. It examines how construction professionals perceive the efficacy of existing RM practices in addressing anticipated GenAI-related risks under conditions of early-stage adoption. This perspective is particularly relevant for SCPs, where the consequences of inadequate risk governance are amplified by sustainability requirements, fragmented project delivery structures, and lifecycle accountability.

In response to this gap, this study aims to examine the risks associated with integrating GenAI into risk management within sustainable construction projects. The specific objectives are: (1) to identify and categorise the key risks associated with GenAI implementation in risk management for SCPs; (2) to examine the perceived efficacy of current RM practices in addressing these anticipated risks; and (3) to propose optimal strategies to mitigate the impact of GenAI-related risks.

2 Literature review

2.1 Sustainable construction projects (SCPs)

SCPs are increasingly recognised as a central pathway for delivering built assets that align with long-term environmental, social, and economic priorities (Fischer et al., 2016; Gan et al., 2015). SCPs emphasise strategies that minimise adverse impacts across the building lifecycle by reducing waste and carbon emissions, improving energy efficiency, and using durable, resource-efficient materials that limit the need for frequent replacement (Bhatti and Nazir, 2024; Wijayasekera et al., 2022). These projects also incorporate sustainable design principles that consider ecological sensitivity, site characteristics, and the responsible management of natural resources (Liudmila and Balkiz, 2019; Fischer et al., 2016). Key features include the use of environmentally friendly materials, water conservation measures, and design approaches that enhance occupant wellbeing while protecting biodiversity (Hauke et al., 2016; Bal et al., 2013). SCPs prioritise long-term resource efficiency and environmental stewardship to ensure that construction activities remain viable over time, strengthening their role in advancing sustainability within the construction sector (Sfakianaki, 2015; Toochukwu, 2025).

The characteristics of SCPs extend beyond environmental considerations to encompass social and economic dimensions (Edum-Fotwe and Price, 2009; Zabihi and Habib, 2012). Socially, SCPs aim to support occupant health, community wellbeing, and user comfort (Arif et al., 2016; McMillan and Varga, 2022). Economically, they seek to optimise operational efficiency, reduce lifecycle costs, and deliver long-term asset value (Altaf et al., 2023). Toochukwu (2025) argued that the integration of advanced design practices, innovative materials, and high-performance building systems distinguishes SCPs from conventional projects, which often focus on short-term cost objectives rather than long-term resilience and sustainability performance.

SCPs also encompass a range of project types that reflect different sustainability priorities (Sfakianaki, 2015). These include energy-efficient buildings designed to minimise operational energy demand, low-carbon projects that reduce embodied emissions through material choices, water-efficient developments that optimise consumption and recycling, and nature-sensitive projects that preserve biodiversity and ecological balance (Chu et al., 2022; Abdous et al., 2024; Hauke et al., 2016). Additional SCP types involve passive design strategies, adaptive reuse of existing structures, and the integration of renewable energy systems (Okiye et al., 2023). Although these categories vary in emphasis, they share core objectives related to improving environmental performance, enhancing long-term value, and supporting broader sustainability goals (Sfakianaki, 2015; Toochukwu, 2025).

The holistic nature of SCPs introduces significant managerial and technical complexity (Sabini and Silvius, 2023). Project teams must balance environmental performance requirements with social and economic objectives while engaging with a wider range of stakeholders, stricter regulatory expectations, and higher levels of transparency (Bal et al., 2013). Compliance with sustainability standards and certification systems also increases administrative and organisational demands (Mori Junior et al., 2016). In addition, SCPs are typically delivered through fragmented and temporary project organisations, where multiple stakeholders interact across different lifecycle stages. This structure increases the need for consistent, reliable, and traceable information to support decision-making. Any deficiencies in data quality, communication, or governance can therefore have amplified consequences, affecting not only project performance but also sustainability compliance, certification outcomes, and long-term asset value. These characteristics make SCPs particularly sensitive to emerging digital risks, including those associated with GenAI integration. As a result, delivering SCPs effectively requires rigorous planning, continuous monitoring, and advanced risk management approaches capable of addressing the heightened uncertainty and interdependence that characterise these projects.

2.2 Generative artificial intelligence (GenAI)

2.2.1 Definition and key concept

The development of GenAI is rooted in the evolution of machine learning and neural network research over several decades (Anysz et al., 2021; Chou et al., 2024). Early AI systems focused on rule-based reasoning and simple pattern recognition, but the introduction of artificial neural networks marked a turning point by enabling models to learn directly from data (Abiodun et al., 2019). The emergence of deep learning further accelerated progress, allowing computers to identify complex patterns within large datasets (Hadid et al., 2024; Abiodun et al., 2019; Erfani and Cui, 2022). The first major step towards generative capability came with the advancement of autoencoders and sequence-based models, which enabled machines to reconstruct and generate new data (Pillai and Matus, 2020). Over time, innovations such as adversarial learning, transformer-based architectures, and diffusion-based generative techniques transformed GenAI into a sophisticated family of models capable of producing highly realistic and context-aware outputs (Yazdi et al., 2024; Costantino et al., 2015). These developments positioned GenAI as a powerful tool capable of addressing complex decision-making and analytical challenges across various domains, including engineering and construction (Hadid et al., 2024; Parveen, 2018).

GenAI refers to advanced AI models designed to produce new content, patterns, or insights derived from learned data (Nidhisree et al., 2024; Mohamed et al., 2026b). Unlike conventional AI systems that primarily classify or predict outcomes, GenAI generates original outputs that resemble or extend existing information (Du et al., 2024; Wu et al., 2017). These outputs may include text, visual content, predictive scenarios, optimisation alternatives, or structured analytical insights that support decision-making in dynamic project environments (Rane et al., 2023; Davahli et al., 2021). GenAI operates through a range of deep learning architectures. Among the earliest generative models were autoencoders, which compress and reconstruct data, enabling the creation of new variations. This foundation led to Variational Autoencoders (VAEs), which learn underlying data distributions and generate new samples, making them useful for producing design alternatives or construction scenarios (Ren, 2022). A major breakthrough occurred with the introduction of Generative Adversarial Networks (GANs), where two neural networks compete to create increasingly realistic synthetic data. GANs support applications such as visual risk scenario modelling, site simulation, and synthetic dataset generation for training safety or defect detection systems (Dunmore et al., 2023).

Another transformative development was the emergence of transformer-based language models, which enabled high-performance natural language generation, document understanding, and analytical reasoning across large volumes of textual project information. These models support tasks such as contract interpretation, automated reporting, risk statement generation, and unstructured data analysis. Diffusion models later expanded generative capabilities to high-resolution images and complex spatial representations, contributing to architectural visualisation and construction scenario synthesis (Hagos et al., 2024; Mohamadi et al., 2023). GenAI also includes reinforcement learning-based generative models, which learn optimal behaviours through iterative interactions within simulated environments (Cao et al., 2023). These models generate adaptive strategies and decision sequences that respond to evolving project risks or operational constraints. Recent advancements have produced hybrid generative systems that integrate deep learning with optimisation, simulation, and probabilistic reasoning, enabling GenAI to address multidimensional engineering problems involving uncertainty, interdependencies, and incomplete data (Dunmore et al., 2023; Kasgari et al., 2020).

A defining characteristic of GenAI is its adaptability (Li et al., 2024). Through continual exposure to new information or feedback, these models refine their generative accuracy, improve contextual awareness, and respond dynamically to evolving project conditions. This capacity to synthesise new insights from large-scale, diverse, and unstructured datasets distinguishes GenAI from traditional analytical tools. As a result, GenAI has become an increasingly valuable asset for engineering and construction applications, especially in areas that require predictive modelling, scenario analysis, knowledge extraction, and real-time decision support (Caldas et al., 2020; Gheibi et al., 2021).

2.2.2 GenAI for risk management in construction

Risk management in construction is complicated by the dynamic, fragmented, and data-intensive nature of project environments (Yazdi et al., 2024). Construction projects generate large volumes of information across design, procurement, site operations, safety inspection, and supply chain activities (Pan and Zhang, 2021). Much of this information is unstructured and updated frequently, making it difficult for traditional methods to capture emerging risks, identify hidden patterns, or respond to rapid changes on-site (Afzal et al., 2020). These limitations highlight the need for analytical tools that can process diverse data sources and support more proactive, data-driven decision-making (Waqar et al., 2023; Yigitcanlar et al., 2022).

GenAI aligns strongly with these needs because it can analyse complex datasets, recognise subtle patterns, and generate new insights that inform risk-related decisions (Zhao et al., 2024; Mohamed et al., 2025b). Its ability to synthesise diverse information types such as text, images, schedules, and numerical data enables it to translate large and unstructured construction datasets into meaningful risk intelligence (Regona et al., 2024). This capacity makes GenAI particularly suited to environments where information evolves continuously and risks emerge across multiple project stages (Aladağ, 2023; Lachhab et al., 2018).

One area where GenAI adds distinctive value is in risk identification. Construction risks are often embedded in narrative documents such as site diaries, method statements, design comments, and inspection reports (Eber, 2020). Manual review of these documents is time-consuming and prone to oversight. GenAI can automatically extract relevant information, detect recurring concerns, and generate structured risk statements. This supports more comprehensive identification of potential risks, including those that may not be immediately visible to project teams (Mohamed et al., 2025a; Regona et al., 2024; Fridgeirsson et al., 2023).

In addition to improving identification, GenAI strengthens risk prediction through the generation of scenarios that reflect possible project outcomes (Pham and Han, 2023). Construction projects are influenced by variables such as weather conditions, material availability, labour fluctuations, and design revisions. GenAI can simulate alternative scenarios based on these variables, allowing project teams to anticipate delays, safety risks, cost deviations, and quality-related risks before they materialise (Yaseen et al., 2020). This capability supports a shift from reactive problem-solving to more proactive planning.

GenAI also enhances the monitoring and communication of risks as projects progress (Zhou et al., 2023). Construction environments change rapidly, requiring continuous updates to risk registers and mitigation plans (Al-Mhdawi et al., 2024b; Al-Mhdawi et al., 2024c). GenAI can process new information in real time and update risk assessments accordingly, enabling project teams to adjust their strategies as conditions evolve (Zhou et al., 2023; Yazdi et al., 2024). At the communication level, GenAI can generate summaries, visual reports, and explanations tailored to different stakeholders, strengthening shared understanding and coordination across project teams (Raj et al., 2025; Pan and Zhang, 2021).

These capabilities are particularly relevant in SCPs, where risk management depends on integrating heterogeneous data related to environmental performance, regulatory compliance, lifecycle assessment, and stakeholder reporting. Decision-making in such contexts requires high levels of data accuracy, traceability, and transparency. Any limitations in the quality or interpretation of GenAI-generated outputs may therefore have amplified consequences, affecting not only project performance but also sustainability certification outcomes, environmental compliance, and long-term asset value.

These operational improvements extend to strategic decision-making (Yaseen et al., 2020). GenAI can generate alternative design options, compare mitigation strategies, and support evaluations that balance cost, safety, time, and sustainability considerations (Hadid et al., 2024). This enhances the ability of organisations to make informed decisions that align with long-term project objectives and sustainability goals (De Almeida, 2024; Paramesha et al., 2024). Although GenAI offers substantial benefits, its integration into construction risk management introduces new risks. These include risks related to data quality, model transparency, organisational readiness, and the reliability of automated outputs in high-stakes decisions. Construction datasets can be inconsistent or incomplete, and GenAI’s reliance on such data may lead to biased or misleading risk insights (Mohamed et al., 2025a).

These risks are particularly critical in SCPs, where project outcomes are subject to heightened regulatory scrutiny and long-term sustainability expectations. In this context, even minor inaccuracies in data or model outputs can influence compliance decisions, stakeholder trust, and lifecycle performance. As a result, assessing how existing RM practices are perceived to cope with such emerging risks provides an important exploratory step in understanding organisational preparedness prior to widespread GenAI adoption in the construction sector.

2.3 Related work

As summarised in Table 1, prior studies addressing GenAI in CRM and sustainable construction projects (SCPs) can broadly be grouped into four themes. The first theme highlights how GenAI can enhance data processing, automate documentation, and support early risk detection. For instance, Pan and Zhang (2021), Afzal et al. (2020), Afzal et al. (2021), and Salih and El-Adaway (2024) demonstrate that AI-driven models can process large, unstructured construction datasets, generate risk scenarios, and improve predictive accuracy in project planning and control. These studies collectively underline the potential of GenAI to transform traditional risk management through real-time analytics and improved decision support.

The second theme focuses on the challenges and risks associated with generative AI adoption and digital transformation in construction. Studies by Pan and Zhang (2021), Regona et al. (2024), Alt et al. (2024); Choi et al. (2021), and Rane (2023) identify risks related to data quality, algorithmic bias, organisational resistance, technological incompatibility, and cyber-security threats. A third theme examines ethical and governance considerations related to AI in engineering and infrastructure. Liang et al. (2024) and Heersmink et al. (2011), for example, propose frameworks addressing accountability, transparency, data privacy, and fairness in AI deployment. The fourth theme focuses on the efficacy of risk management practices and the application of fuzzy logic in construction risk assessment. Agyekum-Mensah and Knight (2017), Osei-Kyei and Chan (2018), and Al-Mhdawi et al. (2023b) evaluate how stakeholder engagement, communication quality, risk culture, and organisational support influence project performance.

