AI-Driven Risk Management and Sustainable Decision-Making: Role of Perceived Environmental Responsibility

Abstract

Risk management driven by AI has the potential to play an important role in sustainable decision-making by offering predictive insights and scenario modeling capabilities. This proactive approach empowers top management to align informed decisions in accordance with sustainability goals while optimizing resource allocation and mitigating risks. While existing research examined the benefits of AI risk management, this study addresses the underexplored question of how AI risk management impacts sustainable decision-making, particularly when considering the mediating role of perceived environmental responsibility. A structural equation modeling (SEM) technique was employed on a dataset comprising 428 senior managers from the Pakistani construction sector. The results revealed that AI-driven risk management is highly relevant to sustainable decision-making. Further, perceived environmental responsibility was found to have a partial mediating effect. These results hold a significant promise for organizations seeking to integrate AI for environmental sustainability goals. The findings of the study enhance the understanding of how AI-driven risk management is a driving mechanism empowering decision-makers to make more sustained decisions in the construction sector.

1. Introduction

Sustainable development is heavily concerned and meeting the prerequisites of humanity while instantaneously progressing social and technological avenues. During the contemporary digital epoch, sustainability and artificial intelligence (AI) are inextricably connected by leveraging different applications and initiatives in order to address environmental and socio-economic challenges [1,2]. AI tools are incredibly imperative due to their indispensable role in advancing sustainability efforts by empowering and fostering effective resource allocation, strengthening environmental surveillance and assessment, and enabling data-driven decision-making procedures. Furthermore, AI is equally contributing to the risk management practices of organizations. Senior leadership cannot avoid risk management, as it assists in anticipating and mitigating potential negative outcomes that could impede the organizational objectives [3,4]. Instead, it enables managers to assess and address internal and external uncertainties. In fact, organizations can resist times of unpredictability by proactively managing risks.

The existing literature on decision-making identified different enablers that are raising organizations’ concerns for sustainable decision-making practices [1,5]. In the context of AI, as suggested by Yaseen (2021), leveraging AI for proactive risk identification, organizations can analyze vast datasets to pinpoint potential risks across various operational facets, including environmental impact, supply chain resilience, and regulatory compliance [6]. Moreover, AI-driven risk assessments offer a comprehensive approach, considering environmental, social, and economic factors, facilitating holistic evaluations and the prioritization of actions to minimize adverse effects on sustainability objectives [3]. AI-enabled data-driven decision-making assists businesses to act on insights from past data and prediction models, ensuring that actions are in accordance with sustainability objectives and risk-free. In addition to maximizing sustainability impact while minimizing waste and optimizing overall performance, AI improves resource allocation by spotting inefficiencies and redirecting resources accordingly. By keeping checks on data in real-time, organizations are able to adjust to unfamiliar circumstances rapidly, availing full advantage of sustainability potential while mitigating risks [7,8]. Therefore, risk management enabled by AI not only aids sustainability initiatives but also facilitates organizations’ survival and thriving in a constantly changing world.

Pakistan is experiencing rapid urbanization and infrastructure development, resulting in a huge demand for construction projects [9,10]. The construction industry in Pakistan holds an important position in facilitating social as well as economic growth. From an economic point of view, it plays an enormous part in the nation’s Gross Domestic Product (GDP) by creating job prospects, driving investment, and spurring the country’s economic expansion [7,11]. Additionally, both local and international investors contribute to the industry, bringing in capital that fuels economic growth. Through building essential infrastructure like roads, bridges, schools, and hospitals, the construction sector significantly improves the quality of life for people across Pakistan, making everyday living better and more comfortable [11]. Nevertheless, this swift change also intensifies environmental concerns, which are very urgent to take into consideration. Unfortunately, policymakers often lack the necessary empirical knowledge to address these issues effectively. According to Shah and Longsheng (2020), this sector is heavily causing environmental issues like pollution, the depletion of resources, and the destruction of habitats [12]. However, despite the proliferation of AI technologies in various domains, the specific mechanisms through which AI influences decision-making processes, particularly in the context of risk management and sustainability, remain underexplored. This is more particular in the context of the construction industry in developing countries. There is a need to figure out how using AI in risk management can help construction companies identify and address environmental risks efficiently by promoting sustainable decision-making practices [13,14]. This study, therefore, has the primary objective to examine the impact of AI-driven risk management on sustainable decision-making in the Pakistani construction industry.

Making informed choices that prioritize environmental concerns and sustainability is also based on intrinsic motivation [15]. In a recent study, Faraz et al. (2021) argued that if individuals and organizations perceive it as their ethical obligation to protect and conserve the natural environment, they will show greater mindfulness and awareness [16]. Companies with a strong sense of environmental responsibility may be more proactive in identifying and addressing environmental risks and can stimulate innovation in AI-driven risk management practices. By prioritizing environmental responsibility, organizations can harness AI technologies to gain deeper insights regarding risks, fostering sustainable decision-making not only within the construction sector but also across other industries [17]. This study aims to examine the mediating role of perceived environmental responsibility on the relationship between AI-driven risk management and sustainable decision-making.

