Does ESG Practices Influence Financial Companies’ Performance? The Moderating Role of AI Use

Abstract

A This study examines the interplay between environmental, social, and governance (ESG) practices, artificial intelligence (AI) adoption, and financial performance within Saudi Arabia’s financial sector. It investigates whether AI adoption moderates the ESG–performance relationship, reflecting the sector’s ongoing digital transformation under Vision 2030. Drawing on 224 firm-year observations across banks, diversified financials, real estate investment trusts (REITs), and insurance companies, the study employs content analysis of annual reports to identify AI implementation. Panel regression models are used to test the effects of ESG practices on both accounting-based (ROE) and market-based (Tobin’s Q) performance measures, while examining AI’s moderating role. The results reveal that ESG practices significantly enhance accounting-based performance, particularly return on equity, while board size exerts a positive and board independence a negative influence. However, ESG does not significantly affect market-based valuation (Tobin’s Q). Notably, AI adoption negatively moderates the ESG–financial performance link, suggesting short-term challenges in integrating digital transformation with sustainability strategies. This study contributes to literature in three key ways. First, it provides new evidence from financial institutions in a developing economy—Saudi Arabia—where ESG and AI integration remains underexplored. Second, unlike previous research that proxies AI adoption through R&D expenditure, this study captures actual deployment of AI tools in operational activities. Third, it extends the ESG–performance debate by introducing AI adoption as a novel moderating factor. The findings offer actionable insights for managers and policymakers in emerging markets, underscoring the importance of developing organizational capabilities that harmonize AI-driven innovation with ESG principles to foster sustainable long-term value creation.

1. Introduction

Recent advances in artificial intelligence (AI) have attracted considerable global attention and accelerated its adoption across a wide range of business functions. In management and organizational settings, AI applications include natural language processing, chatbots, machine learning, decision-support systems, and predictive analytics. Despite its growing use, the impact of AI on corporate performance remains inconclusive (Wamba-Taguimdje et al., 2020; Domini et al., 2023). While some studies suggest that AI enhances firm performance and supports long-term sustainability (Zhang & Yang, 2024; Vinuesa et al., 2020), others argue that excessive reliance on AI may jeopardize organizational stability because of the risks associated with digitalization, particularly in firms characterized by weak governance structures (Rana et al., 2022; Song et al., 2025).

More recently, AI has become an important enabler of corporate sustainability initiatives. Organizations increasingly employ AI-based tools to assess ESG-related risks, opportunities, and stakeholder sentiment through the analysis of large volumes of structured and unstructured data. Platforms such as IBM Watson and Google Cloud NLP facilitate ESG analytics, while machine-learning applications, including Amazon SageMaker, are used to predict ESG risks. In addition, automated reporting systems, such as SAP Sustainability Solutions, improve ESG reporting quality and regulatory compliance. Collectively, these developments indicate that firms are investing in AI technologies not only to strengthen sustainability practices but also to improve overall organizational performance.

Although the literature has extensively examined AI adoption and its implications for ESG practices and firm performance, financial institutions have largely been excluded from empirical investigations (Shen, 2023; Alodat & Hao, 2025). Given the distinctive characteristics of the financial sector, sector-specific analyses are necessary to generate more reliable insights into the relationships among ESG practices, AI adoption, and financial performance (Gholami et al., 2022). Addressing this gap, the present study focuses on four major financial subsectors—banking, diversified financial services, real estate investment trusts (REITs), and insurance—to examine whether AI adoption influences the ESG–performance relationship.

The importance of this investigation is reinforced by the evolving role of ESG within financial institutions. Following the global financial crisis, financial firms have faced growing pressure from regulators, investors, and other stakeholders to enhance transparency, accountability, and social responsibility through ESG disclosures. Unlike non-financial firms, financial institutions operate within a highly regulated environment characterized by capital adequacy requirements, central bank supervision, and sustainable-finance frameworks, including the Saudi Central Bank (SAMA) Sustainable Finance Framework. Consequently, ESG performance serves not only as a reputational signal but also as an indicator of risk management quality, regulatory compliance, and access to financing.