Although these studies provide valuable insights, they tend to address GenAI-related risks and RM practices in a generalised manner, without fully considering the distinctive characteristics of SCPs. In particular, limited attention has been given to how fragmented project delivery structures, multi-stakeholder coordination, regulatory pressures, and lifecycle sustainability requirements shape the nature and consequences of GenAI-related risks in this context. As a result, the implications of GenAI adoption in SCPs remain insufficiently explored.

Prior work therefore falls short in several key areas. First, existing studies do not systematically identify and classify the full spectrum of GenAI-related risks that arise when GenAI is embedded in RM processes for SCPs. Many contributions treat AI-related risks at a general organisational level or focus primarily on performance benefits, without capturing the interdependencies between technical, organisational, and sustainability-related risk dimensions. Second, most contributions focus on conventional construction or broader digital transformation, paying limited attention to the stricter regulatory, environmental, and data-sensitivity demands of SCPs. This limitation reduces the applicability of existing findings to contexts where decision-making must satisfy not only cost and time objectives but also sustainability compliance, certification standards, and long-term lifecycle performance. Third, existing studies provide limited examination of whether current RM practices are adequately prepared to address anticipated GenAI-related risks under conditions of early-stage adoption. Instead, prior work often assumes fully developed implementation or emphasises technological capabilities, with comparatively less attention given to organisational preparedness and perception-based evaluation. Finally, there is a lack of research that integrates a structured GenAI risk taxonomy with an exploratory, perception-based assessment of RM practice efficacy tailored to SCPs. Such an approach is particularly important in this sector, where temporary project organisations, fragmented information flows, and sustainability-driven accountability may significantly influence how risks are perceived, prioritised, and managed. In response to these gaps, this study adopts an exploratory, early-warning perspective to examine GenAI-related risks in risk management for SCPs, drawing on insights from United Kingdom construction professionals.

3 Methodology

A fixed-stage mixed-method approach was adopted in this research for data collection, analysis, and processing, as illustrated in Figure 1. The details of each methodological stage are presented in the following subsections.

FIGURE 1

3.1 Stage one: identification of the key risks emerging from GenAI integration into RM of SCPs

This stage adopted an SLR aligned with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) framework to identify and classify the key risks associated with integrating GenAI into RM within SCPs. A PRISMA-guided approach was adopted because it offers a transparent, replicable, and logically sequenced procedure for identifying, screening, assessing, and including relevant studies, thereby reducing the likelihood of informal or inconsistent study selection (Almashhour et al., 2025; Elseknidy et al., 2025). Consistent with recent literature-based studies in construction and risk-related research, this stage was designed to establish a credible evidence base for the later empirical phases of the study and to ensure that the resulting risk classification was grounded in a systematically refined body of knowledge (Khalid et al., 2024; Mohamed et al., 2026a; Suffo and Brome, 2018; Gupta et al., 2023).

Given the early stage of GenAI adoption in construction, this stage focused on identifying both observed and anticipated risks associated with GenAI integration into RM processes. This approach supports the exploratory nature of the study by capturing emerging risk themes that may not yet be fully reflected in routine industry practice, but are conceptually and empirically grounded in the existing literature. The review progressed through four PRISMA steps: identification, screening, eligibility, and inclusion, as illustrated in Figure 2. However, the stage was intended not only to collect literature, but also to refine the evidence into a form suitable for structured risk identification. Accordingly, the selected studies were examined to identify recurrent GenAI-related risks, merge conceptually overlapping items, and organise the refined risk factors into coherent categories.

FIGURE 2

Particular attention was given to risks relevant to construction and SCP contexts, including those linked to data reliability, regulatory compliance, lifecycle performance, and multi-stakeholder coordination. The final output of this stage was a taxonomy comprising 30 risk factors grouped into five main categories: input quality-related risks, technological adaptability-related risks, ethical and governance-related risks, information integrity-related risks, and financial risks.

3.1.1 Identification step

A comprehensive search strategy was developed to guide the identification of relevant literature. Scopus was selected as the primary database because of its broad coverage of peer-reviewed publications across construction management, engineering, and technology-related disciplines, making it particularly suitable for an interdisciplinary topic such as GenAI-enabled RM in SCPs (Farfan Chilicaus et al., 2025; Gao et al., 2023). The review focused on literature published between 2014 and 2025 in order to capture the development of AI-driven and, more recently, GenAI-related literature relevant to RM in construction contexts. To improve consistency and quality in source selection, the review prioritised English-language studies published in peer-reviewed and academically credible outlets. Journal quality was also considered where appropriate, with impact factor used as an additional indicator of publication quality rather than as a sole basis for inclusion (Naoum et al., 2015; Siraj and Fayek, 2019; Vaishya et al., 2020). The search was performed within the title, abstract, and keyword fields, and the search terms were organised into three thematic groups: GenAI-related terms, RM-related terms, and sustainable construction/project-context terms. These keyword groups are presented in Table 2, while the complete Scopus search string used in stage one is provided in Table 3. Boolean operators such as AND OR were used to combine the search terms in a way that broadened retrieval while maintaining conceptual relevance to the study objectives. This process generated an initial pool of 271 records.

3.1.2 Screening step

Following identification, the retrieved records were screened in a staged and increasingly focused manner. The first screening pass removed duplicate and clearly irrelevant studies, reducing the sample from 271 to 138 records. This step ensured that broadly captured search outputs that lacked sufficient relevance to GenAI, RM, or sustainable construction were filtered out before more detailed assessment. The second screening pass involved reviewing the titles of the remaining studies. At this stage, records that did not demonstrate an adequate thematic connection to the integration of GenAI into RM or to closely related construction and project-management contexts were excluded, reducing the sample from 138 to 112 studies. The third screening pass involved abstract review, which provided a more refined assessment of each study’s relevance, scope, and likely contribution. This stage reduced the sample further from 112 to 98 studies, which were then advanced to the eligibility stage. Together, these screening steps strengthened the conceptual focus of the review while preserving transparency in the refinement process.

3.1.3 Eligibility step

A full-text and content-based eligibility assessment was then conducted for the 98 retained studies using the inclusion and exclusion criteria presented in Table 4. At this stage, each article was examined in greater depth to determine whether it made a meaningful contribution to the identification of GenAI-related risk categories and specific risk factors within the RM of SCPs. The assessment considered conceptual relevance, methodological contribution, and the extent to which each study discussed, classified, or evidenced risks, limitations, barriers, or implementation challenges associated with GenAI or closely related AI applications in risk-oriented construction settings. In line with the exploratory positioning of the study, both empirical evidence and forward-looking discussions of potential risks were considered relevant, provided they were grounded in credible analytical or conceptual arguments. After this full-text review, 36 studies were excluded. The excluded studies generally lacked sufficient relevance to the core research focus, did not address RM or sustainable construction in a meaningful way, or concentrated only on general AI benefits without discussing associated risks or implementation challenges. Following this step, 62 studies were retained for the final analytical phase.

3.1.4 Inclusion and content analysis

Following the eligibility assessment, a formal inclusion step was undertaken to ensure that only studies directly aligned with the research objectives were retained. At this stage, each study was re-examined to confirm that it explicitly addressed risks, limitations, or challenges associated with GenAI integration in RM or closely related construction contexts, and that it provided sufficient detail to support systematic extraction of risk-related content. Studies lacking extractable risk information or presenting only general discussions of AI benefits were excluded to ensure analytical consistency and traceability. A conventional content analysis approach was subsequently applied because it is particularly well suited to emerging research areas in which concepts remain dispersed and categories need to be derived inductively from the literature rather than imposed in advance (Elo and Kyngas, 2008; Hsieh and Shannon, 2005). Within this process, risk-related statements, concepts, descriptions, and discussion points were extracted from each study and recorded as initial risk items. To improve transparency and reproducibility, the extraction process followed a structured coding procedure. During the initial coding phase, each study was reviewed systematically, and all references to GenAI-related risks were identified and assigned preliminary codes that closely reflected the terminology used in the source material. For example, statements referring to inaccurate datasets were coded as “inaccurate or incomplete data” (R01), while references to limited data access were coded as “data unavailability” (R02), and discussions of biased datasets were coded as “data bias” (R03). This process generated a comprehensive pool of initial codes representing diverse expressions of risk across the literature.

The extracted items were then compared across the selected studies to identify conceptual overlap, repeated patterns, and variations in terminology. Where multiple studies referred to closely related risks using different wording, the items were merged and refined into common risk factors. This consolidation stage involved grouping semantically similar codes into unified risk factors. For instance, codes such as limited data access and missing datasets were merged into “data unavailability” (R02), while codes such as incorrect interpretation of AI outputs were grouped under “misinterpretation of results” (R10), and references to lack of technical skills were consolidated into “absence of expertise” (R09). To ensure robustness, each refined factor was retained only when supported by multiple studies or justified through strong conceptual relevance within the literature.

These refined factors were then grouped into broader thematic categories according to their dominant characteristics and conceptual similarity. The categorisation process was guided by both inductive patterns emerging from the data and alignment with construction risk management literature. For example, factors such as “inaccurate or incomplete data” (R01), “data unavailability” (R02), and “data bias” (R03) were grouped under input quality-related risks, as they directly affect the reliability of data inputs used by GenAI systems. Similarly, factors such as “insufficient training” (R07), “absence of expertise” (R09), and “human error” (R11) were grouped under technological adaptability-related risks, while factors including “algorithm bias” (R17) and “unclear responsibility” (R20) were classified under ethical and governance-related risks. Factors such as “data breach” (R23) and “vulnerability to adversarial cyberattacks” (R26) were grouped under information integrity-related risks, and cost-related factors such as “high initial costs” (R27) were categorised under financial risks. This structured grouping ensured conceptual coherence within categories while maintaining clear differentiation between them.

The categorisation process also considered the relevance of each risk factor to the specific conditions of SCPs, including data sensitivity, regulatory requirements, and lifecycle accountability, to ensure that the resulting taxonomy reflects sector-specific risk characteristics. To enhance reliability, the coding and categorisation process was conducted iteratively, with repeated reviews to refine factor definitions, remove redundancy, and ensure internal consistency. Particular attention was given to distinguishing closely related factors, such as separating “data bias” (R03) from “algorithm bias” (R17), to maintain conceptual clarity across categories.

To this end, this iterative process of extraction, comparison, consolidation, and categorisation produced the final taxonomy of 30 risk factors.

3.2 Stage two: development of a multi-criteria assessment model for evaluating GenAI-related risks and RM practices

In this stage, the authors built upon prior research in the field of RM to identify key indicators for evaluating the perceived efficacy of RM practices, including studies such as, Agyekum-Mensah and Knight (2017) and Osei-Kyei and Chan (2018). The study of Al-Mhdawi et al. (2023b) was particularly instrumental, as it derived a comprehensive set of indicators through expert interviews with senior professionals in the construction industry, each possessing at least 15 years of experience and holding active memberships in reputable international bodies such as the Association for Project Management (APM), the American Society of Civil Engineers (ASCE), and the Institution of Civil Engineers (ICE).

From this process, six core indicators were identified, each accompanied by specific assessment factors: (1) Stakeholder Involvement Level, measured by the relative significance of each stakeholder to the other (RS) and their level of interest in the project (LI); (2) Organisational Communication Level, evaluated through organisational communication culture (OCC) and the usability of communication tools (UCT); (3) Risk Culture, assessed via the willingness to collaborate and share knowledge (WCKS) and the presence of conflicting values (CV); (4) Risk Management Training Level, based on the availability of expertise for training (AET) and availability of resources (AR); (5) Risk Management Policies, examined through upper management support and openness (UMSO) and the perceived potential benefits of RM practices (PB); and (6) Continuous Risk Monitoring Level, assessed via the level of RM integration with other management processes (RMIMP) and project parties’ integration (PPI).

In addition, a Risk Significance Level (RSL) indicator was incorporated into the model to assess the significance of each risk based on four factors: Impact Level (IL), Probability Level (PL), Detectability Level (DL), and Exposure Level (EL). This indicator ensures that the assessment of RM practices reflects the relative importance of each risk from a perception-based perspective rather than assuming uniform impact across risk categories. Given the early stage of GenAI adoption in construction, the model was designed to assess how RM practices are perceived to cope with anticipated GenAI-related risks rather than to evaluate performance under fully implemented conditions. The model therefore serves as an exploratory tool for examining organisational preparedness and identifying potential gaps in existing RM structures.