Addressing these problem areas is crucial not only for advancing academic knowledge in the field but also for providing actionable insights that can guide the development of a resilient, sustainable, and technologically advanced construction sector. This study seeks to bridge these critical gaps and contribute to the ongoing discourse on the role of AI in shaping the future. Specifically, this study aims to answer the following questions: (a) To what extent does the integration of AI-driven risk management systems influence sustainable decision-making? (b) How does the perceived environmental responsibility of senior managers mediate the relationship between AI-driven risk management and sustainable decision-making? This study holds significant potential to gain insight into the correlation between AI-driven risk management and sustainable decision-making, particularly within the context of the construction industry in Pakistan. This study contributes to the literature on AI-driven risk management and sustainability while offering pragmatic suggestions for improved decision-making processes in the construction sector and beyond. By bridging critical gaps in understanding and offering evidence-based recommendations, this study can pave the way for a more prosperous, environmentally friendly, and technologically sophisticated future.

The rest of the paper is designed as follows: the literature review section offers a thorough understanding of the constructs, such as sustainable decision-making, risk management, AI’s role in risk management, and how traditional approaches to risk management are distinct from AI-driven risk management. Hypothesis development and the framework of the study are also presented in the literature review section. The methodology section portrays the target population and sampling method. The results section offers the measurement model and structural model analysis in the form of tables and figures. Finally, the discussion section elaborates on the results of the study and presents the theoretical and practical implications of the study. The study ends with a conclusion that draws a complete picture of the research.

2. Literature Review

2.1. Sustainable Decision-Making

The evolution of sustainable decision-making in the literature has emerged in different stages following a journey through various phases, reflecting transits in our awareness and understanding regarding environmental awareness and socio-economic values. In the 1970s–1990s era, a phase can be termed Emergence and Awareness focused on increasing consciousness about environmental concerns [18,19] like resource depletion [20] and pollution [21,22,23] (see Table 1). Among the early literature, the book The Limits to Growth by Meadows et al. (1972) and the report “Our Common Future” by Brundtland and Comum (1987) introduced the idea of sustainable development. Subsequently, another phase that is more towards compliance and regulation in the period of the 1990s to 2000s emphasized the rules and regulations for environmental safety and protection (Logsdon & Wartick, 1995) [24,25,26]. It was also when ideas like corporate social responsibility (CSR) and corporate environmental responsibility (CER) gained fame and became a part of literary discussions [27,28]. Organizations started to acknowledge and take responsibility for their environmental and societal impact.

In the next phase, comprising the 2000s to the 2010s, named integration and mainstreaming, sustainability became more than just a buzzword; rather it was considered a core part of business operations and an integral part of decision-making processes [29,30]. Researchers attempted to identify the potential financial benefits linked with green environmental practices and being socially responsible [31]. During this period, concepts like green innovation and sustainable supply chains received scholars’ and practitioners’ attention [32,33]. From 2010 till the present, the concept of sustainable decision-making has greatly transformed. In this innovation and transformation phase, the major focus was shifted toward advanced technologies like circular economy and renewable energy driving transformative changes in sustainability practices [34,35,36,37]. These technologies are pushing us toward a more sustainable future. We are witnessing a lot of talk about sustainable business models and creating smart cities with an enhanced understanding of how social issues, the environment, and the economy are connected. As the literature progresses, both researchers and practitioners are actively looking for innovative solutions to address pressing sustainability challenges in our ever-changing world.

Table 1. Brief overview of sustainable decision-making evolution.

Phase	Timeframe	Characteristics	References
Emergence and Awareness	1970 to 1990	Rise of environmental awareness and consciousness and introduction of sustainable development concept	[18,19,20]
Compliance and Regulation	1990 to 2000	Regulatory emphasis: Prioritizing government policies, international agreements, and corporate environmental compliance.	[29,30]
Integration and Mainstreaming	2000 to 2010	Mainstreaming sustainability in business practices, focus on sustainable supply chains, green innovation, and corporate sustainability reporting	[30,34]
Innovation and Transformation	2010 to present	Emphasis on innovative sustainability solutions and disruptive technologies and business models, highlighting interconnected social, environmental, and economic systems	[35,37]

Table 1. Brief overview of sustainable decision-making evolution.

Phase	Timeframe	Characteristics	References
Emergence and Awareness	1970 to 1990	Rise of environmental awareness and consciousness and introduction of sustainable development concept	[18,19,20]
Compliance and Regulation	1990 to 2000	Regulatory emphasis: Prioritizing government policies, international agreements, and corporate environmental compliance.	[29,30]
Integration and Mainstreaming	2000 to 2010	Mainstreaming sustainability in business practices, focus on sustainable supply chains, green innovation, and corporate sustainability reporting	[30,34]
Innovation and Transformation	2010 to present	Emphasis on innovative sustainability solutions and disruptive technologies and business models, highlighting interconnected social, environmental, and economic systems	[35,37]

2.2. AI’s Contribution towards Sustainable Decisions

Artificial intelligence (AI) is considered a promising technology and embraced as a game-changer in the era of sustainability. The recent literature suggested that AI enhances the decision-making process in organizations due to its potential to analyze vast datasets, make predictions, and identify patterns [38,39,40]. Researchers emphasized that AI-driven management assists and optimizes resource allocation [41], lessening bad environmental impacts (Van Wynsberghe, 2021) and improving decision-making processes toward sustainability goals (Rodgers et al., 2023) [39,42]. Additionally, AI tools such as natural language processing and machine learning have been leveraged to solve complex sustainability issues ranging from climate change mitigation to social equity.