Financial institutions also occupy a unique position within the ESG ecosystem. They are evaluated based on their own ESG performance while simultaneously influencing the sustainability outcomes of the broader economy through lending, underwriting, and investment activities (Tang et al., 2016). As a result, the mechanisms linking ESG to financial performance in financial institutions differ from those typically observed in industrial firms, where operational efficiency and supply-chain management often play a more prominent role.

Similarly, AI adoption in financial institutions is concentrated in risk assessment, credit evaluation, fraud detection, regulatory compliance, and customer service applications. These uses are subject to stringent model-risk management and explainability requirements, distinguishing them from the product and process innovations commonly associated with AI in manufacturing and other non-financial sectors. Consequently, AI is more likely to influence regulatory and risk-related outcomes than operational efficiency within financial institutions. These distinctive characteristics provide a strong rationale for examining the ESG–performance relationship within the financial sector separately rather than combining financial and non-financial firms in a single analysis.

The financial services industry offers a unique setting for examining the relationship between ESG practices, AI adoption, and financial performance. Compared with non-financial industries, financial institutions operate in a highly regulated environment and are subject to greater stakeholder scrutiny and more extensive disclosure requirements. As a result, transparency, accountability, and responsible business practices play a particularly important role in this sector.

In addition, financial institutions depend more on information processing, data analytics, and risk management than on physical production assets, making AI a strategic capability rather than simply an operational tool. AI technologies are now widely integrated into core financial activities, including credit risk assessment, fraud detection, anti-money laundering compliance, customer relationship management, and portfolio optimization. This extensive use of AI enables financial institutions to collect, analyze, and incorporate ESG-related information more effectively into strategic decisions and financial operations.

These industry-specific characteristics suggest that the relationship between ESG and financial performance, as well as the moderating role of AI, may differ from those observed in manufacturing and other service sectors. Accordingly, focusing on the financial services industry provides an appropriate and meaningful context for investigating these relationships.

Saudi Arabia offers an especially relevant context for this investigation. As one of the leading economies in the Middle East, the Kingdom has actively promoted both AI adoption and sustainable business practices. Within the financial sector, institutions have increasingly integrated AI-driven technologies into fraud detection, credit assessment, predictive analytics, customer-service platforms, compliance monitoring, and risk-management systems. These developments have been strongly supported by the Vision 2030 agenda, which aims to diversify the economy and accelerate technological innovation. The establishment of the Saudi Data and Artificial Intelligence Authority (SDAIA) in 2019 further strengthened the national commitment to AI development and implementation.

At the same time, Saudi Arabia has intensified its focus on ESG and sustainability. The Saudi Central Bank’s Sustainable Finance Framework, together with increasing expectations regarding sustainability disclosure, has encouraged financial institutions to integrate ESG considerations into governance structures, investment decisions, and reporting practices. National sustainability initiatives and commitments to global development objectives have further heightened stakeholder expectations concerning corporate responsibility and transparency. The simultaneous advancement of AI adoption and ESG integration creates a unique environment for examining how technological innovation influences the ESG–financial performance relationship.

Prior studies provide mixed evidence regarding the impact of ESG practices on financial performance in financial institutions. Several studies report a positive association between ESG performance and financial outcomes (Gholami et al., 2022; Andrieș & Sprincean, 2023), whereas others find insignificant or negative effects (Soana, 2011; Shakil et al., 2019; Bătae et al., 2020). These conflicting findings suggest that the ESG–performance relationship may depend on contextual or firm-specific factors that have not yet been adequately explored.

Against this backdrop, we propose AI adoption as a potentially important moderator of the ESG–performance relationship. AI has been increasingly applied across sectors including healthcare, agriculture, education, taxation, fraud detection, and customer services (Liu et al., 2023). Although AI adoption in financial services continues to expand, its implementation remains uneven across institutions. According to the World Bank (2024), financial firms are gradually integrating AI to improve decision-making, customer engagement, and risk management. Similarly, the McKinsey Global Institute (2024) estimates that generative AI could create annual value ranging from USD 200 billion to USD 340 billion within the global banking and financial services industry.