The SCP context is important not only for framing the study but also for the logic of the assessment model itself. In this research, the significance of each identified GenAI-related risk was evaluated before being incorporated as an input into the RM practice efficacy model. This means that the contextual setting directly affects the assessment outcome. A risk that appears moderate in a general construction or oil and gas setting may carry greater significance in SCPs due to fragmented project delivery, temporary collaboration among multiple parties, design–construction coordination demands, site-based uncertainty, contractual complexity, and sustainability-related performance requirements. Accordingly, the evaluation framework was applied specifically within the SCP context to ensure that both the risk significance assessment and the RM practice efficacy results reflect sector-specific conditions.

To strengthen the development of the indicator set, the authors adopted a structured Delphi process involving six experts in construction management and risk management after the initial indicators had been identified from prior literature. This method allows the systematic collection of expert judgements through iterative consultation, facilitating the refinement of research constructs and the development of consensus among specialists on complex or emerging topics (Ojiako et al., 2025).

The purpose of this process was to enhance the conceptual clarity and contextual suitability of the indicators for evaluating perceived RM preparedness in relation to GenAI-related risks rather than to validate established operational performance. Two iterative rounds of the Delphi process were conducted to achieve expert consensus. In the first round, the experts reviewed the proposed indicators and their associated factors in terms of definitional clarity, potential overlap between constructs, relevance to the study objectives, and applicability to the sustainable construction context. Based on the feedback received, revisions were made to improve wording, remove ambiguity, and ensure alignment between the indicators, their associated factors, and the structure of the survey instrument. In the second round, the revised indicators were presented again to the same experts for further evaluation and confirmation. This round focused on validating the clarity, coherence, and contextual relevance of the refined indicators and ensuring agreement on their suitability for inclusion in the study. This iterative process resulted in the retention of six core RM practice indicators and one RSL indicator, which were subsequently carried forward to questionnaire design and pilot testing.

3.3 Stage three: rating efficacy of RM practices

In this stage, a structured questionnaire survey was employed to measure the perceived efficacy of existing RM practices in addressing the risks identified in Stage One. This approach allowed data to be gathered in a consistent and comparable manner, enabling a systematic analysis of practitioners’ perceptions while ensuring respondent anonymity, which often enhances the reliability and candour of responses (Bird, 2019; Chartres et al., 2019).

3.3.1 Survey development

The questionnaire was specifically designed to assess the perceived efficacy of RM practices in relation to the six main evaluation indicators identified in Stage Two: stakeholder involvement, organisational communication, risk culture, RM training, RM policies, and continuous risk monitoring. Each indicator was assessed through two associated factors, reflecting the underlying elements that contribute to RM practice performance.

In addition to the six indicators, a RSL indicator was introduced to ensure that the assessment of RM practices in addressing each risk reflects its relative importance. The structure of the questionnaire, as shown in Supplementary Appendix SA, was directly informed by the indicator development stage. Each survey item was mapped to one of the indicators or associated factors identified in Stage Two, ensuring alignment between the conceptual model and the measurement instrument. This step was intended to support content coherence between the literature-derived constructs, the Delphi-based refinement process, and the final questionnaire administered to respondents.

Participants were asked to evaluate the efficacy of RM practices in addressing risks associated with integrating GenAI within SCPs using a five-point Likert scale. The linguistic ratings Very Low, Low, Medium, High, and Very High were later mapped into the fuzzy inference system as input variables. The survey consisted of two key sections: the first collected background information such as respondents’ professional roles, years of experience, academic qualifications, and their familiarity with GenAI, while the second section focused on rating the perceived efficacy of current RM practices in relation to the identified GenAI-related risks.

3.3.2 Pilot testing

Before large-scale distribution, a pilot test was carried out to verify the clarity, structure, and interpretability of the survey instrument. Pilot testing serves as a vital step in evaluating the reliability and validity of survey items, ensuring that questions are properly worded, relevant, and capable of capturing the intended responses (Saunders et al., 2019; Bryman, 2016). In this study, 15 UK-based construction professionals with experience in project management and GenAI applications participated in the pilot survey, as shown in Figure 3. Feedback from the participants led to minor revisions, including simplifying technical terminology, refining the indicator descriptions, and adding a question regarding respondents’ direct experience with GenAI. The improved version demonstrated enhanced readability and alignment with the study objectives.

FIGURE 3

3.3.3 Survey administration

The finalised questionnaire was distributed to 104 construction management professionals across the United Kingdom. Selection criteria included: (1) active engagement in the construction industry or academic research related to construction management, and (2) familiarity with or practical experience using GenAI in project environments. Respondents represented a diverse range of roles including project managers, consultants, and academics. Out of the 104 distributed questionnaires, 89 were returned, with 80 valid responses retained after excluding incomplete submissions, yielding a 76.9% valid response rate. A detailed overview of the respondents’ demographics and professional characteristics is presented in the results and analysis section. Given the early stage of GenAI adoption in construction, the survey was designed to capture informed expert judgement regarding the preparedness of existing RM practices rather than to evaluate performance under fully implemented conditions. Respondents were therefore asked to assess RM practices based on their professional knowledge, experience, and familiarity with GenAI-related applications within risk management and sustainable project environments. Accordingly, the resulting data represent perception-based evaluations of RM practice efficacy in relation to anticipated GenAI-related risks. These responses provide an exploratory indication of organisational preparedness and potential gaps in existing RM structures rather than a direct measurement of mature, standardised industry-wide practice.

3.4 Stage four: development and application of the fuzzy-based model to assess RM practice effectiveness

A fuzzy-based assessment model was developed using MATLAB R2024b to evaluate the perceived efficacy of RM practices in addressing the risks associated with integrating GenAI into RM within SCPs. The model comprised seven fuzzy controllers, two fuzzy subsystems, and a final controller for overall assessment quantification. Six controllers accepted two inputs and included 25 IF–THEN conditional rules to produce a single output, while the fuzzy controller representing the RSL accepted four inputs and included 625 rules. Additionally, Subsystem one processed the crisp outputs from Controllers one to three and included 125 IF–THEN rules. Similarly, Subsystem two used the outputs of Controllers 4 to 6 as inputs and generated another output based on 125 IF–THEN rules.

Finally, the overall controller combined the outputs of Subsystems 1, 2, and RSL, using 125 IF–THEN rules to produce a single overall output representing the quantified RM Practices Level (RMPL) in relation to each identified risk. The RMPL values ranged from one to 5, where one indicated very low perceived efficacy of RM practices, and five indicated very high perceived efficacy. The IF–THEN conditional statements for the entire model were formulated based on previous work by Al-Mhdawi et al. (2023b), where expert judgement was applied to ensure validity and reduce bias through multiple rounds of consultation with professionals possessing significant experience in fuzzy systems and the construction sector. The selection of experts, as indicated in Al-Mhdawi et al. (2023b), was based on their demonstrated knowledge of RM and sustainable construction, as well as their familiarity with GenAI technologies in project environments.

The model consisted of three core processes: fuzzification, fuzzy inference, and defuzzification. Triangular membership functions were employed for both input and output variables due to their simplicity, computational efficiency, and suitability for representing subjective judgements. For the fuzzy inference process, Mamdani’s Fuzzy Inference System (MFIS) was applied, given its intuitive nature, widespread use in the literature, and appropriateness for processing subjective and linguistic data. Furthermore, defuzzification was performed using the centroid of area method, a commonly accepted approach that aligns well with expert-driven evaluations. It is important to note that the fuzzy-based model does not measure actual RM performance under fully implemented GenAI conditions. Instead, it quantifies expert perceptions regarding the efficacy of existing RM practices in addressing anticipated GenAI-related risks. The model therefore serves as an analytical tool for translating subjective expert judgements into structured and comparable outputs within an exploratory assessment framework.

Following this, an expert validation interview was conducted to confirm the accuracy and practical relevance of the assessment results. The interview findings indicated a strong level of agreement among experts regarding the consistency and interpretability of the fuzzy-based outputs, suggesting that the model provides a coherent representation of expert perceptions regarding RM practice efficacy. These results should be interpreted as reflecting informed professional judgement rather than as direct evidence of real-world RM practice performance.

To further clarify the fuzzy modelling procedure, the five-point Likert-scale responses collected in the survey were mapped directly into five linguistic terms: Very Low, Low, Medium, High, and Very High. These linguistic terms were represented in the model using triangular membership functions defined on the 1–5 scale as follows: Very Low = (1, 1, 2), Low = (1.5, 2, 2.5), Medium = (2.5, 3, 3.5), High = (3.5, 4, 4.5), and Very High = (4, 5, 5). These values represent the lower bound, peak, and upper bound of each linguistic term.

The fuzzy inference system was then applied using representative IF–THEN rules to model the logical relationships between the inputs and outputs, while final crisp values were obtained through centroid defuzzification. To improve methodological transparency, representative examples of the IF–THEN rules used in the fuzzy inference system are presented. Because the complete rule base is extensive, only illustrative examples are included here. For instance: (1) IF (RS is High) AND (LI is High) THEN Stakeholder Involvement Level is High; (2) IF (WCKS is High) AND (CV is Low) THEN Risk Culture is High; and (3) IF (Subsystem one is High) AND (Subsystem two is High) AND (RSL is Medium) THEN RMPL is High. These examples illustrate the logical structure through which the model converts linguistic input assessments into intermediate outputs and the final quantified RM practices level.

Finally, sensitivity analysis was implemented as an additional validation procedure for the fuzzy-based RM practices assessment model because of its robustness, transparency, and suitability for evaluating decision-support systems derived from expert judgement. This method enables systematic examination of how controlled variations in model inputs affect the RMPL output, thereby providing insight into model behaviour, responsiveness, and internal reliability (Saltelli et al., 2008). It is particularly useful in fuzzy inference systems because it helps determine whether the model outputs remain logically stable under small perturbations in subjective input values rather than being excessively influenced by minor uncertainty in the expert ratings (Ferretti et al., 2016).

3.5 Stage five: identification of best-practice strategic responses for managing GenAI-related risks

A semi-structured survey was employed to enable experts to share their insights and identify response strategies that could effectively mitigate each category of risk. The use of semi-structured surveys to gather expert insights represents a robust methodology for identifying effective risk response strategies across diverse project categories (Appiah, 2020). This approach facilitates the elicitation of nuanced perspectives and practical recommendations from experienced professionals, thereby enriching the depth and relevance of the research findings (Surawy-Stepney et al., 2023). The adoption of a semi-structured survey approach, rather than direct one-on-one interviews, was strategically chosen to capitalise on its inherent adaptability in data collection and its potential to capture a broader range of expert insights. This approach is particularly advantageous when dealing with geographically dispersed participants or when aiming to minimise scheduling conflicts (Algassim et al., 2023).

Considering the emerging nature of GenAI adoption in the construction sector, this stage was designed to capture forward-looking expert perspectives on how organisations may respond to anticipated GenAI-related risks, rather than to document established or widely implemented practices. The survey was distributed to 17 experts, allowing them to provide their professional opinions on strategies that could be used to mitigate or avoid risks associated with integrating GenAI into RM for SCPs, based on their experience in the construction management field. Twelve of the experts returned completed surveys, and a minimum of four distinct practices were proposed for each risk category. The sample size of 12 was determined based on guidance from previous research, such as Guest et al. (2006), which suggests that thematic saturation in qualitative research is typically reached with 12 or fewer participants, particularly when the respondents are highly experienced in the subject matter. The responses collected from the experts were subsequently analysed to eliminate redundant strategies and consolidate similar responses that reflected overlapping approaches. This refinement process involved comparing, discussing, and synthesising the responses to ensure clarity, consistency, and thematic integrity in the final list of recommended strategies.

The final set of strategies should therefore be interpreted as indicative best-practice guidance derived from expert consensus, rather than as prescriptive or universally validated solutions. Their value lies in highlighting practical directions for improving organisational preparedness and supporting decision-making in SCPs as GenAI adoption continues to evolve.

4 Results and discussion

4.1 Risks identification and classification

The SLR identified 30 key risks associated with the integration of GenAI into RM within SCPs. These risks were organised into five main categories: input quality risks (Cat1), technological adaptability risks (Cat2), ethical and governance risks (Cat3), information integrity risks (Cat4), and financial risks (Cat5), as shown in Figure 4. These categories were derived based on the underlying sources of the identified risk factors, enabling a structured classification framework.

FIGURE 4

The first category includes inaccurate or incomplete data (R01), data unavailability (R02), data bias (R03), extensive data complexity (R04), and data overfitting (R05). These risks highlight the foundational importance of high-quality, comprehensive, and representative datasets in ensuring accurate and reliable GenAI performance. Poor data quality or insufficient data access can lead to biased models and erroneous risk predictions, undermining trust and decision-making efficacy within RM processes. In the context of SCPs, these risks are amplified due to the reliance on heterogeneous data sources, including environmental metrics, lifecycle performance data, and regulatory compliance information, which increases the difficulty of maintaining data consistency and reliability.