Using cutting-edge data analytics, AI enables the dissection of huge amounts of information related to environmental, social, and economic factors and empowers top management to uncover complicated patterns, discern trends, and pinpoint potential future avenues for further improvements [43,44]. This, in turn, culminates in the leaders’ ability to make more data-driven and strategic decisions. However, it may not be reachable without efficient risk management, as risk management is crucial in recognizing, assessing, and mitigating potential threats [38]. By systematically addressing risks, organizations can predict problems and take preemptive actions in order to minimize the negative impact. Grasping and handling risks allows leaders to make informed choices, contemplating potential outcomes and uncertainties while making any decision.

2.3. Risk Management—Traditional vs. AI-Driven

Traditional risk management techniques and AI-driven risk management approaches significantly differ in their methods of data handling processes, analysis, and decision-making. In the traditional approach, risk management heavily depends on manual efforts to gather and scrutinize data from records and rely on insights from experts [45]. In contrast, AI-driven risk management exploits machine learning and sophisticated algorithm methods to effectively examine and evaluate immense volumes of data [46]. AI systems can decipher complex data patterns, assess correlations, and forecast real-time potential hazards empowering organizations to react promptly and efficiently.

Traditional approaches may struggle to scale and adapt to changing risk landscapes [45,47] whereas AI-driven systems are highly scalable and adaptable. Traditional risk management relies on human judgments and familiarities, low initial investment, adaptability, and subjectivity; however, it may have limitations in scalability, response time, and predictive analysis (see

Table 2. Comparative analysis—Traditional vs. AI-driven risk management.

Aspect	Traditional Risk Management	AI-Driven Risk Management
Familiarity	✓	✗
Human Judgment	✓	✓
Low Initial Investment	✓	✗
Adaptability	✓	✗
Subjectivity	✓	✗
Limited Scalability	✓	✓
Slow Response Time	✓	✗
Difficulty in Predictive Analysis	✓	✗
Advanced Analytics	✗	✓
Predictive Capabilities	✗	✓
Automation	✗	✓
Real-time Insights	✗	✓
Complexity	✓	✗
Data Privacy Concerns	✓	✓
Potential for Bias	✓	✓
Overreliance on Technology	✗	✓

Source: self-constructed from [4,43,45,49,50,51,52].

). On the other hand, AI-driven risk management offers advanced analytics, predictive capabilities, automation, real-time insights, and a reduced potential for bias, but it may require greater technical expertise and raise concerns about data privacy and overreliance on technology [4,48]. Overall, AI-driven methods prioritize advanced analytics and automation, while traditional approaches are based on human judgment and familiarity. However, each of these approaches offers distinct advantages and complications in managing risks.

2.4. AI Techniques for Risk Management Processes

2.4.1. Cost Modeling Methods

Cost modeling methods are compassed through a three-stage model comprising input, process, and output with a key aim of encompassing cost data, evaluation data, productivity, and safety data [53,54]. To mitigate the volatility in project costs, a regularization neural network was formulated particularly for cost estimation purposes. An illustration of this is to forecast the costs associated with a reinforced concrete structural unit that is engineered to offer more stable and accurate cost estimation. Moreover, the utilization of artificial neural networks (ANNs) was carried out to monitor project productivity and safety [55]. A three-layered network containing a fuzzy output structure appeared to be most effective considering the subjective characteristics of a significant portion of the data. In order to improve the accuracy of scheduling analyses and risk identification, extensive AI databases have been established. These databases enable project managers to input new project tasks and instantaneously acquire feedback regarding predicted actual durations [56]. The verification process involved contrasting the AI output with real project data, demonstrating its validity, and offering invaluable information about potential project influencers for the intent of risk mitigation strategy. This method brings substantial time efficiency when compared to conventional spreadsheet approaches and advances the managerial ability for sustainable decision-making.

2.4.2. Time Modeling Methods

AI approaches for risk management processes contain a range of time modeling methods that attempt to predict the timing and duration of prospective risks accurately. These techniques deploy AI algorithms to evaluate past data, determine trends, and predict the timing of upcoming occurrences [57]. A time series analysis is a commonly employed methodology that entails the examination of sequential datasets in order to detect patterns, seasonal variations, and anomalies. Time modeling in risk management often involves the utilization of techniques such as autoregressive integrated moving averages (ARIMA) and seasonal decomposition of time series (STL) [58,59,60]. Furthermore, machine learning methods, such as recurrent neural networks (RNNs) and long short-term memory (LSTM) networks, possess the capability to effectively capture temporal relationships within datasets, enabling them to generate predictions pertaining to forthcoming time points [61]. Top management can improve their capacities to anticipate and mitigate risks that occur over time by integrating time modeling methodologies into their risk management procedures, resulting in improved strategic planning and decision-making.

2.4.3. Risk Modeling Methods

The application of artificial intelligence (AI) technology has had a profound impact on risk management practices, namely in the context of risk modeling procedures. These techniques utilize sophisticated algorithms and machine learning methods to analyze massive amounts of data, identify complex patterns, and forecast possible dangers with unprecedented accuracy in comparison to conventional methods [62]. Prominent risk modeling techniques in the field of artificial intelligence encompass machine learning algorithms such as decision trees and neural networks, natural language processing for the analysis of textual data, and fuzzy logic for managing uncertainty [62,63]. The integration of AI methodologies into risk management practices facilitates enhanced understanding, quantification, and the mitigation of risks in various industries, including finance, insurance, cybersecurity, and project management. In turn, this integration fosters more informed decision-making processes and yields improved outcomes in risk management.