Despite growing interest in both ESG and AI, empirical evidence on their combined effects remains limited, particularly within financial institutions operating in emerging markets such as Saudi Arabia (Al-Baity, 2023). Existing inconsistencies in the ESG–performance literature may reflect differences in institutional environments, regulatory structures, and industry characteristics. By focusing on Saudi Arabia, this study contributes to a deeper understanding of how country-specific and industry-specific conditions shape the interplay among ESG practices, AI adoption, and financial performance.

Saudi Arabia occupies a prominent position within the Gulf Cooperation Council (GCC) due to its sophisticated financial sector and rapidly expanding capital market. Its economic transformation has been accompanied by significant progress in innovation and technological development. According to the Global Innovation Index (2023), Saudi Arabia ranked 48th among middle-income economies, representing one of the most notable improvements globally during the post-pandemic period.

The Kingdom has also demonstrated a strong commitment to sustainability at both corporate and national levels. In 2021, the Saudi Tadawul Group introduced ESG disclosure guidelines to enhance transparency and align listed firms with international best practices. At the national level, Saudi Arabia has committed to achieving net-zero carbon emissions by 2060 through investments in carbon capture technologies, clean hydrogen production, and other sustainability initiatives. These efforts are guided by the Carbon Circular Economy framework, which emphasizes reducing, reusing, recycling, and removing carbon emissions.

Parallel to these sustainability initiatives, Saudi Arabia continues to position itself as a regional leader in AI development. Since its establishment in 2019, SDAIA has played a central role in advancing AI adoption across public and private sectors, strengthening digital capabilities, and promoting data-driven innovation. These efforts support economic diversification, enhance public services, and contribute to the development of a knowledge-based economy. By 2023, Saudi-based AI firms had attracted approximately USD 1.7 billion in investment, while the AI workforce expanded at an annual growth rate of 51% between 2018 and 2022 (Al Ayyash, 2024).

Recent evidence from Saudi Arabia further highlights the importance of examining the interaction between AI and ESG. Ebnaoof (2026) reports that AI adoption is associated with higher-quality ESG disclosure by improving data collection, reporting accuracy, transparency, and stakeholder communication among Saudi listed firms. Similarly, Hamdouni (2025) finds that AI implementation is positively related to firms’ ESG performance, particularly through enhanced monitoring systems, predictive analytics, and more effective sustainability-oriented decision-making. Collectively, these studies suggest that AI can serve as an important enabler of corporate sustainability by strengthening ESG reporting and performance.

Despite these advances, important questions remain unanswered. Existing Saudi evidence has focused primarily on the direct effect of AI on ESG disclosure or ESG performance, with limited attention to whether AI influences the extent to which ESG practices translate into superior financial performance. This distinction is important because improvements in ESG reporting or sustainability outcomes do not necessarily imply stronger financial returns. Moreover, implementing AI requires substantial investments in digital infrastructure, organizational capabilities, and human capital, which may either reinforce or offset the financial benefits associated with ESG initiatives. Consequently, the moderating role of AI in the ESG–financial performance relationship remains an open empirical question, particularly within Saudi Arabia’s highly regulated financial sector, where institutions are simultaneously advancing digital transformation and sustainable finance under the Vision 2030 agenda. Addressing this gap, the present study examines whether AI adoption strengthens or weakens the relationship between ESG practices and financial performance among Saudi financial institutions.

Additional research highlights the growing integration of AI within Saudi financial institutions. Applications of machine learning and natural language processing have enhanced customer engagement, strengthened risk-management systems, and improved decision-making processes (Eskandarany, 2024b). Together, these developments illustrate how Saudi financial institutions are leveraging both AI and ESG frameworks to align with international standards while supporting the objectives of Vision 2030.

Using a sample of 224 firm-year observations from Saudi financial institutions, our findings indicate that ESG performance improves profitability, measured by return on equity (ROE). We also find that board size positively influences financial performance, whereas board independence is associated with weaker financial outcomes. More importantly, AI adoption negatively moderates the ESG–performance relationship, suggesting that financial institutions may face transitional challenges when integrating AI technologies with sustainability strategies.