The second category, Technological Adaptability Risks (R06–R16), covers inconsistent connectivity (R06), insufficient training (R07), incompatibility with legacy systems (R08), absence of expertise (R09), misinterpretation of results (R10), human error (R11), failure of model training (R12), wrong model selection (R13), lack of awareness (R14), cultural resistance (R15), and lack of trust (R16). These risks collectively represent the technical and organisational barriers that hinder the seamless integration of GenAI technologies into construction workflows. Such risks are particularly significant in SCPs, where project delivery is often fragmented across multiple stakeholders and temporary project teams, limiting the standardisation and continuity required for effective technology integration. These risks demonstrate the sector’s need for stronger digital infrastructure, continuous professional development, and robust validation mechanisms to enhance the adaptability of GenAI-based RM processes.

The third category, Ethical and Governance Risks (R17–R22), includes algorithm bias (R17), absence of regulatory frameworks (R18), confidentiality breaches (R19), unclear responsibility (R20), legal risk (R21), and non-compliance with policies (R22). These risks underline the importance of establishing transparent, accountable, and ethically grounded governance structures for GenAI adoption. Without proper regulation, oversight, and ethical considerations, the integration of AI tools into RM may expose projects to compliance violations, privacy breaches, and reputational damage. These risks are heightened in SCPs due to stricter regulatory scrutiny, sustainability certification requirements, and increased expectations for transparency in environmental and social performance reporting.

The fourth category, Information Integrity Risks (R23–R26), comprises data breach (R23), overdependence on synthetic data (R24), inconsistent data formatting (R25), and vulnerability to adversarial cyberattacks (R26). These risks directly threaten the authenticity, accuracy, and security of the information used in GenAI models. Inadequate protection of information integrity can compromise model reliability, leading to corrupted insights and potential project-wide disruptions. In SCP environments, where decisions often involve long-term environmental and safety implications, compromised data integrity may have extended consequences beyond immediate project performance.

The fifth category, Financial Risks (R27–R30), includes high initial costs (R27), return-on-investment outcome discrepancies (R28), training cost and staff development (R29), and customisation and integration expenses (R30). These financial considerations highlight the substantial investment required to implement GenAI technologies effectively. Beyond technical feasibility, economic sustainability remains a critical determinant of successful GenAI adoption, especially for large-scale construction projects with constrained budgets. This is particularly relevant for SCPs, where additional costs associated with sustainability compliance and performance monitoring may intensify financial pressures on project delivery.

To this end, these 30 risks illustrate the multi-dimensional nature of risks associated with GenAI integration in RM for SCPs, spanning data integrity, technical capability, organisational culture, ethical compliance, and financial viability.

4.2 Risk significance analysis

A part of the assessment model, as shown in Figure 5, was used to evaluate the RSL based on four factors: IL, PL, DL, and EL. The results of this assessment are summarised in Table 5, which presents the mean scores for each indicator, along with the ranking of risks within their respective categories and overall. The analysis revealed that the most critical risks influencing the integration of GenAI into risk management (RM) for sustainable construction projects (SCPs) include R11: Human Error, ranked first overall; R02: Data Unavailability, ranked second; R07: Insufficient Training, ranked third; R23: Data Breach, ranked fourth; and R03: Data Bias, ranked fifth.

FIGURE 5

These top-ranked risks indicate that the reliability and efficacy of GenAI-driven RM systems depend heavily on data accessibility, model transparency, user competency, and information security. Furthermore, integrating the RSL into the subsequent fuzzy-based efficacy quantification model improved the accuracy of the evaluation. The use of RSLs as weighted inputs in measuring the efficacy of RM practices added significant value by linking risk criticality directly to management performance. This approach enhances model reliability and ensures a realistic and context-sensitive assessment of RM practice efficacy within AI-enabled sustainable construction environments.

4.3 RM practices efficacy survey respondents’ profile

The survey respondents represented a diverse range of roles within the construction industry. The majority were Project Managers (71%), followed by other project management roles (13%), academics (12%), and consultants (5%). This distribution indicates a strong representation of project management professionals, suggesting that most participants were directly engaged in the supervision and execution of construction projects. However, the relatively smaller proportion of consultants and academics may have limited the diversity of viewpoints, particularly concerning advisory and theoretical insights. In terms of professional experience, respondents demonstrated a wide range of project management backgrounds. The largest group had 6–15 years of experience (29%), followed closely by those with 16–25 years (27%) and more than 25 years (24%), while 20% reported 1–5 years of experience. This composition suggests that the sample primarily consisted of mid-to senior-level professionals, reflecting perspectives from individuals with extensive exposure to real-world construction risk management practices. The range of experience levels contributes to a balanced understanding of both emerging and long-established challenges across professional stages. Regarding educational qualifications, the respondents were generally well educated, with 52% holding bachelor’s degrees, 36% master’s degrees, and 12% doctorate degrees. This indicates that a substantial portion of the sample possessed postgraduate-level qualifications, reinforcing the credibility of the insights obtained, especially in discussions relating to advanced construction and risk management frameworks.

With respect to familiarity with GenAI in construction management, the survey results revealed that 55% of respondents considered themselves to have intermediate experience, 38% were at the beginner level, and only 7% identified as experts. This distribution reflects the early stage of GenAI adoption in the construction sector, where most professionals are still developing their practical understanding and applications of this emerging technology. These findings highlight the importance of targeted training and capacity-building initiatives to strengthen the effective integration of GenAI into construction risk management practices. Figure 6 presents detailed visual summaries of the respondents’ profiles based on their professional roles, years of experience, educational qualifications, and levels of GenAI familiarity within construction management.

FIGURE 6

4.4 Architecture of the developed model

As outlined in Stage Four of the methodology, the authors used the Fuzzy Logic Toolbox in MATLAB R2024b to design and implement the proposed RM practices efficacy assessment model. The model comprised 16 input variables in total, including 12 RM-practice input variables distributed across six two-input fuzzy controllers (RS, LI, OCC, UCT, AET, AR, WCKS, CV, UMSO, PB, RMIMP, and PPI) and four risk significance input variables (IL, PL, DL, and EL) used in a separate RSL controller. In addition, the model included seven fuzzy controllers, two fuzzy subsystems, 650 conditional statements for the controllers and subsystems, and a single output variable, namely, the RM practices level (RMPL), as presented in Figure 5. During model implementation, input values rated through the fuzzy survey responses served as inputs for developing the fuzzy inference system. These values were fuzzified using a triangular fuzzy membership function. In addition, 650 IF–THEN statements were developed under the Mamdani inference system. Furthermore, the RMPL for each respondent was computed through defuzzification of the output membership function using the centroid of area method. Table 6 shows the RMPL level in relation to the top two significant risks in each risk category.

This multi-layered fuzzy logic structure facilitated a comprehensive and systematic quantification of the efficacy of RM practices, enabling a more rigorous and data-driven evaluation of how effectively these practices address GenAI-related risks (Al-Mhdawi et al., 2022; Al-Mhdawi et al., 2023b; Mohamed et al., 2025b). The fuzzy logic controllers were designed using structured IF–THEN rule sets, as detailed in Stage Four of the methodology. The overall architecture of the fuzzy-based efficacy assessment model for RM practices is illustrated in Figure 5. To further enhance the interpretability of the fuzzy model, three-dimensional surface visualisations were produced using MATLAB’s Fuzzy Logic Surface Viewer. These visualisations demonstrate how the output representing the quantified efficacy level of RM practices varies in response to changes in input variables. Figures 7–9 present surface mapping examples for the fuzzy controller, with each plot showing two input variables and one output. For instance, Figure 7 illustrates how the presence of conflicting values (CV) and the willingness to collaborate and share knowledge (WCKS) influence the overall risk culture within the organisation as an input to measure and evaluate RMPL, with blue shades representing the low-risk culture within an organisation and green shades representing the high-risk culture. These visual tools provide an intuitive and transparent understanding of how different factors interact within the fuzzy model, thereby supporting more informed and evidence-based decision-making in evaluating RM practices.

FIGURE 7

FIGURE 8

FIGURE 9

Based on Table 5, which ranks each risk factor according to its significance, the top two risks in each category were identified as follows: for Category 1 (R02 and R03); for Category 2 (R11 and R07); for Category 3 (R20 and R19); for Category 4, R23 and R25; and for Category 5, R28 and R27. Table 5 reveals only small variations between the RMPL values across the identified risks, indicating that the overall organisational capacity to manage GenAI-related risks remains relatively consistent but moderate. This narrow range of efficacy scores suggests that the assessed RM practices show relatively limited variation across the identified risks. This may indicate a broadly similar level of perceived efficacy across risk categories, although it does not on its own provide conclusive evidence about overall organisational preparedness. Such consistency in perceived RM performance across risk categories has also been observed in complex project environments where emerging technologies are introduced without fully differentiated risk governance frameworks (Al-Mhdawi et al., 2023a).

The top-ranked risk, R23, achieved the highest RMPL, demonstrating that current RM practices are most effective when dealing with data security and cyber-related threats. This outcome suggests that RM practices are perceived to be relatively more effective for addressing data security and cyber-related risks compared to several other identified risks, which aligns with prior studies emphasising the prioritisation of cybersecurity and data protection in digital transformation initiatives within construction and infrastructure sectors (Rane, 2023; Regona et al., 2024). However, this interpretation is based on comparative RMPL values and should not be considered direct evidence of broader organisational readiness or investment commitment.

Additionally, the second-ranked risk, R19, also received a high RMPL value, which may indicate comparatively stronger perceived efficacy of RM practices in addressing information sensitivity and intellectual property-related risks. This observation is consistent with existing literature highlighting increased organisational attention to data governance and confidentiality risks in AI-enabled systems (Regona et al., 2024; Mohamed et al., 2025a). However, this interpretation reflects respondent perceptions captured through RMPL values and should not be taken as direct evidence of the existence or maturity of formal governance procedures within organisations.

The third-ranked risk, R11, further illustrates the impact of workforce capability on GenAI performance. Its relatively high RMPL value may suggest that respondents perceive RM practices related to human factors, such as training and procedural controls, to be comparatively more effective than in other areas, which corresponds with previous research identifying training and human competency as critical enablers of successful AI implementation in construction management (Afzal et al., 2020; Nabawy and Gouda Mohamed, 2024). However, the persistence of human-related risks still reflects limitations in automation and oversight.

Conversely, the lowest-ranked risks R28, R02, and R03 reflect notable weaknesses in current RM practices. These results may indicate perceived limitations in current readiness to address data quality and accessibility-related risks, as well as uncertainty regarding the financial justification for GenAI adoption. This finding is consistent with prior studies that identify data availability, data quality, and cost-related uncertainties as key barriers to AI adoption in the built environment (Jallow et al., 2023; Nabawy and Gouda Mohamed, 2024). The findings suggest that organisations may not yet have fully developed mechanisms to ensure reliable data collection, curation, and integration, which can limit GenAI’s analytical potential and affect decision-making outcomes. However, these results should be interpreted as reflecting perceived efficacy and early-stage professional judgement, rather than direct measurement of mature, fully institutionalised GenAI practice in the construction sector.

Overall, the relatively close RMPL values across risks indicate that organisations are not fully confident in the efficacy of their RM systems in dealing with emerging GenAI challenges, particularly in the absence of specialised frameworks and adaptive training initiatives. The results also reveal that conflict value (CV) tends to be high among stakeholders for risks such as R19, R25, R28, R23, and R07, suggesting weak consensus and a fragmented risk culture when confronting GenAI-related risks. This observation aligns with previous studies indicating that emerging technologies often introduce uncertainty and divergent stakeholder perceptions, which can weaken alignment and slow adoption processes (Yao and Garcia de Soto, 2024). Furthermore, the average stakeholder interest and engagement levels recorded across categories indicate limited involvement in AI-driven RM decision processes, which may reduce the overall efficacy of implemented risk mitigation strategies. Similar patterns of limited stakeholder engagement in digital innovation processes have been reported in construction projects undergoing digital transformation (Pan and Zhang, 2021). Enhancing collaboration, clarifying responsibilities, and fostering a shared understanding of AI-related risks are therefore essential steps towards improving the perceived readiness and resilience of RM practices in sustainable construction projects.

4.5 Sensitivity analysis

To strengthen the validation of the fuzzy RM practices efficacy model, a sensitivity analysis was conducted to examine how controlled variations in input values affect the model outputs. Sensitivity analysis is widely used to assess how changes in model inputs influence outputs and to evaluate the robustness of model-based conclusions (Pannell, 1997). In multi-criteria decision contexts, it is particularly useful for identifying influential inputs and testing whether the interpretation of results remains stable under controlled perturbation (Więckowski and Sałabun, 2023). In this study, the sensitivity analysis was applied across the model inputs and risks, while RS and the risks R02, R11, R20, and R25 are presented here only as an illustrative example to demonstrate the procedure and the nature of the resulting changes in RMPL. RS was varied by ±10% while the remaining inputs were held constant, and the resulting RMPL values were examined for four representative risks drawn from different categories, namely, R02, R11, R20, and R25. The results showed that the model responded in a directionally consistent manner across all four risks. When RS was increased by 10%, the RMPL values increased in all cases, while a 10% reduction in RS led to a decrease in RMPL values for all four risks. For R02, the baseline RMPL value of 3.142 changed to 3.216 under RS +10% and to 3.024 under RS −10%, corresponding to +2.36% and −3.76%, respectively.