2.4.4. Hybrid Models

Hybrid models of AI-driven risk management embrace a wide range of novel approaches intended for addressing the complicated processes and ambiguities associated with risk assessment and mitigation [64]. The integration of various AI methods, such as fuzzy logic, neural networks, and genetic algorithms, appears in hybrid models like Fuzzy Hybrid Methods (FHMs) [49], Fuzzy-Analytical Network Processing (FANP) [65], and Fuzzy Monte Carlo Simulation (FMCS) [66]. Hybrid models deliver robust frameworks for comprehensive risk analysis and decision assistance by integrating several methodologies. These models proficiently address interdependencies among risk aspects and effectively tackle uncertainties. These models bring intriguing possibilities for the advancement of risk management methods in various industries, including risk evaluation and sustainable decision-making.

2.5. Theoretical Foundation

The innovation diffusion theory (IDT) describes innovation adoption and spread through social systems over the long run. First coined by Everett Rogers in 1962, the IDT proposes that the adoption of innovations has a general chart consisting of different stages influenced by a multitude of factors [67]. According to the IDT, the adoption process can be categorized into five stages, such as knowledge, persuasion, decision, implementation, and confirmation [68,69]. People become conscious of the innovation and are flooded with the idea of its probability. With regard to the process of decision-making, it entails that the use of new technologies or practices is not a random thing; it is adopted following a certain pattern. The first step of decision-makers is to gather information regarding a new idea, then analyze its potential benefits, make a decision to use it, implement the innovation, and finally confirm its usefulness [70]. Understanding these stages is a crucial component since it assists decision-makers to be prepared to foresee such challenges, and they can plan in advance.

The social cognitive theory (SCT), theorized by Albert Bandura in 1986, is based on the premise that the cognitive processes of an individual affect the way he/she behaves in a social setting [71]. The SCT contends that people learn by being shown, imitating, and emulating the actions of others. This theory emphasizes the role of self-efficacy and reciprocal determinism [72,73]. Self-efficacy means the level of belief an individual has in the ability to successfully carry out something necessary to accomplish the targeted results. On the other hand, reciprocal determinism refers to the dynamic relationship among personal factors, behavior, and the environment in shaping human behavior. Through observation, individuals develop perceptions of what constitutes responsible environmental behavior and may be motivated to emulate these behaviors themselves [74].

Synthesizing the information from the IDT and SCT, the study can give a vivid notion of how senior managers in the construction industry view and use AI-powered risk management systems to aid them in making sustainable decisions (see Figure 1). Likewise, the SCT clarifies the ways that the perceived environmental responsibility influences the formation of the AI-driven risk management and sustainable decision-making relationship.

2.6. Hypothesis Development

2.6.1. AI-Driven Risk Management and Sustainable Decision-Making

The integration of AI into risk management methods has been considered a transformative way with key implications for sustainable decision-making. By incorporating the analytical power and predictive capability of AI, organizations can not only identify and assess risks more effectively but also make informed decisions that are in line with sustainability objectives. According to Adam et al. (2022), organizations anticipate and mitigate risks with greater efficiency and accuracy to make informed decisions [75]. Similarly, Killen et al. (2020) concluded that by assessing vast datasets and evaluating patterns, AI predicts potential risks related to resource depletion, environmental impact, regulatory compliance, and supply chain disruptions [38]. Prior studies emphasized that AI-driven risk management models can predict environmental risks in construction projects, allowing stakeholders to implement proactive measures to minimize negative environmental outcomes [30,39,76]. Given this, this study formulated the following hypothesis:

H1. 

AI-driven risk management is positively associated with sustainable decision-making.

2.6.2. Perceived Environmental Responsibility

Environmental responsibility is regarded as the moral duty that falls upon every person, company, and community to minimize their harmful footprints on this planet [77]. As van der Werff et al. (2021) emphasized, environmental responsibility is an intrinsic motivation to save precious natural resources and encourage deep-seated care for our environment [78]. Communities previously divided find common ground through shared green initiatives that foster unity at grassroot levels through global movements echoing calls to action across continents whispering winds of change. Therefore, protecting the environment is considered a shared responsibility.

Business operations are susceptible to environmental risks ranging from pollution, habitat impairment, and regulatory non-compliance, which can be better spotted with the use of AI-based risk management systems [79]. These systems help organizations comply with their environmental chores by incorporating environmental data with predictive analytics to identify risks and suggest methods to mitigate them [80]. Environmental responsibility aided these systems, which find inefficiencies and suggest ways to fix them. Having increased concern, organizations can make decisions that promote long-term sustainability and prevent harm to the environment by aligning risk management processes with concepts of environmental responsibility. According to Zheng et al. (2020), making decisions that involve economic and social aspects is an integral part of being environmentally responsible [77]. Decision-makers acquire data-driven insights into the environmental effects of different options by employing AI-based risk management. This assists in prioritizing goals related to sustainability while making decisions. Organizations can minimize environmental harm and promote sustainability in the long run by making decisions that are in line with environmental responsibility principles. This discussion led us to form the following hypothesis:

H2. 

AI-driven risk management is positively associated with perceived environmental responsibility.

H3. 