This study contributes to the literature in several ways. First, unlike prior studies that primarily focus on industrial firms in developed economies (M. A. Khan, 2022; Halid et al., 2023), we examine the financial sector within a developing-country context. Second, whereas most previous studies proxy AI adoption using research and development expenditures (Acharya & Arnold, 2019; Yamashita et al., 2021; Besiroglu et al., 2024), we employ a more direct measure based on content analysis of annual reports to capture actual AI deployment. Third, we extend the debate surrounding the ESG–performance nexus by examining the moderating role of AI adoption. Given the mixed findings reported in prior studies (Bătae et al., 2020), our analysis provides new evidence on how AI integration shapes the relationship between ESG performance and financial outcomes.

Recent studies have increasingly examined technology and digital transformation as mechanisms that strengthen the relationship between ESG practices and firm performance. For example, Fu and Li (2023) report that digital transformation enhances the positive effect of ESG on corporate financial performance among Chinese listed firms, while Shafique et al. (2026) show that AI-powered big data analytics improve ESG performance through AI-enabled green learning capabilities, with responsible leadership further strengthening this relationship. Although these studies provide valuable evidence that digital technologies can reinforce sustainability outcomes, they primarily examine broad digital transformation initiatives or AI capabilities in non-financial settings using digital transformation indices or survey-based measures.

Our study extends this emerging literature in several important ways. First, rather than measuring digital transformation or perceived AI capabilities, we employ a content-analysis-based measure of actual AI implementation derived from firms’ annual reports, thereby capturing verified organizational deployment of AI technologies. Second, we focus on financial institutions, a setting that differs fundamentally from manufacturing and hospitality because financial firms operate in a highly regulated, information-intensive environment where AI is integrated into core activities such as credit assessment, fraud detection, risk management, regulatory compliance, and customer service. Consequently, the interaction between AI adoption, ESG practices, and financial performance may differ from that observed in non-financial industries.

Finally, this study contributes to the limited evidence from emerging economies by providing the first empirical examination of the moderating role of AI adoption in the ESG–financial performance relationship within Saudi Arabia’s financial sector. While previous Saudi studies have primarily investigated AI applications in areas such as banking and cybersecurity (Eskandarany, 2024a; Alraddadi, 2023), little is known about whether AI alters the effectiveness of ESG practices in improving financial performance. Our findings indicate that the positive technology-enabled ESG complementarity documented in some non-financial sectors cannot be assumed to apply universally, highlighting the importance of industry context when evaluating the role of AI in sustainable corporate performance. The remainder of the paper is organized as follows. Section 2 presents the theoretical framework. Section 3 develops the research hypotheses. Section 4 describes the methodology, while Section 5 reports and discusses the empirical findings. Section 6 concludes the study.

2. Theoretical Framework

This section develops the theoretical framework underpinning our research model by following a deductive structure: for each theory, we first outline its core tenets in general terms before deriving its specific implications for the variables examined in this study. Four theoretical perspectives are introduced in turn. Managerial Stakeholder Theory is presented first to establish why ESG practices are expected to influence financial performance. The Resource-Based View follows, explaining why AI tools constitute a potentially valuable organizational resource. Because neither of these theories addresses how and why individual organizations decide to adopt AI in the first place, the Technology Acceptance Model is then introduced to explain the behavioral mechanism underlying AI adoption decisions within financial institutions. Finally, Dynamic Capabilities Theory is presented as the integrating framework that links these three perspectives, explaining why the interaction between ESG practices and AI use—rather than either factor in isolation—is expected to shape financial performance.

In contrast to the prevalent emphasis on increasing shareholder value advocated by neoclassical thinking, Managerial Stakeholder Theory asserts that banks, as profit-seeking entities, should prioritize addressing the diverse needs of stakeholders, including customers, employees, investors, and the community (Friedman & Miles, 2002). Accordingly, investments in ESG initiatives by banks may cater to broader stakeholder concerns, fostering trust and enhancing customer satisfaction, ultimately contributing to improved financial performance (Ersoy et al., 2022). Managerial Stakeholder Theory serves as a theoretical framework for understanding the positive influence of ESG activities on firm value (McWilliams & Siegel, 2000). Furthermore, technological innovations have the potential to tackle societal and environmental challenges while promoting shared benefits and values among companies and their stakeholders (Y. Li et al., 2022).