Overall, this illustrative example shows that the model is responsive to perturbations in RS in a controlled and interpretable manner. This is consistent with the role of sensitivity analysis as a tool for examining output responsiveness and the robustness of conclusions rather than for applying a fixed universal threshold of acceptable change (Pannell, 1997). At the same time, the results indicate that when baseline RMPL values are relatively close, even limited perturbations may influence the relative ordering of some risks. This observation is consistent with recent work on sensitivity analysis in multi-criteria decision analysis, which emphasises that ranking stability should be examined carefully when only small differences separate alternatives or cases (Więckowski and Sałabun, 2023). Accordingly, the sensitivity analysis supports the internal coherence of the developed fuzzy model, while also indicating that fine-grained comparisons between risks with very similar RMPL values should be interpreted cautiously. Figure 10 presents the baseline, RS −10%, and RS +10% RMPL values for the four selected risks as an illustrative example of the sensitivity analysis procedure.

FIGURE 10

4.6 Model validation and verification

As outlined in Stage Four, a validation exercise was conducted involving a panel of senior management professionals who were invited to review the methodological steps and outputs of the proposed fuzzy assessment model. The evaluation focused on two key criteria: the perceived practicality of use and ease of use of the model. A total of six construction experts participated in this process. It is important to note that these individuals were independent of the earlier phases of data collection and model development, which helped to strengthen the objectivity of the evaluation process. The professional backgrounds of the panel members are detailed in Table 6.

The purpose of this validation exercise was not to verify real-world performance under fully implemented GenAI conditions, but rather to assess whether the model provides a coherent, usable, and contextually relevant framework for exploring the perceived efficacy of RM practices in relation to anticipated GenAI-related risks in SCPs. In this sense, the validation focused on the model’s suitability as an exploratory decision-support tool within an emerging area of practice.

Feedback from the panel indicated a high level of agreement regarding the model’s relevance and usability. All participants acknowledged that the model was practical for application within the construction industry and intuitive to use. They reported that the model was straightforward to interpret and could be readily customised to suit the distinctive features of SCPs. Moreover, the panel confirmed that the fuzzy assessment framework addressed the critical dimensions necessary for examining the perceived efficacy of RM practices when considering GenAI-related risks in SCPs. These assessments support the internal validity, interpretability, and contextual suitability of the model. They also indicate that the framework has potential value for both academic analysis and practitioner reflection, particularly as a structured means of examining preparedness and identifying possible gaps in RM arrangements before wider GenAI adoption becomes established in construction practice.

4.7 Indicators of the efficacy assessment model

4.7.1 Stakeholder involvement level

Effective RM in construction projects requires the active engagement of stakeholders throughout all phases, from risk identification and assessment to allocation, monitoring, and control (Song et al., 2025). Stakeholders such as project managers, clients, contractors, and consultants contribute unique expertise and perspectives, which enhance communication, clarify risk ownership, and support transparent and collaborative decision-making processes (Bal et al., 2013; Suffo and Brome, 2018; Afzal et al., 2021). The extent of stakeholder involvement often depends on their relative influence and interest in the project, both of which affect how risks are perceived, prioritised, and addressed.

In the context of SCPs, where complexity and long-term performance requirements are heightened, stakeholder involvement becomes even more critical, particularly with the integration of GenAI into RM practices (Ajirotutu et al., 2024; Toochukwu, 2025). Although GenAI offers advanced capabilities for predictive analysis and data-driven decision-making, its usefulness depends on stakeholders’ ability to interpret, validate, and apply generated insights appropriately (Pan and Zhang, 2021; Holzmann and Lechiara, 2022; Giraud et al., 2023). Accordingly, stakeholder involvement is treated in this study as an important indicator for assessing the perceived efficacy of GenAI-enabled RM practices, because it reflects the extent to which AI-supported insights can be accepted, interpreted, and operationalised within collaborative SCP environments.

4.7.2 Organisational communication level

In SCPs, effective communication, defined as the systematic exchange of project-related information, is critical throughout the project lifecycle. Project communication management ensures that information is accurately generated, shared, and utilised, directly influencing risk identification, stakeholder alignment, and overall project performance (Whyte et al., 2016; Yao and García de Soto, 2024; Wen et al., 2025). As SCPs often involve interdisciplinary teams and multiple stakeholders, the demand for clear, timely, and reliable communication intensifies.

Within the context of RM, organisational communication depends not only on formal communication structures but also on communication culture and the accessibility and usability of digital tools. This becomes especially important in GenAI-enabled RM because AI-generated outputs must be communicated across project interfaces in a form that supports interpretation, validation, and timely action (Pan and Zhang, 2021; Fridgeirsson et al., 2023; Wen et al., 2025). In SCPs, where sustainability reporting, compliance evidence, and lifecycle information flows are central, organisational communication can therefore be understood as an indicator of the perceived capacity to translate AI-generated risk insights into coordinated project action.

4.7.3 Risk culture

Risk culture in SCPs refers to the collective values, beliefs, and attitudes that influence how individuals and teams perceive and respond to risk within an organisation (Moczydłowska et al., 2023). A strong risk culture promotes proactive risk identification, open communication, and shared accountability (Hassan et al., 2022; Sheedy and Canestrari-Soh, 2023). Organisational leaders play a central role in shaping and reinforcing this culture (Summerill et al., 2010; Hassan et al., 2022). Hassan et al. (2022) explain that the maturity of risk culture depends on factors such as collaboration, knowledge sharing, and openness in communication. Conversely, conflicting values or limited engagement can hinder effective RM. In the context of GenAI integration, risk culture becomes particularly significant because trust in AI-generated outputs, willingness to question automated recommendations, and openness to new forms of decision support are all shaped by organisational norms and behavioural expectations (Holzmann and Lechiara, 2022; Hashfi and Raharjo, 2023; Liang et al., 2024). This indicator therefore captures the perceived readiness of organisations to engage with AI-informed RM practices in SCP settings.

4.7.4 Risk management training level

Providing RM training improves a team’s capability to identify, assess, and mitigate risks effectively while reinforcing consistent application of tools and techniques (Agyekum-Mensah and Knight, 2017; Aladağ, 2023). The quality of training depends on the availability of skilled personnel and adequate resources. GenAI integration further increases the importance of training, as practitioners must be able to understand, interpret, and critically evaluate AI-generated outputs rather than rely on them uncritically (Pan and Zhang, 2021; Aladağ, 2023; Taiwo et al., 2024). This risk is particularly relevant in SCPs, where decisions often carry long-term sustainability implications and require defensible judgement across environmental, economic, and social dimensions (Dobrovolskienė and Tamošiūnienė, 2015; Sabini and Silvius, 2023). Therefore, this indicator reflects the perceived efficacy of organisational capability to support the responsible use of GenAI within RM processes.

4.7.5 Risk management policies

Risk management policies provide a structured framework guiding how risks are identified, assessed, and controlled, supporting organisations in achieving strategic objectives and maintaining resilience (Boughaba and Bouabaz, 2020; Wen et al., 2025). Their implementation depends on leadership commitment, organisational support, and recognition of RM benefits. In GenAI-enabled RM, policies must also accommodate risks related to data governance, accountability, transparency, confidentiality, and regulatory compliance (Pillai and Matus, 2020; Liang et al., 2024; Regona et al., 2024). These risks are particularly important in SCPs, where governance expectations are heightened by sustainability certification demands, reporting obligations, and long-term performance accountability (Mori Junior et al., 2016; Elseknidy et al., 2025; Tettey et al., 2025). Accordingly, this indicator reflects the perceived efficacy of existing policy arrangements for accommodating emerging AI-related risks in SCPs.

4.7.6 Continuous risk monitoring level

Continuous risk monitoring enables organisations to track and respond to emerging risks throughout the project lifecycle (Zhao et al., 2023). It supports early detection, timely intervention, and organisational learning. Within GenAI-enabled RM, continuous monitoring becomes increasingly important because AI systems operate in data-rich and evolving environments, where output reliability may change as project conditions change (Yazdi et al., 2024; Wen et al., 2025). In SCPs, this need is intensified by long project durations, changing sustainability targets, compliance checks, and lifecycle accountability requirements (Fischer et al., 2016; Sabini and Silvius, 2023; Elseknidy et al., 2025). This indicator therefore reflects the perceived capacity of organisations to maintain ongoing oversight of AI-related risks and adapt their RM responses over time.

4.8 Best practices for mitigating GenAI-Related risks in RM for SCPs

As described in Stage Five, a semi-structured survey was used to capture expert insights on strategies for mitigating risks associated with integrating GenAI into RM practices for SCPs. This approach enabled the collection of context-specific and experience-based perspectives from professionals across the construction sector. The survey was distributed to 17 construction management professionals, of whom 12 responded, resulting in a 70.6% response rate. In line with Guest et al. (2006), this sample size is sufficient to achieve thematic saturation when participants possess substantial expertise. The synthesised strategies are summarised in Table 8. In the absence of widespread, mature deployment of GenAI in construction practice, the identified strategies should be interpreted as forward-looking and exploratory rather than as established or validated industry practices. They reflect informed expert judgement regarding how organisations may respond to anticipated GenAI-related risks within SCP environments rather than documented evidence of fully implemented solutions. This interpretation is consistent with prior research on emerging technologies, where early-stage studies often rely on expert-driven insights to anticipate risks and guide preparedness before large-scale implementation occurs (Ojiako et al., 2026; Yigitcanlar et al., 2022). Although several of the identified strategies may appear general, they were elicited specifically in relation to GenAI-related risks in SCPs. At this stage of adoption, many responses focus on foundational measures such as training, governance, data quality, and monitoring. This pattern aligns with previous studies on digital transformation in construction, which highlight that early responses to emerging technologies tend to prioritise organisational readiness, capability development, and governance structures before more specialised practices are fully developed (Pan and Zhang, 2021; Al Jundi and Chaudhry, 2025).

However, their relevance lies in how these measures are applied within the distinctive characteristics of SCPs, including fragmented project delivery structures, temporary multi-party collaboration, contractual interfaces, design–construction coordination, and sustainability-driven performance requirements. These sector-specific conditions have been widely recognised as key factors shaping the implementation of innovation and risk management practices in construction projects (Bal et al., 2013; Sabini and Silvius, 2023), reinforcing the need to interpret the identified strategies within the SCP context rather than as generic organisational responses.

These findings suggest that while foundational RM practices remain applicable, their implementation in SCPs requires contextual adaptation to address the combined challenges of sustainability objectives and emerging AI technologies. More visible sector-specific differentiation is likely to emerge when such strategies are translated into project-level implementation plans, resource allocation decisions, and operational practices.

5 Conclusion

This research presents a structured and exploratory risk assessment framework developed to evaluate the significance of risks associated with integrating GenAI into RM within the context of SCPs, with a particular focus on assessing perceived organisational preparedness rather than established implementation performance. The study also examines the perceived efficacy of existing RM practices in addressing identified risks and identifies appropriate best practices for mitigating these risks. The research was designed to identify and categorise key risk factors emerging from GenAI integration, quantify their significance based on probability, impact, detectability, and exposure, and assess how effectively existing RM practices are perceived to address the most critical risks within each category. A structured, multi-stage methodological approach was adopted. The initial stage involved a systematic literature review, resulting in 62 high-quality studies that informed the development of a comprehensive GenAI risk taxonomy. A multi-criteria assessment model based on fuzzy set theory was then developed to evaluate both risk significance and RM practice efficacy. Survey data from 80 construction professionals were used to capture expert judgement, which was subsequently analysed through the fuzzy model. Additional expert input was incorporated to refine the model and identify best-practice mitigation strategies.

The findings indicate that human error, data unavailability, and insufficient training represent the most critical perceived challenges affecting the integration of GenAI into RM for SCPs. These risks reflect the socio-technical nature of GenAI adoption in construction, where data dependency, human interpretation, and organisational capability interact within complex project environments. In SCPs, these challenges are further amplified due to fragmented delivery structures, multi-stakeholder coordination, regulatory pressures, and sustainability performance requirements (Bal et al., 2013; Sabini and Silvius, 2023; Datta et al., 2023).

The results further indicate that the efficacy of current RM practices in addressing GenAI-related risks is perceived to be moderate to low. This outcome suggests that existing RM systems, which were primarily developed for conventional project risks, may not yet be fully adapted to address the uncertainty, data dependency, and governance challenges associated with GenAI integration (Afzal et al., 2021; Regona et al., 2024). Key limitations include insufficient expertise, limited awareness of GenAI-specific risks, weak alignment between stakeholders, and a lack of structured organisational support mechanisms. Importantly, these findings should be interpreted as perception-based insights generated under conditions of early-stage GenAI adoption rather than as direct evidence of mature, fully implemented RM practices in the construction sector. The study therefore provides an early-warning and preparedness-oriented perspective, highlighting potential gaps that may become critical as GenAI adoption increases.