Perceived environmental responsibility is positively associated with sustainable decision-making.

H4. 

Perceived environmental responsibility mediates the relationship between AI-driven risk management and sustainable decision-making.

3. Methodology

The research methodology employed in this study aimed to systematically investigate the impact of AI-driven risk management on sustainable decision-making in the construction sector of Pakistan. This was an exploratory study with a quantitative approach utilizing a structured questionnaire survey for data collection from senior management professionals actively involved in decision-making processes within the targeted industry. A cross-sectional survey technique was employed to test the respondents’ opinions at a specific point in time.

3.1. Sampling Strategy

A non-probability convenience sampling technique was employed to select participants for the questionnaire survey. The target population for the current study was the Pakistani construction sector. Choosing this sector as a target research population is of significant importance, as this sector is heavily contributing to the economic growth of a country as well as playing a role in socio-cultural values [7,11]. In the context of Pakistan, the construction sector contributes 2.3% to 2.85% to GDP [81]. It plays a vital role in economic development and job creation as well as infrastructure development. However, there are several challenges faced by the Pakistani construction sector in terms of sustainable operations and environmentally friendly initiatives [11]. We, thus, directed this study to the construction sector to explore the factors contributing to sustainable decision-making. For this sake, we consulted the Pakistan Engineering Council (PEC) and randomly selected 680 companies registered with the PEC.

Data collection was performed through a survey questionnaire derived from the commonly employed scales used in prior research. The survey instrument was created in the English language. The respondents demonstrated proficiency in English due to its status as the official language of employment and the medium of instruction in high schools and universities in Pakistan [11]. The senior management, including project managers, were approached and requested to participate in the study. There are different rules of thumb when choosing a sample size to represent the target population. According to the 10-times sampling criterion laid out by Maxwell (2000), the maximum number of formative indicators that can be employed in the SEM approach is multiplied by 10 [82]. According to these guidelines, 150 participants (10 × 15) will be needed for the study. Lin et al. (2020) emphasized that a minimum of 100 samples are required for partial least square structural equation modeling (PLS-SEM) [83]. From the 680 distributed questionnaires, the final useable sample size comprised 428 respondents, indicating a response rate of 63%.

Table 3. Demographic Information (N = 428).

Education
	Diploma	Undergraduate	Master	PhD	Total
Female	12	69	49	2	132
Male	33	115	142	6	296
Age
	20–29 years	30–39 years	40–50 years	>50 years	Total
Female	37	51	30	14	132
Male	42	139	88	27	296
Experience
	<2 years	3–7 years	7–10 years	>10 years	Total
Female	9	71	33	19	132
Male	27	119	92	58	296

presents the demographic information of the respondents.

3.2. Measures

The designed questionnaire had three sections. In the first section, personal information including gender and age was acquired. In the second section, professional information such as education and job experience were asked. The third section contained the 14 questions for three constructs (see Appendix A). For the independent variable, a 6-item scale developed by Wong et al. (2022) was used to analyze AI-driven risk management (AIRM) with an alpha reliability of 0.86 [84]. The dependent variable, sustainable decision-making was examined through a 3-item scale adopted from Ben Amara and Chen (2021) [85]. The alpha value for this scale is 0.75. Finally, perceived environmental responsibility (PER), the mediating variable, was measured with a 5-item scale developed by Zheng et al. (2020) with an alpha value of 0.73 [77]. The research instrument utilized a five-point Likert scale to gauge responses, where participants could choose from the following options: 1—Strongly Disagree, 2—Disagree, 3—Neutral, 4—Agree, and 5—Strongly Agree. Employing this scale facilitated the collection of quantitative data efficiently and conveniently, allowing for a structured approach to measuring participants’ perspectives.

3.3. Common Method Variance Measures

We chose structural equation modeling (SEM), specifically the partial least squares (PLS) approach, to validate our measurements and test our complex model. This method is appropriate for incorporating latent variables, crucial for correlational research [86]. Since we gathered both independent and dependent variables from the same respondents, we needed to ensure that common method bias did not skew our results.

To address this concern, we used four different methods. First, we conducted Harman’s one-factor test involving a principal component factor analysis [87]. This test did not show substantial common method bias as indicated by the variance explained. Second, we applied a partial correlation method [88,89], which also did not reveal any significant changes in variance explained, further suggesting no substantial bias. Third, we employed Lindell and Whitney’s method, using a theoretically unrelated construct (marker variable) to adjust correlations among our main constructs [90]. This test, using e-business use as the marker variable, showed no significant correlation, indicating minimal common method bias. Finally, our correlation matrix did not display any highly correlated variables, which is typically indicative of common method bias [91]. In summary, these tests collectively indicated that common method bias is not a significant concern in our study.

4. Results

4.1. Measurement Model (Outer Model)

The reliability and validity of the measurement model should be verified before examining the structural model. For this, factor loading (FL) has been assessed to check the strength of the relationship between indicators and the latent construct it is devised to measure. While alpha (α) examines the internal consistency (reliability) of a set of questions employed to assess a single latent construct. On the other hand, the average variance extracted (AVE) reflects the amount of variance in the indicators explicated by the latent construct on average. Finally, composite reliability (CR) measures the internal consistency of a set of indicators for a latent construct. The fulfillment of all three requirements is required to establish the convergent validity of the scales [92,93]: (1) all indicator factor loadings (FL) should surpass a threshold of 0.50; composite reliabilities (CR) should exceed 0.8; and the average variance extracted (AVE) for each construct should be beyond 0.5. The alpha value (α) should be higher than 0.70, as suggested by Hair et al. (2013) [93]. As shown in

Table 4. Construct validity and reliability.