According to the Resource-Based Theory, organizations establish a competitive advantage by acquiring resources that are valuable, rare, inimitable, and non-substitutable in order to achieve sustained long-term success (Wernerfelt, 1984). These resources enable firms to implement strategies that drive efficiency, effectiveness, and competitive superiority (Albitar et al., 2023). Drawing from the Resource-Based perspective, it is anticipated that banks can achieve long-term value by investing in unique and scarce resources such as ESG initiatives and AI tools. The adoption of ESG practices and new technologies in operational and financial activities is distinct for each bank, enhancing their sustainability and financial performance uniquely (Fischer & Sawczyn, 2013). For instance, contemporary tools such as machine learning models, predictive analytics, and big data analytics play a pivotal role in enabling banks to assess and improve their ESG performance by providing data-driven insights and reinforcing decision-making in the context of sustainability and responsible banking practices. Similarly, Corporate Social Theory underscores that ESG-focused banking strategies enhance both reputation and financial performance. Consequently, ESG activities might represent a significant tool for firms to enhance customer, employee, and investor satisfaction (Soana, 2011).

While Stakeholder Theory explains why ESG matters to financial institutions and the RBV explains why AI tools can be a valuable organizational resource, neither theory addresses the behavioral question of why and how individual financial institutions decide to adopt AI in their operations in the first place. The Technology Acceptance Model (TAM) is the relevant theory for this purpose (Davis, 1989). TAM posits that an individual’s or organization’s intention to use a new technology, and ultimately its actual use, is determined primarily by two beliefs: perceived usefulness, the degree to which the technology is believed to improve job or organizational performance, and perceived ease of use, the degree to which using the technology is believed to be free of effort. Higher perceived usefulness and ease of use generate a more favorable attitude toward technology, which in turn increases behavioral intention and actual adoption (Venkatesh & Davis, 2000). Applied to our setting, TAM suggests that AI adoption in Saudi financial institutions is shaped by decision-makers’ beliefs about whether AI tools will meaningfully improve risk management, customer service, and ESG-related reporting (perceived usefulness), and about the organizational and technical ease of deploying such tools within existing infrastructure and regulatory constraints (perceived ease of use). This behavioral mechanism complements the RBV: while the RBV explains why AI is potentially valuable as a resource, TAM explains the adoption process through which that potential is, or is not, realized in practice, which is directly relevant to interpreting the still-developing extent of AI deployment we observe across our sample.

Taken together, Managerial Stakeholder Theory, the Resource-Based View, and TAM are not treated here as separate, unrelated explanations for ESG and AI, but as complementary accounts that must be integrated to derive a coherent prediction for their joint effect on financial performance. Stakeholder Theory explains why ESG practices are valuable—they build trust and legitimacy with the diverse stakeholders on whom banks depend; the RBV explains why AI tools are potentially valuable—as resources that, in principle, are scarce and difficult to imitate; and TAM explains the behavioral process through which firms actually come to adopt and use those tools. However, none of these three theories alone specifies what happens when an ESG-oriented resource (stakeholder-facing sustainability practices) is combined with a technology-oriented resource (AI tools) once adopted within the same firm. We address this gap by drawing on Dynamic Capabilities Theory (Teece et al., 1997), which holds that resources generate competitive advantage only to the extent that a firm possesses the capacity to integrate, reconfigure, and orchestrate them coherently. Under this integrated framework, AI does not moderate the ESG–performance relationship directly through its technical functionality, but through the firm’s capability to align AI’s data-processing and decision-support functions with its stakeholder-oriented ESG objectives—for example, by directing AI-based analytics toward ESG risk identification, reporting, or stakeholder communication rather than treating the two as parallel, disconnected initiatives. This generates a specific, falsifiable mechanism: where organizational capability to integrate AI with ESG objectives is high, AI use should reinforce the ESG–performance link (the basis for H2); where such integrative capability is still developing—as is plausible in a financial sector undergoing early-stage AI adoption under an evolving regulatory regime, consistent with the TAM-based adoption mechanism outlined above—AI use may instead compete with ESG initiatives for managerial attention and organizational resources, weakening rather than strengthening the relationship. This dynamic-capabilities lens therefore provides an ex-ante theoretical basis, rather than a post hoc explanation, for why the sign of AI’s moderating effect is an empirical question and for how we interpret the direction of the result reported below.