The proposed fuzzy-based assessment model demonstrates that effective evaluation of GenAI-related risks in SCPs requires simplicity, adaptability, and the ability to incorporate subjective expert judgement within complex decision environments. The model captures RM efficacy across six interrelated domains, enabling a structured and context-sensitive analysis of organisational readiness. Furthermore, the study identifies a set of best-practice strategies derived from expert input. Although several of these strategies may appear broadly applicable, their relevance in this study lies in their interpretation within the distinctive context of SCPs, where temporary project organisations, contractual fragmentation, and sustainability-driven constraints shape how risk management practices are implemented (Toochukwu, 2025; Abdous et al., 2024). These strategies emphasise data governance, training, ethical oversight, stakeholder coordination, and phased implementation as critical enablers of effective GenAI integration.

5.1 Theoretical and practical implications

This research contributes to the literature at the intersection of GenAI and RM in SCPs through the development of a context-sensitive and perception-based assessment framework grounded in fuzzy set theory. It extends existing RM research by explicitly addressing the gap between emerging technological potential and organisational preparedness, particularly within the unique delivery conditions of sustainable construction projects. The study also advances theoretical understanding by demonstrating how RM efficacy can be conceptualised as a multi-dimensional construct influenced by stakeholder dynamics, organisational processes, and technology-related uncertainty. This perspective is particularly relevant for emerging technologies such as GenAI, where formalised practices are still evolving (Mohamed et al., 2026a; Yigitcanlar et al., 2022). From a practical perspective, the findings provide actionable insights for construction practitioners and policymakers. The identification of critical risks highlights the need for targeted investment in data infrastructure, workforce capability, and governance systems. The proposed assessment model offers a structured tool for organisations to evaluate their preparedness and identify gaps in RM practices before large-scale GenAI adoption occurs.

5.2 Research limitations and future directions

Despite its contributions, this study has several limitations. First, the research relies on expert judgement collected through surveys. Although many respondents possessed substantial experience in construction management, only a small proportion had advanced experience with GenAI applications. As a result, the findings reflect informed professional perceptions rather than direct observation of mature, large-scale implementation in practice. Second, the study is based primarily on UK-based experts, which may limit generalisability. Differences in regulatory environments, digital maturity, and resource availability may influence how GenAI-related risks are perceived and managed across regions. Future research should extend this work through longitudinal studies and real-world case applications as GenAI adoption becomes more established in construction practice. Such studies would enable validation of the proposed framework under operational conditions and support the development of more implementation-focused risk management models.

References

• 1

  AbdousM.AslaniA.NoorollahiY.ZahediR.yousefiH. (2024). Design and analysis of zero-energy and carbon buildings with renewable energy supply and recycled materials. Energy Build.324, 114922. 10.1016/j.enbuild.2024.114922

  • CrossRef
  • Google Scholar

• 2

  AbiodunO. I.JantanA.OmolaraA. E.DadaK. V.UmarA. M.LinusO. U.et al (2019). Comprehensive review of artificial neural network applications to pattern recognition. IEEE Access7, 158820–158846. 10.1109/access.2019.2945545

  • CrossRef
  • Google Scholar

• 3

  AfzalF.YunfeiS.JunaidD.HanifM.S. (2020). Cost-risk contingency framework for managing cost overrun in metropolitan projects: Using fuzzy-AHP and simulation. Int. J. Manag. Proj. Bus., 13 (5), 1121–1139. 10.1108/IJMPB-07-2019-0175

  • CrossRef
  • Google Scholar

• 4

  AfzalF.YunfeiS.NazirM.BhattiS. M. (2021). A review of artificial intelligence based risk assessment methods for capturing complexity-risk interdependencies: cost overrun in construction projects. Int. J. Manag. Proj. Bus.14 (2), 300–328. 10.1108/ijmpb-02-2019-0047

  • CrossRef
  • Google Scholar

• 5

  Agyekum-MensahG.KnightA. D. (2017). The professionals’ perspective on the causes of project delay in the construction industry. Eng. Construction Architectural Management24 (5), 828–841. 10.1108/ecam-03-2016-0085

  • CrossRef
  • Google Scholar

• 6

  AjirotutuR. O.MatthewB.GarbaP.JohnsonS. O. (2024). AI-driven risk mitigation: transforming project management in construction and infrastructure development. World J. Adv. Eng. Technol. Sci.13 (02), 611–623. 10.30574/wjaets.2024.13.2.0628

  • CrossRef
  • Google Scholar

• 7

  Al JundiY.ChaudhryH. N. (2025). A critical review of the carbon–energy nexus within the construction sector’s embodied emissions: a case study in the United Arab Emirates. Energies18 (10), 2654. 10.3390/en18102654

  • CrossRef
  • Google Scholar

• 8

  Al-MhdawiM. K. S.BritoM. P.Abdul NabiM.El-AdawayI. H.OnggoB. S. (2022). Capturing the impact of COVID-19 on construction projects in developing countries: a case study of Iraq. J. Management Engineering38 (1), 05021015. 10.1061/(asce)me.1943-5479.0000991

  • CrossRef
  • Google Scholar

• 9

  Al-MhdawiM. K. S.QaziA.AlzarradA.DacreN.RahimianF.BuniyaM. K.et al (2023a). Expert evaluation of ChatGPT performance for risk management process based on ISO 31000 standard. SSRN Electron. J., 1–6. 10.2139/ssrn.4504409

  • CrossRef
  • Google Scholar

• 10

  Al-MhdawiM. K. S.BritoM.OnggoB. S.QaziA.O’connorA.NamianM. (2023b). Construction risk management in Iraq during the COVID-19 pandemic: challenges to implementation and efficacy of practices. J. Constr. Eng. Manag.149 (9), 04023086. 10.1061/jcemd4.coeng-13099

  • CrossRef
  • Google Scholar

• 11

  Al-MhdawiM. K. S.BritoM.OnggoB. S.QaziA.O'ConnorA. (2024a). COVID-19 emerging risk assessment for the construction industry of developing countries: evidence from Iraq. Int. Journal Construction Management24 (7), 693–706. 10.1080/15623599.2023.2169301

  • CrossRef
  • Google Scholar

• 12

  Al-MhdawiM. K. S.O'ConnorA.QaziA. (2024b). Structural equation modeling and fuzzy set theory: advancing risk assessment in oil and gas construction projects. Environ. Impact Assess. Rev.109, 107622. 10.1016/j.eiar.2024.107622

  • CrossRef
  • Google Scholar

• 13

  Al-MhdawiM. K. S.QaziA.KarakhanA. A.RahimianF. P.AbualqumbozM.Al MhdawiA. K.et al (2024c). Implementation of a combined fuzzy controller model to enhance risk assessment in oil and gas construction projects. IEEE Access12, 68319–68333. 10.1109/access.2024.3399129

  • CrossRef
  • Google Scholar

• 14

  AladağH. (2023). Assessing the accuracy of ChatGPT use for risk management in construction projects. Sustainability15 (22), 16071. 10.3390/su152216071

  • CrossRef
  • Google Scholar

• 15

  AlgassimH.SepasgozarS. M.OstwaldM.DavisS. (2023). A qualitative study on factors influencing technology adoption in the architecture industry. Buildings13 (4), 1100. 10.3390/buildings13041100

  • CrossRef
  • Google Scholar

• 16

  AlmashhourR.Al-MhdawiM. K. S.DaghfousA.QaziA.OjiakoU. (2025). Traditional to sustainable risk management in the construction industry: a systematic literature review. Int. J. Manag. Proj. Bus.18 (3), 528–565. 10.1108/ijmpb-01-2025-0021

  • CrossRef
  • Google Scholar

• 17

  AltT.IbischA.MeiserC.WilhelmA.ZimmerR.DitzJ.et al (2024). Generative AI models: opportunities and risks for industry and authorities. arXiv Preprint arXiv:2406.04734. 10.48550/arXiv.2406.04734

  • CrossRef
  • Google Scholar

• 18

  AltafM.AlaloulW. S.MusaratM. A.QureshiA. H. (2023). Life cycle cost analysis (LCCA) of construction projects: sustainability perspective. Environ. Dev. Sustain.25 (11), 12071–12118. 10.1007/s10668-022-02579-x

  • CrossRef
  • Google Scholar

• 19

  AnyszH.ApolloM.GrzylB. (2021). Quantitative risk assessment in construction disputes based on machine learning tools. Symmetry13 (5), 744. 10.3390/sym13050744

  • CrossRef
  • Google Scholar

• 20

  AppiahB. L. (2020). Risk management processes and analysis in projects construction industry. Environ. Eng.5 (4), 92-101.

  • Google Scholar

• 21

  ArifM.KatafygiotouM.MazroeiA.KaushikA.ElsarragE. (2016). Impact of indoor environmental quality on occupant well-being and comfort: a review of the literature. Int. J. Sustain. Built Environ.5 (1), 1–11. 10.1016/j.ijsbe.2016.03.006

  • CrossRef
  • Google Scholar

• 22

  BalM.BrydeD.FearonD.OchiengE. (2013). Stakeholder engagement: achieving sustainability in the construction sector. Sustainability5 (2), 695–710. 10.3390/su5020695

  • CrossRef
  • Google Scholar

• 23

  BehroozH.LipizziC.KorfiatisG.IlbeigiM.PowellM.NouriM. (2023). Towards automating the identification of sustainable projects seeking financial support: an AI-powered approach. Sustainability15 (12), 9701. 10.3390/su15129701

  • CrossRef
  • Google Scholar

• 24

  BhattiM. A. A.NazirM. U. (2024). The impact of project process management on sustainable project success in the construction sector: the moderating role of risk management practices. Bull. Bus. Econ. (BBE)13 (2), 1065–1072. 10.61506/01.00462

  • CrossRef
  • Google Scholar

• 25

  BirdD. K. (2009). The use of questionnaires for acquiring information on public perception of natural hazards and risk mitigation–a review of current knowledge and practice. Nat. Hazards Earth Sys. Sci.9 (4), 1307-1325. 10.5194/nhess-9-1307-2009

  • CrossRef
  • Google Scholar

• 26

  BoughabaA.BouabazM. (2020). Identification and risk management related to construction projects. Adv. Comput. Des.5 (4), 445–465. 10.12989/acd.2020.5.4.445

  • CrossRef
  • Google Scholar

• 27

  BrymanA. (2016). Social research methods. Oxford university press.

  • Google Scholar

• 28

  CaldasR. D.RodriguesA.GilE. B.RodriguesG. N.VogelT.PelliccioneP. (2020). “A hybrid approach combining control theory and AI for engineering self-adaptive systems,” in Proceedings of the IEEE/ACM 15th International Symposium on Software Engineering for Adaptive and Self-Managing Systems, 9–19.

  • Google Scholar

• 29

  CaoY.ShengQ. Z.McAuleyJ.YaoL. (2023). Reinforcement learning for generative AI: a survey. arXiv Preprint arXiv:2308.14328. 10.48550/arXiv.2308.14328

  • CrossRef
  • Google Scholar

• 30

  ChartresN.BeroL. A.NorrisS. L. (2019). A review of methods used for hazard identification and risk assessment of environmental hazards. Environment international123, 231-239. 10.1016/j.envint.2018.11.060

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 31

  ChoiS. W.LeeE. B.KimJ. H. (2021). The engineering machine-learning automation platform (EMAP): a big-data-driven AI tool for contractors’ sustainable management solutions for plant projects. Sustainability13 (18), 10384. 10.3390/su131810384

  • CrossRef
  • Google Scholar

• 32

  ChouJ. S.ChongP. L.LiuC. Y. (2024). Deep learning-based chatbot by natural language processing for supportive risk management in river dredging projects. Eng. Appl. Artif. Intell.131, 107744. 10.1016/j.engappai.2023.107744

  • CrossRef
  • Google Scholar

• 33

  ChuW.VicidominiM.CaliseF.DuićN.ØstergaardP. A.WangQ.et al (2022). Recent advances in low-carbon and sustainable, efficient technology: strategies and applications. Energies15 (8), 2954. 10.3390/en15082954

  • CrossRef
  • Google Scholar

• 34

  CostantinoF.Di GravioG.NoninoF. (2015). Project selection in project portfolio management: an artificial neural network model based on critical success factors. Int. J. Proj. Manag.33 (8), 1744–1754. 10.1016/j.ijproman.2015.07.003

  • CrossRef
  • Google Scholar

• 35

  DattaS. D.TayehB. A.HakeemI. Y.Abu AishehY. I. (2023). Benefits and barriers of implementing building information modeling techniques for sustainable practices in the construction industry—A comprehensive review. Sustainability15 (16), 12466. 10.3390/su151612466

  • CrossRef
  • Google Scholar

• 36

  DavahliM. R.KarwowskiW.FiokK.WanT.ParsaeiH. R. (2021). Controlling safety of artificial intelligence-based systems in healthcare. Symmetry13 (1), 102. 10.3390/sym13010102

  • CrossRef
  • Google Scholar

• 37

  de AlmeidaM. V. F. (2024). Applying Artificial Intelligence on Project Management Office Improvement (master's Thesis). Portugal: Universidade NOVA de Lisboa.