Construct	FL	α	CR	AVE
AI-driven risk management (AIRM)		0.865	0.903	0.652
AIRM—1 Our firm promotes AI tools to simulate different project scenarios and assess their associated risks.	0.735
AIRM—2 Our firm executed AI-powered systems to monitor project progress and identify potential safety hazards.	0.858
AIRM—3 Our firm use AI-powered analytics to predict and mitigate cost overruns.	0.853
AIRM—4 Our firm use AI risk management tools project schedules and resource allocation.	0.751
AIRM—5 Our firm allocates budget for implementing AI-driven risk management solutions.	0.833
Perceived environmental responsibility (PEM)		0.742	0.837	0.562
PEM—1 Environmental safeguard begins with me.	0.758
PEM—2 I think I should have wider obligation for safeguarding the environment.	0.759
PEM–4 I am able to take responsibility for environmental conservation in construction projects.	0.767
PEM—5 Environmental conservation is my obligation.	0.714
Sustainable decision-making (SDM)		0.753	0.810	0.587
SDM—1 Managers or entrepreneurs in our firm recurrently involve employees or workers in critical decisions to adopt eco-innovation practices	0.786
SDM—2 Our firm policies are considerably affected by the view of employees about eco-innovation	0.730
SDM—3 Employees or workers realize that they are involved in crucial enterprise’s decisions to adopt eco-innovation practices	0.781

Notes: FL = Factor loading, α = Cronbach Alpha, AVE = Average variance extracted, CR = Composite reliability.

and Figure 2, all three requirements have been satisfied, confirming the presence of convergent validity.

To ensure the discriminant validity, we employed two methods: Fornell and Larcker’s criterion and the Heterotrait–Monotrait (HTMT) method. According to Fornell and Larcker, a latent variable’s square root of the average variance extracted (AVE) should be higher than its correlations with other latent variables [92]. Our results confirm this criterion, indicating the presence of discriminant validity in the model (see

Table 5. Fornell–Larcker Criterion.

Constructs	AIRM	PER	SDM
AI-driven Risk Management	0.808
Perceived Environmental Responsibility	0.608	0.750
Sustainable Decision-Making	0.590	0.680	0.766

). Additionally, we utilized the HTMT method, which has become popular recently [94]. First, we looked at whether HTMT values exceeded certain thresholds. Some scholars suggest a threshold of 0.85, while others propose 0.90, especially when correlations are close to one [95]. In this study, all HTMT values were below 0.85, indicating adequate discriminant validity (see

Table 6. Heterotrait–Monotrait Ratio (HTMT).

Constructs	AIRM	PER	SDM
AI-driven Risk Management
Perceived Environmental Responsibility	0.756
Sustainable Decision-Making	0.778	0.840

).

4.2. Structural Model Assessment

The study employs a structural equation modeling (SEM) analysis to investigate the direct effects between AI-driven risk management (AIRM), perceived environmental responsibility (PER), and sustainable decision-making (SDM). Consistent with Chin (2009), we run bootstrapping with 5000 subsamples to ensure robustness and reliability in the analysis [96]. We examined the T-value that assesses the significance of the relationship between two variables (latent or observed). A high T-value (positive or negative) along with a low p-value indicates a statistically significant relationship. On the other hand, the p-value represents the probability of observing the obtained results (or more extreme results) if there were truly no relationship between the variables. First, the relationship between AIRM and PER (H1) shows a strong positive association with a beta value of 0.608 (see

Table 7. Path Modeling.

	Relationship	Beta Value	Std. Error	T-Value	p Value	Decision
H1	AIRM->PER	0.608	0.041	15.005	0.000	Supported
H2	AIRM->SDM	0.280	0.049	5.749	0.000	Supported
H3	PER->SDM	0.510	0.048	10.634	0.000	Supported

Note: AIRM = AI-driven risk management; PER = Perceived Environmental Responsibility; SDM = Sustainable decision-making.

). This means that as AI-driven risk management increases, so does perceived environmental responsibility. This relationship is statistically significant (beta = 0.608, p = 0.000); therefore, H1 was found to be supported (see Figure 3). The results support the idea that AI-driven risk management positively influences perceived environmental responsibility.

Moving on to H2, which examines the relationship between AIRM and SDM, we find a positive relationship with a beta value of 0.280 and p-value of 0.000. This confirms that H2 was found to be supported and statistically significant, reinforcing the idea that AI-driven risk management has a tangible impact on sustainable decision-making. H3 focuses on the relationship between PER and SDM. The analysis reveals a strong positive relationship with a beta value of 0.510 and a p-value of 0.000. This indicates that higher levels of perceived environmental responsibility are associated with increased sustainable decision-making and confirms that H3 was also found supported in this study.

Examining the quality of a model and checking fit indices is important to ensure the reliability and validity of research findings. These indices offer quantitative measures of how well the model aligns with the observed data and how accurately it predicts outcomes [97]. In general, three components of quality fit indices are the coefficient of determination (R2), predictive relevance (Q2), and standardized root mean square residual (SRMR). As shown in

Table 8. The quality of the model and fit indices.