3. Hypotheses Development

3.1. Financial Performance and ESG

The European Union has consistently revised sustainability regulations, particularly with respect to disclosure and transparency obligations for Principles for Responsible Investment (PRI) members. In its 2025 strategic plan, PRI has emphasized strengthening impact and advancing the integration of ESG principles into corporate decision-making processes. Globally, numerous initiatives reflect the increasing alignment of ESG factors with strategic business operations (M. A. Khan, 2022). Within this framework, corporate value creation has become increasingly interlinked with the Sustainable Development Goals (SDGs).

Both regulators and business leaders widely contend that firms adopting sustainability-oriented practices are more likely to build stronger customer loyalty, thereby improving their financial performance. From an academic standpoint, extensive literature supports the notion that companies committed to sustainability and ethical practices often develop enhanced reputational capital in the marketplace.

Scholars have also explored how the mode of sustainability reporting influences the relationship between ESG disclosure and corporate performance. Mervelskemper and Streit (2017), for example, report that market valuation differs depending on whether a firm issues integrated reports or standalone sustainability reports. Likewise, Bissoondoyal-Bheenick et al. (2023) find that firm size and media visibility positively affect the ESG–performance nexus among G20 firms, consistent with stakeholder theory.

Broadly, the literature examining the ESG–performance relationship can be categorized into three main streams. The first highlights a positive link, arguing that transparent ESG disclosures reduce information asymmetry, lower contracting costs, and ultimately improve financial performance (Brooks & Oikonomou, 2018). The second stream identifies a negative association between ESG practices and performance (Crisóstomo et al., 2011; Nollet et al., 2016). A third group, represented by Atan et al. (2018) and Gholami et al. (2022), finds no statistically significant relationship between the two variables.

Given these mixed results, scholars have increasingly examined how corporate governance mechanisms shape the ESG–performance nexus (Devinney, 2009; Zaman et al., 2022). For instance, Alodat and Hao (2025) show that sustainability committees and gender-diverse boards strengthen the link between ESG disclosure and firm performance across European companies. Similarly, L. Aladwey and Alsudays (2023) demonstrate that ESG scores mediate the relationship between board gender diversity and firm value. However, a majority of these studies remain focused on non-financial sectors, leaving the role of ESG in banking and financial institutions relatively underexplored.

More recent scholarship has begun addressing this gap. Miralles-Quirós et al. (2019), for example, observe a positive association between corporate governance and environmental performance among international commercial banks. In emerging economies, Azmi et al. (2021) establish that ESG efforts enhance bank value by improving cash flows and efficiency, while simultaneously reducing the cost of equity capital, though not the cost of debt. Similarly, Ersoy et al. (2022) report a U-shaped relationship between ESG and bank value in the U.S., suggesting that excessive ESG investment may diminish value due to high associated costs. Di Tommaso and Thornton (2020) further find that higher ESG scores reduce risk-taking behaviors in banks, though this effect is accompanied by lower market valuations, implying possible “overinvestment” in sustainability initiatives.

Sectoral comparisons reinforce the unique dynamics within financial institutions. Gholami et al. (2022) show that ESG disclosure positively affects profitability exclusively in the financial sector, with no parallel effect observed in non-financial firms. Evidence on specific ESG components also varies. For instance, Soana (2011) finds no significant relationship between corporate social performance and financial performance in banks, whereas Wu and Shen (2013), Bolton (2013), and Siueia et al. (2019) argue that CSR policies contribute positively to bank profitability. Conversely, Shakil et al. (2019) report that governance structures alone have limited influence on bank performance.

Taken together, the existing evidence suggests that the relationship between ESG practices and financial performance remains inconclusive, particularly within financial institutions where regulatory requirements, stakeholder expectations, and sustainability commitments differ from those observed in non-financial sectors. In Saudi Arabia, these dynamics are further reinforced by Vision 2030 initiatives, which encourage financial institutions to integrate sustainability considerations into their business strategies while maintaining strong financial outcomes. Consequently, examining whether ESG practices contribute to superior financial performance in Saudi financial institutions represents an important empirical question.

Accordingly, the study proposes the following hypothesis:

H1. 

ESG practices are positively associated with financial performance.