  • Google Scholar

• 38

  DobrovolskienėN.TamošiūnienėR. (2015). An index to measure sustainability of a business project in the construction industry: lithuanian case. Sustainability8 (1), 14. 10.3390/su8010014

  • CrossRef
  • Google Scholar

• 39

  DuH.NiyatoD.KangJ.XiongZ.ZhangP.CuiS.et al (2024). The age of generative AI and AI-generated everything. Ieee Netw.38 (6), 501–512. 10.1109/mnet.2024.3422241

  • CrossRef
  • Google Scholar

• 40

  DunmoreA.Jang-JaccardJ.SabrinaF.KwakJ. (2023). A comprehensive survey of generative adversarial networks (GANs) in cybersecurity intrusion detection. IEEE Access11, 76071–76094. 10.1109/access.2023.3296707

  • CrossRef
  • Google Scholar

• 41

  EberW. (2020). Potentials of artificial intelligence in construction management. Organ. Technol. and Manag. Constr. Int. J.12 (1), 2053–2063. 10.2478/otmcj-2020-0002

  • CrossRef
  • Google Scholar

• 42

  Edum-FotweF. T.PriceA. D. (2009). A social ontology for appraising sustainability of construction projects and developments. Int. J. Proj. Manag.27 (4), 313–322. 10.1016/j.ijproman.2008.04.003

  • CrossRef
  • Google Scholar

• 43

  EloS.KyngäsH. (2008). The qualitative content analysis process. J. Advanced Nursing62 (1), 107–115. 10.1111/j.1365-2648.2007.04569.x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 44

  ElseknidyM.Al-MhdawiM. K. S.QaziA.OjiakoU.MahammediC.RahimianF. P. (2025). Developing a sustainability-driven risk management framework for green building projects: a literature review. J. Clean. Prod.519, 145891. 10.1016/j.jclepro.2025.145891

  • CrossRef
  • Google Scholar

• 45

  ErfaniA.CuiQ. (2022). Predictive risk modelling for major transportation projects using historical data. Automation Constr.139, 104301. 10.1016/j.autcon.2022.104301

  • CrossRef
  • Google Scholar

• 46

  Farfán ChilicausG. C.Licapa-RedolfoG. S.Arbulú BallesterosM. A.Corrales OtazúC. D.Apaza MirandaS. J.Flores CastilloM. M.et al (2025). Digital transformation and sustainability in post-pandemic supply chains: a global bibliometric analysis of technological evolution and research patterns (2020–2024). Sustainability17 (7), 3009. 10.3390/su17073009

  • CrossRef
  • Google Scholar

• 47

  FaridaN.HandayaniN. U.WibowoM. A. (2019). Developing indicators of green construction of green supply chain management in construction industry: a literature review. In IOP Conf. Ser. Mater. Sci. Eng.598, 1, 012021. 10.1088/1757-899x/598/1/012021

  • CrossRef
  • Google Scholar

• 48

  FerrettiV.SaltelliA.TarantolaS. (2016). Trends in sensitivity analysis practice in the last decade. Sci. Total Environ.568, 666–670. 10.1016/j.scitotenv.2016.06.118

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 49

  FischerD.HaukeB.BragançaL.AndradeJ. B.MateusR. (2016). What Does ‘Sustainable Construction’Mean? an Overview.

  • Google Scholar

• 50

  FridgeirssonT. V.IngasonH. T.JonassonH. I.GunnarsdottirH. (2023). A qualitative study on artificial intelligence and its impact on the project schedule, cost and risk management knowledge areas as presented in PMBOK®. Appl. Sci.13 (19), 11081. 10.3390/app131911081

  • CrossRef
  • Google Scholar

• 51

  GanX.ZuoJ.YeK.SkitmoreM.XiongB. (2015). Why sustainable construction? Why not? An owner's perspective. Habitat International47, 61–68. 10.1016/j.habitatint.2015.01.005

  • CrossRef
  • Google Scholar

• 52

  GaoB.WangY.XieH.HuY.HuY. (2023). Artificial intelligence in advertising: advancements, challenges, and ethical considerations in targeting, personalization, content creation, and ad optimization. Sage Open13 (4), 21582440231210759. 10.1177/21582440231210759

  • CrossRef
  • Google Scholar

• 53

  GheibiO.WeynsD.QuinF. (2021). Applying machine learning in self-adaptive systems: a systematic literature review. ACM Trans. Aut. Adapt. Syst. (TAAS)15 (3), 1–37. 10.1145/3469440

  • CrossRef
  • Google Scholar

• 54

  GiraudL.ZaherA.HernandezS.AkramA. A. (2023). The impacts of artificial intelligence on managerial skills. Int. J. Constr. Manag.32 (3), 566–599. 10.1080/12460125.2022.2069537

  • CrossRef
  • Google Scholar

• 55

  GuestG.BunceA.JohnsonL. (2006). How many interviews are enough? An experiment with data saturation and variability. Field Methods18 (1), 59–82. 10.1177/1525822x05279903

  • CrossRef
  • Google Scholar

• 56

  GuptaM.AkiriC.AryalK.ParkerE.PraharajL. (2023). From ChatGPT to ThreatGPT: impact of generative AI in cybersecurity and privacy. IEEE Access. 10.1109/ACCESS.2023.3300381

  • CrossRef
  • Google Scholar

• 57

  HadidA.ChakrabortyT.BusbyD. (2024). When geoscience meets generative AI and large language models: foundations, trends, and future challenges. Expert Syst.41 (10), e13654. 10.1111/exsy.13654

  • CrossRef
  • Google Scholar

• 58

  HagosD. H.BattleR.RawatD. B. (2024). Recent advances in generative ai and large language models: current status, challenges, and perspectives. IEEE Transactions Artificial Intelligence. 10.1109/TAI.2024.3444742

  • CrossRef
  • Google Scholar

• 59

  HashfiM. I.RaharjoT. (2023). Exploring the challenges and impacts of artificial intelligence implementation in project management: a systematic literature review. Int. J. Adv. Comput. Sci. Appl.14 (9), 366–376. 10.14569/ijacsa.2023.0140940

  • CrossRef
  • Google Scholar

• 60

  HassanM. K.AbdulkarimM. E.IsmaelH. R. (2022). Risk governance: exploring the role of organisational culture. J. Account. and Organ. Change18 (1), 77–99. 10.1108/jaoc-01-2021-0003

  • CrossRef
  • Google Scholar

• 61

  HaukeB.KuhnhenneM.LawsonM.VeljkovićM.SiebersR. (2016). “What does ‘sustainable construction’mean? An overview,” in Sustainable Steel Buildings A Practical Guide for Structures and Envelopes. Blackwell: Wiley.

  • Google Scholar

• 62

  HeersminkR.van den HovenJ.van EckN. J.van den BergJ. (2011). Bibliometric mapping of computer and information ethics. Ethics Information Technology13, 241–249. 10.1007/s10676-011-9273-7

  • CrossRef
  • Google Scholar

• 63

  HolzmannV.LechiaraM. (2022). Artificial intelligence in construction projects: an explorative study of professionals’ expectations. Eur. J. Bus. Manag. Res.7 (3), 151–162. 10.24018/ejbmr.2022.7.3.1432

  • CrossRef
  • Google Scholar

• 64

  HomK. (2024). Circular economy: the 4Rs reduce-reuse-recycle-recover. In IOP Conf. Ser. Earth Environ. Sci.1402, 1, 012042. 10.1088/1755-1315/1402/1/012042

  • CrossRef
  • Google Scholar

• 65

  HossnyH. E.IbrahimA. H.ElnadyA. (2021). Assessment of construction project complexity. Open Civ. Eng. J.15 (1), 414–423. 10.2174/1874149502115010414

  • CrossRef
  • Google Scholar

• 66

  HsiehH. F.ShannonS. E. (2005). Three approaches to qualitative content analysis. Qual. Health Research15 (9), 1277–1288. 10.1177/1049732305276687

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 67

  IluyomadeT. D.OkwanduA. C. (2024). Innovative materials in sustainable construction: a review. Int. J. Sci. and Res. Arch.12 (1), 2435-2442.

  • Google Scholar

• 68

  JallowH.RenukappaS.SureshS.RahimianF. (2023). Artificial intelligence and the UK construction industry–empirical study. Eng. Manag. J.35 (4), 420–433. 10.1080/10429247.2022.2147381

  • CrossRef
  • Google Scholar

• 69

  KasgariA. T. Z.SaadW.MozaffariM.PoorH. V. (2020). Experienced deep reinforcement learning with generative adversarial networks (GANs) for model-free ultra reliable low latency communication. IEEE Trans. Commun.69 (2), 884–899. 10.1109/tcomm.2020.3031930

  • CrossRef
  • Google Scholar

• 70

  KhalidJ.ChuanminM.AltafF.ShafqatM. M.KhanS. K.AshrafM. U. (2024). AI-driven risk management and sustainable decision-making: role of perceived environmental responsibility. Sustainability16 (16), 6799. 10.3390/su16166799

  • CrossRef
  • Google Scholar

• 71

  LachhabM.BélerC.CoudertT. (2018). A risk-based approach applied to system engineering projects: a new learning-based multi-criteria decision support tool based on an ant colony algorithm. Eng. Appl. Artif. Intell.72, 310–326. 10.1016/j.engappai.2018.04.001

  • CrossRef
  • Google Scholar

• 72

  LiJ.ZhangM.LiN.WeynsD.JinZ.TeiK. (2024). Generative AI for self-adaptive systems: state of the art and research roadmap. ACM Trans. Aut. Adapt. Syst.19 (3), 1–60. 10.1145/3686803

  • CrossRef
  • Google Scholar

• 73

  LiangC. J.LeT. H.HamY.ManthaB. R.ChengM. H.LinJ. J. (2024). Ethics of artificial intelligence and robotics in the architecture, engineering, and construction industry. Automation Constr.162, 105369. 10.1016/j.autcon.2024.105369

  • CrossRef
  • Google Scholar

• 74

  LiudmilaC.BalkizY. (2019). “Bridging sustainable construction technologies and heritage: novel approach to the conservation of the built environment,” in Sustainable Construction Technologies (Butterworth-Heinemann), 395–442.

  • Google Scholar

• 75

  McMillanL.VargaL. (2022). A review of the use of artificial intelligence methods in infrastructure systems. Eng. Appl. Artif. Intell.116, 105472. 10.1016/j.engappai.2022.105472

  • CrossRef
  • Google Scholar

• 76

  MoczydłowskaJ.SadkowskaJ.ŻelazkoB.CiocoiuC. N.StawickaE. (2023). Understanding risk culture in the context of a sustainable project: a preliminary study. Sustainability15 (6), 5302. 10.3390/su15065302

  • CrossRef
  • Google Scholar

• 77

  MohamadiS.MujtabaG.LeN.DorettoG.AdjerohD. A. (2023). ChatGPT in the age of generative AI and large language models: a concise survey. arXiv Preprint arXiv:2307.04251. 10.48550/arXiv.2307.04251

  • CrossRef
  • Google Scholar

• 78

  MohamedM. A. H.Al-MhdawiM. K. S.OjiakoU.DacreN.QaziA.RahimianF. (2025a). Generative AI in construction risk management: a bibliometric analysis of the associated benefits and risks. Urbanization, Sustain. Soc.2 (1), 196–228. 10.1108/uss-11-2024-0069

  • CrossRef
  • Google Scholar

• 79

  MohamedM. A. H.Al-MhdawiM. K. S.RahimianF. P.OjiakoU.O’ConnorA.MahammediC. (2025b). “Exploring the risks of integrating generative artificial intelligence into construction risk management: insights from a systematic literature review,” in 6th International Conference on Civil and Building Engineering Informatics: ICCBEI 2025.

  • Google Scholar

• 80

  MohamedM. A. H.Al-MhdawiM. K. S.OjiakoU.QaziA.MahammediC.DacreN. (2026a). A GenAI-driven risk management framework for sustainable development projects. Proj. Leadersh. Soc.7, 100217. 10.1016/j.plas.2026.100217

  • CrossRef
  • Google Scholar

• 81

  MohamedM. A. H.Al-MhdawiM. K. S.Shadlooye ChianehG.OjiakoU. (2026b). Development of a hierarchical fuzzy assessment model for quantifying GenAI risks in sustainable construction projects. Front. Built Environ.12, 1842237. 10.3389/fbuil.2026.1842237

  • CrossRef
  • Google Scholar

• 82

  Mori JuniorR.FranksD. M.AliS. H. (2016). Sustainability certification schemes: evaluating their effectiveness and adaptability. Corp. Gov.16 (3), 579–592. 10.1108/cg-03-2016-0066

  • CrossRef
  • Google Scholar

• 83

  NabawyM.Gouda MohamedA. (2024). Risks assessment in the construction of infrastructure projects using artificial neural networks. Int. J. Constr. Manag.24 (4), 361–373. 10.1080/15623599.2022.2156902

  • CrossRef
  • Google Scholar

• 84

  NaoumS.EgbuC. (2015). Critical review of procurement method research in construction journals. Procedia economics and finance. 21(1), 6–13. 10.1016/s2212-5671(15)00144-6

  • CrossRef
  • Google Scholar

• 85

  NidhisreeC.PaulA.VenunadhA.BhowmickR. S. (2024). “Generative AI under scrutiny: assessing the risks and challenges in diverse domains,” in 2024 IEEE 6th International Conference on Cybernetics, Cognition and Machine Learning Applications (ICCCMLA) (IEEE), 243–248.