Variables	R2	Q2	SRMR
SDM	0.51	0.33	0.064
PER	0.37	0.14

Note: SDM = Sustainable decision-making; PER = Perceived Environmental Responsibility.

, it is observed that the R2 value for SDM is 0.51, which indicates that approximately 51% of the variance in sustainable decision-making can be explained by the predictor variables included in the model. On the other hand, the R2 value for PER is 0.37, suggesting that around 37% of the variance in perceived environmental responsibility is accounted for by the predictors in the model. Q2 value indicates how well the model predicts new data through cross-validation methods [98]. For SDM, the Q2 value is 0.33, implying that the model has moderate predictive relevance for sustainable decision-making. Finally, SRMR measures the discrepancy between the observed correlations and the model-implied correlations [99]. Lower values indicate a better fit. The SRMR value is 0.064 for SDM, suggesting a reasonably good fit between the observed data and the model’s predictions for sustainable decision-making. In summary, the model appears to have a moderate-to-good fit for sustainable decision-making, as evidenced by the R2 and SRMR values.

We analyzed the mediation effect of PER on the relationship between AIRM and SDM. As shown in

Table 9. Specific Indirect Effect.

	Relationship	Beta value	Std. Error	T-Value	p-Value	Confidence Interval (BC)		Decision
						5% LL	95% UL
H4	AIRM->PER->SDM	0.310	0.035	8.878	0.000	0.261	0.375	Supported

Note: AIRM = AI-driven risk management; PER = perceived environmental responsibility; SDM = sustainable decision-making.

, AIRM positively influences SDM through PER (β = 0.310, p < 0.001) and confirms the PER with a mediating effect. Therefore, hypothesis H4 is found supported. However, this was a partial mediation that occurs when both indirect and direct effects are significant [100].

5. Discussion

Risk management holds immense value for organizations in achieving long-term goals. This is more particular in the construction industry, where making accurate decisions is often challenging due to the uncertainty and complexity inherent in the projects. Recently, AI-driven risk management has grown to be a key component, assisting well-informed decision-making and the allocation of resources [101]. By utilizing AI technologies, organizations can improve their capacity to evaluate and reduce risks. Integrating AI-driven risk management into decision-making processes serves as a milestone in promoting resilient and sustainable development in the construction industry. This study examined how sustainable decision-making can be achieved through AI-driven risk management. In addition, the study examined the impact of AI-driven risk management on sustainable decision-making through perceived environmental responsibility.

The results revealed that AI-driven risk management is positively associated with sustainable decision-making. This is in line with some previous studies [102,103], where researchers found that AI can play a positive role in improving the organizational decision-making process. For instance, predictive analytics and machine learning algorithms could be used in the forecasting of environmental risks, in the optimization of resource allocation, and in the development of sustainable infrastructure. The role of AI in sustainable decision-making is pivotal for the realization of Pakistan’s vision for environmental responsibility and sustainable construction. Some recent studies urged that business operations and decision-making have always been characterized by an intrinsic link with risk [104,105]. This study also answers the calls for future empirical research by some scholars [106,107] who have highlighted the need to explore the role of risk management strategies combined with sustainability management processes to demonstrate their impact on the decision-making process.

Killen, Geraldi, and Kock (2020) argued that decision-makers utilize a range of methods and software systems to aid decision-making [38]. At most times, these are defective, either for being a haphazard way of doing things or for using mathematical or statistical methods whose accurate understanding by the decision-makers is low. This problem is visible not only in risk management where organizations that do not understand contemporary risk management theory often use ineffective methods and tools to identify and assess risks. However, these decision and risk management disciplines have become far removed from practitioners and scholars failing to have a good understanding of each other’s domains while producing research that is too theoretical with no relevance to real-world decisions. Some researchers agree that AI can swing the balance with predictive insights and scenario modeling that enable decision-makers to not only anticipate uncertainties but also deal with risks efficiently [48,106]. Based on the findings of this study, it can be illustrated that AI-driven risk management is a reliable predictor of sustainable decision-making.

The study also examined the role of perceived environmental responsibility as a mediator for the relationship between AI-driven risk management and sustainable decision-making. AI-driven risk management systems not only enable the identification and mitigation of risks impacting environmental factors but also give stakeholders the tools to make rational decisions toward sustainable goals. The study’s emphasis on perceived environmental responsibility as a mediator reveals an additional mechanism through which AI-driven risk management influences the decision-making process that leads to sustainability in construction. In such cases, when construction firms believe that they are environmental stewards, they are more likely to adopt AI-driven risk management practices in their decision-making processes, so they come with more sustainable outcomes [77]. Given this, developing a culture of responsibility towards the environment in the construction sector can make AI-powered risk management more effective in making sustainable choices, thus contributing to the environmentally friendly construction approach and ensuring the desired outcome of the project. An awareness of increased environmental responsibility during decision-making practices leads managers to scale the AI-based risk management outcomes with sustainability goals. This signifies the necessity to build a sense of environmental stewardship among the higher level of management. Initiatives aimed at fostering environmental awareness and responsibility consciousness can improve the positive impact of AI technologies on the decision-making approach that favors sustainability.