  • Google Scholar

• 86

  NishantR.KennedyM.CorbettJ. (2020). Artificial intelligence for sustainability: Challenges, opportunities, and a research agenda. Int. J. Inf. Manag.53, 102104. 10.1016/j.ijinfomgt.2020.102104

  • CrossRef
  • Google Scholar

• 87

  OjiakoU.Al-MhdawiM. K. S.UbaidA. M.ChipuluM.AdedejiT.AlRaeesiE. J. H. (2025). Intentional misrepresentations of project information. J. Leg. Aff. Dispute Resolut. Eng. Constr.17 (4), 03125003. 10.1061/jladah.ladr-1318

  • CrossRef
  • Google Scholar

• 88

  OjiakoU.HiggsM.Al NaqbiK. K. M.Al-MhdawiM. K. S.ChipuluM.WilliamsT.et al (2026). Dark personality traits in project management: a bibliometric analysis and agenda for future research. Proj. Leadersh. Soc.7, 100214. 10.1016/j.plas.2026.100214

  • CrossRef
  • Google Scholar

• 89

  OkiyeS. E.OhakawaT. C.NwokediegwuZ. S. (2023). Framework for integrating passive design strategies in sustainable green residential construction. Int. J. Sci. Res. Civ. Eng.7 (6), 17–44. 10.32628/ijsrce237524

  • CrossRef
  • Google Scholar

• 90

  Osei-KyeiR.ChanA. P. (2018). Stakeholders’ perspectives on the success criteria for public-private partnership projects. Int. J. Strategic Prop. Manag.22 (2), 131–142. 10.3846/ijspm.2018.444

  • CrossRef
  • Google Scholar

• 91

  PanY.ZhangL. (2021). Roles of artificial intelligence in construction engineering and management: a critical review and future trends. Automation Constr.122, 103517. 10.1016/j.autcon.2020.103517

  • CrossRef
  • Google Scholar

• 92

  PannellD. J. (1997). Sensitivity analysis of normative economic models: theoretical framework and practical strategies. Agric. Econ.16 (2), 139-152. 10.1111/j.1574-0862.1997.tb00449.x

  • CrossRef
  • Google Scholar

• 93

  ParameshaM.RaneN.RaneJ. (2024). Enhancing resilience through generative artificial intelligence such as ChatGPT. SSRN 4832533.

  • Google Scholar

• 94

  ParveenR. (2018). Artificial intelligence in the construction industry: legal issues and regulatory challenges. Int. J. Civ. Eng. Technol.9 (13), 957–962.

  • Google Scholar

• 95

  PhamH. T.HanS. (2023). Natural language processing with multitask classification for semantic prediction of risk-handling actions in construction contracts. J. Comput. Civ. Eng.37 (6), 04023027. 10.1061/jccee5.cpeng-5218

  • CrossRef
  • Google Scholar

• 96

  PillaiV. S.MatusK. J. (2020). Towards a responsible integration of artificial intelligence technology in the construction sector. Sci. Public Policy47 (5), 689–704. 10.1093/scipol/scaa073

  • CrossRef
  • Google Scholar

• 97

  RajA.AroraL.GirijaS. S.KapoorS.PradhanD.ShetgaonkarA. (2025). “AI and generative AI transforming disaster Management: a survey of damage assessment and response techniques,” in IEEE 49th Annual Computers, Software, and Applications Conference (COMPSAC) (IEEE). 1834-1840. arXiv Preprint arXiv:2505.08202. 10.1109/COMPSAC65507.2025.00251

  • CrossRef
  • Google Scholar

• 98

  RaneN. (2023). “Integrating building information modelling (BIM) and artificial intelligence (AI) for smart construction schedule, cost, quality, and safety management: challenges and opportunities,” in Cost, Quality, and Safety Management: Challenges and Opportunities.

  • Google Scholar

• 99

  RaneN.ChoudharyS.RaneJ. (2023). Integrating ChatGPT, bard, and leading-edge generative artificial intelligence in building and construction industry: applications, framework, challenges, and future scope. Framew. Challenges, Future Scope. 10.2139/ssrn.4645597

  • CrossRef
  • Google Scholar

• 100

  RegonaM.YigitcanlarT.HonC.TeoM. (2024). Artificial intelligence and sustainable development goals: systematic literature review of the construction industry. Sustain. Cities Soc.108, 105499. 10.1016/j.scs.2024.105499

  • CrossRef
  • Google Scholar

• 101

  RenZ. (2022). “The advance of generative model and variational autoencoder,” in 2022 IEEE Conference on Telecommunications, Optics and Computer Science (TOCS) (IEEE), 268–271.

  • Google Scholar

• 102

  SabiniL.SilviusG. (2023). “Embracing complexity in sustainable project management,” in Research Handbook on Complex Project Organizing (Cheltenham: Edward Elgar Publishing), 312–321.

  • Google Scholar

• 103

  SalihF.El-AdawayI. H. (2024). “Achieving project objectives and improving functions: the benefits of AI and construction technologies,” in Construction Research Congress 2024, 1–11.

  • Google Scholar

• 104

  SaltelliA.RattoM.AndresT.CampolongoF.CariboniJ.GatelliD.et al (2008). Global Sensitivity Analysis: The Primer. John Wiley and Sons.

  • Google Scholar

• 105

  SaundersM. (2009). Research methods for business students. Person Education Limited.

  • Google Scholar

• 106

  SfakianakiE. (2015). Resource-efficient construction: rethinking construction towards sustainability. World J. Sci. Technol. Sustain. Dev.12 (3), 233–242. 10.1108/wjstsd-03-2015-0016

  • CrossRef
  • Google Scholar

• 107

  SheedyE.Canestrari‐SohD. S. (2023). Does executive accountability enhance risk management and risk culture?Account. and Finance63 (4), 4093–4124. 10.1111/acfi.13087

  • CrossRef
  • Google Scholar

• 108

  SirajN. B.FayekA. R. (2019). Risk identification and common risks in construction: literature review and content analysis. J. Construction Engineering Management145 (9), 03119004. 10.1061/(asce)co.1943-7862.0001685

  • CrossRef
  • Google Scholar

• 109

  SongJ.MunyindaM.Adu SarfoP. (2025). Examining the impact of risk management practices on sustainable project performance in the construction industry: the role of stakeholder engagement. Front. Built Environ.11, 1575827. 10.3389/fbuil.2025.1575827

  • CrossRef
  • Google Scholar

• 110

  SuffoM.BromeR. C. (2018). Methodological proposal for the supply chain risk management within the aeronautical sector. Int. J. Sustain. Aviat.4 (3-4), 290–302. (Yao & Soto, 2024) Yao, D., García de Soto, B. Cyber Risk Assessment Framework for the Construction Industry Using Machine Learning Techniques. 10.1504/ijsa.2018.098427

  • CrossRef
  • Google Scholar

• 111

  SummerillC.PollardS. J.SmithJ. A. (2010). The role of organizational culture and leadership in water safety plan implementation for improved risk management. Sci. Total Environment408 (20), 4319–4327. 10.1016/j.scitotenv.2010.06.043

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 112

  Surawy-StepneyN.ProvostF.BhanguS.CaduffC. (2023). Introduction to qualitative research methods: part 2. Perspect. Clin. Res.14 (2), 95–99. 10.4103/picr.picr_37_23

  • CrossRef
  • Google Scholar

• 113

  TaiwoR.BelloI. T.AbdulaiS. F.YussifA. M.SalamiB. A.SakaA.et al (2024). Generative AI in the construction industry: a state-of-the-art analysis. arXiv Preprint arXiv:2402.09939. 10.48550/arXiv.2402.09939

  • CrossRef
  • Google Scholar

• 114

  TetteyD. J.ChrisbenD.VictoriaM. C.ChukwubuikemE. P.AbdullahiA. (2025). Responsible AI Deployment in Sustainable Project Execution: Ensuring Transparency, Carbon Efficiency and Regulatory Alignment.

  • Google Scholar

• 115

  ToochukwuA. C. (2025). Sustainable construction practices: balancing cost efficiency, environmental impact, and stakeholder collaboration. Int. Res. J. Mod. Eng. Technol. Sci.7 (1), 4118–4140. 10.56726/IRJMETS66703

  • CrossRef
  • Google Scholar

• 116

  VaishyaR.VaishA.HaleemA. (2020). What is there in the scoring and rating of journals?J. Clin. Orthop. Trauma11, S684–S685. 10.1016/j.jcot.2020.07.010

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 117

  WaqarA.OthmanI.Hamah SorN.AlshehriA. M.AlmujibahH.AlotaibiB. S.et al (2023). Modelling relation among implementing AI-based drones and sustainable construction project success. Front. Built Environ.9, 1208807. 10.3389/fbuil.2023.1208807

  • CrossRef
  • Google Scholar

• 118

  WenH.AbouRizkS.MohamedY. (2025). Using large language models to identify project risks for sustainable operations. CIB Conf.1 (1), 294. 10.7771/3067-4883.1757

  • CrossRef
  • Google Scholar

• 119

  WhyteJ.StasisA.LindkvistC. (2016). Managing change in the delivery of complex projects: configuration management, asset information and ‘big data. Int. J. Proj. Manag.34 (2), 339–351. 10.1016/j.ijproman.2015.02.006

  • CrossRef
  • Google Scholar

• 120

  WięckowskiJ.SałabunW. (2023). Sensitivity analysis approaches in multi-criteria decision analysis: a systematic review. Appl. Soft Comput., 148, 110915. 10.1016/j.asoc.2023.110915

  • CrossRef
  • Google Scholar

• 121

  WijayasekeraS. C.HussainS. A.PaudelA.PaudelB.SteenJ.SadiqR.et al (2022). Data analytics and artificial intelligence in the complex environment of megaprojects: implications for practitioners and project organizing theory. Proj. Manag. J.53 (5), 485–500. 10.1177/87569728221114002

  • CrossRef
  • Google Scholar

• 122

  WuD.OlsonD. L.DolguiA. (2017). Artificial intelligence in engineering risk analytics. Eng. Appl. Artif. Intell.65, 433–435. 10.1016/j.engappai.2017.09.001

  • CrossRef
  • Google Scholar

• 123

  YaoD.García de SotoB. (2024). Cyber risk assessment framework for the construction industry using machine learning techniques. Buildings14 (6), 1561. 10.3390/buildings14061561

  • CrossRef
  • Google Scholar

• 124

  YaseenZ. M.AliZ. H.SalihS. Q.Al-AnsariN. (2020). Prediction of risk delay in construction projects using a hybrid artificial intelligence model. Sustainability12 (4), 1514. 10.3390/su12041514

  • CrossRef
  • Google Scholar

• 125

  YazdiM.ZareiE.AdumeneS.BeheshtiA. (2024). Navigating the power of artificial intelligence in risk management: a comparative analysis. Safety10 (2), 42. 10.3390/safety10020042

  • CrossRef
  • Google Scholar

• 126

  YigitcanlarT.RegonaM.XiaB.LiR. Y. M. (2022). Opportunities and adoption challenges of AI in the construction industry: a PRISMA review. J. Open Innovation Technol. Mark. Complex.8 (1), 45. 10.3390/joitmc8010045

  • CrossRef
  • Google Scholar

• 127

  ZabihiH.HabibF. (2012). Sustainability in building and construction: revising definitions and concepts. Int. J. Emerg. Sci.2 (4), 570.

  • Google Scholar

• 128

  ZhaoX.ZhangM.FanX.SunZ.LiM.LiW.HuangL. (2023). Extended reality for safe and effective construction management: State-of-the-art, challenges, and future directions. Buildings, 13 (1), 155. 10.3390/buildings13010155

  • CrossRef
  • Google Scholar

• 129

  ZhaoH.WangX.SunJ.WangY.ChenZ.WangJ.et al (2024). Artificial intelligence powered real-time quality monitoring for additive manufacturing in construction. Constr. Build. Mater.429, 135894. 10.1016/j.conbuildmat.2024.135894

  • CrossRef
  • Google Scholar

• 130

  ZhouH.TangS.HuangW.ZhaoX. (2023). Generating risk response measures for subway construction by fusion of knowledge and deep learning. Automation Constr.152, 104951. 10.1016/j.autcon.2023.104951

  • CrossRef
  • Google Scholar