5.1. Implications

5.1.1. Theoretical Implications

The study proposed and tested a theoretical framework examining the impact of AI-driven risk management on sustainable decision-making. In addition, the study assessed the role of perceived environmental responsibility as a mediator. The findings of the study contribute to the innovation diffusion theory (IDT) by confirming AI-driven risk management as a diffusible innovation that leads to sustainable practices within the construction sector. The findings revealed that the effective management of risk by utilizing AI tools can promote sustainable decision-making at the managerial level. On the other hand, the results align with the social cognitive theory (SCT) by emphasizing the role of internal factors, such as perceived environmental responsibility, in shaping sustainable decision-making through AI-driven risk management. The results confirmed that individuals having a stronger sense of environmental responsibility are more likely to understand AI-driven risk management as valued, accelerating its integration with sustainable decision-making practices. In essence, the proposed framework enlightens the novel perspective on the nexus between technology, individual internal beliefs, and sustainability. This theoretical foundation enhances the understanding and insights on AI-driven risk management as a powerful tool to stimulate sustainable decision-making, especially when coupled with a strong intrinsic motivational factor, such as environmental responsibility.

5.1.2. Practical Implications

Managerial Implications

The significance of these findings is profound for both industry practitioners and policymakers engaged in smart cities and urban planning in Pakistan. The findings stimulate managers working in the construction sector to consider AI as a sustainability enabler by integrating it with risk management. Installing AI-driven risk monitoring systems is not merely an instance of technological breakthrough; this should rather be viewed as a key strategic step towards the attainment of sustainable goals. The findings advocate that the impact of AI on sustainable decision-making is more robust if organizations have a strong commitment to environmental responsibility. Therefore, managers can focus on internal and external training programs, the recognition of environmentally responsible practices, and the formation of measurable environmental goals. This study underscores the crucial role of AI in risk management to enhance sustainability; however, managers must consider that this heavily relies on reliable and high-quality data. Managers, therefore, should have a priority to collect robust data and ensure the data quality to derive accurate results from AI models.

Implications for Employees

The findings of the study are equally important for employees and have some implications for them. With the integration of AI tools in managing risk, employees’ awareness of safety and environmental hazards can be increased. Also, they feel empowered to minimize risks by making informed decisions. This may also motivate them to minimize waste and make a contribution to the corporate sustainability goals. Furthermore, AI tools assist in scenario planning, offering employees the ability to focus on important tasks. The findings of the study emphasized the shifting skillsets and employees’ need to gain new skills regarding interpreting AI outputs and collaborating with AI systems. In the evolving construction industry, it can be projected that employees who efficiently utilize AI tools and are capable of integrating them into the work processes are expected to be well-positioned for career growth.

Societal Implications

The findings of the study have some societal implications. It highlights corporate sustainability by mitigating environmental risks linked with the business operations. AI-driven risk management facilitates identifying the potential risks and diminishing the environmental impact. Additionally, public trust can be gained by fostering transparency in AI implementations in sustainability. By enabling better decision-making and a wider adoption of AI for sustainability, the findings of the study can contribute to attaining broader societal goals related to environmental protection, resource conservation, and a sustainable future.

5.2. Limitations and Future Research Directions

In spite of the study findings stating that an AI-driven risk management system is certainly a part of the strategy for sustainable decision-making in construction, some limitations should be kept in mind. The study emphasizing the construction industry of Pakistan may have some limitations to the generalizability of its conclusions to other industries or geographical regions. The AI risk management approach relationship with sustainable decisions may in Pakistan be influenced by the unique features and context-based factors of the construction sector in a way that is different from other cases. In addition, the choice of structural equation modeling (SEM) as the major analytical method is subject to some constraints too, including the conditions of linear relationships between variables and the possibility of measurement error. In addition, the survey’s reliance on self-reported data from senior managers might be distorted by response bias or social desirability bias, which would, in turn, influence the accuracy and validity of the findings.

In addition, the study focuses on the mediating role of perceived environmental responsibility in AI-driven risk management and sustainable decision-making but ignores the other potential moderators and mediators that can shape the influence between these two factors. For example, organizational culture, regulatory frameworks, or technological infrastructure could also be among the key determinants of the way AI-driven risk management influences sustainability performance in the construction sector. The study’s cross-sectional nature also limits its capability to determine causality or to capture the development of the relationship across time. More evidence in the future can be obtained by using longitudinal studies or by running experimental designs to eventually find out the causal mechanisms responsible for the association between AI-driven risk management and sustainable decision-making. Overall, the findings offer helpful information about the advantages of AI system risk management to sustainability in the construction industry, but there should be more research to tackle these limitations and a more comprehensive understanding of the scenario.

6. Conclusions

This study examined the impact of AI-driven risk management on sustainable decision-making in the construction sector of Pakistan. Investigating this research area is not only important for advancing academic knowledge but also offers actionable insights that can guide the development of a sustainable and technologically advanced construction sector. Furthermore, the findings of the study contribute to ongoing discourse on the role of AI in shaping the future.

The results confirmed a strong and positive correlation between AI-driven risk management and sustainable decision-making. Furthermore, the study established that perceived environmental responsibility strengthens the positive impact of AI-driven risk management on sustainable decision-making. This suggests that a culture of environmental awareness in the organizations can embrace the AI presence and its effectiveness. Despite the study’s limitations, it provides valuable insights into the transformative potential of AI technologies in achieving sustainability goals. Moving forward, the dynamics of these issues should be researched in a variety of contexts and industries to contribute towards more sustainable and resilient decision-making practices globally.