3.2. The Moderating Effect of AI Use

The financial sector was profoundly affected by the global financial crisis of 2007, which highlighted systemic vulnerabilities and prompted regulators to develop policies aimed at enhancing the supervision of banks and other financial institutions. Beyond traditional regulations and central bank guidelines, recent years have witnessed a noticeable acceleration in the adoption of AI tools within the financial sector. Initially, AI applications were primarily directed toward assessing loan risks, but their use has since expanded to decision-making processes that enable stricter oversight of risk distribution across the financial system as a whole. The pace of digital transformation further accelerated during the COVID-19 pandemic, reinforcing the widespread application of AI technologies. These technologies hold the potential to increase banks’ revenues by enabling personalized services for both customers and employees, reducing operational costs through automation, lowering error rates, and optimizing resource allocation.

Recent empirical research has demonstrated the value of AI implementation for improving corporate performance; however, the banking and finance sectors are often underrepresented in such studies (Narwane & Priyadarshinee, 2025; Chen et al., 2022; Bag et al., 2020; Mikalef & Gupta, 2021). Much of the prior literature has concentrated on AI’s impact on innovation processes in areas such as management strategies, technological development, and customer relationship management (Chatterjee et al., 2021; Cui et al., 2021; Varsha et al., 2021). Within the framework of the resource-based view (RBV), banks, as service-oriented institutions, stand to benefit significantly from AI adoption in several domains, including financial product innovation, customer satisfaction, and risk management. Despite these potential advantages, there remains a notable gap in understanding the composition of Artificial Intelligence Capability (AIC) in the banking industry and how such capability influences banks’ overall performance (Huynh et al., 2020). Existing studies investigating the relationship between AI implementation and firm value indicate that, although digital technologies are widely used, their impact on firm performance has produced mixed results across different cultural, contextual, and disciplinary settings (Oduro et al., 2023).

While AI adoption has been associated with improvements in financial performance, its presence also raises challenges for the effectiveness of ESG reporting strategies (Vitolla et al., 2019; Salvi et al., 2020). Specifically, AI may influence the extent to which firms integrate ESG considerations into operational and strategic decision-making, thereby shaping overall performance outcomes. In the context of the banking sector, AI has the potential to act as a moderator in the relationship between bank performance and ESG practices, particularly as banks progressively incorporate sustainability into their operational frameworks. AI tools can significantly enhance the management of ESG-related risks and opportunities. For example, AI systems can process vast datasets to identify ESG-related risks such as climate change impacts, social inequality, or governance shortcomings (P. A. Khan & Johl, 2019). By mitigating these risks, banks may reduce loan defaults, avoid regulatory penalties, and strengthen asset quality, ultimately leading to improved financial performance.

Furthermore, Natural Language Processing (NLP) techniques can be employed to analyze ESG-related big data and evaluate banks’ ESG reputation. Through automating the collection and analysis of unstructured ESG data, AI contributes to the production of more accurate and timely ESG reports. Enhanced ESG reporting, in turn, builds stakeholder trust and attracts ESG-focused investors, while also improving banks’ access to green financing channels. Such outcomes positively influence bank performance and competitiveness. Taken together, these examples illustrate a three-step mechanism linking AI to financial performance through ESG: first, AI tools (e.g., NLP and machine learning) automate the collection, validation, and analysis of ESG data that would otherwise be compiled manually from disparate operational sources; second, this automation improves the efficiency and accuracy of the ESG reporting process itself, reducing the time and staff resources required to produce disclosures and lowering the risk of reporting errors or omissions; and third, the resulting higher-quality, more timely ESG disclosure enhances firm performance indirectly, by strengthening investor and regulator confidence, easing access to ESG-linked financing, and supporting more effective identification and mitigation of ESG-related risks such as loan defaults or compliance penalties. It is through this reporting-efficiency channel, rather than through any direct technical effect of AI on financial outcomes, that AI use is expected to strengthen the ESG–performance relationship.

Accordingly, it is anticipated that the intensity of AI utilization in the banking sector will strengthen the relationship between ESG practices and financial performance, underscoring the strategic significance of integrating technological innovation with sustainability objectives. This perspective is consistent with a rapidly growing body of literature that treats technology and digital transformation as moderating mechanisms in the ESG–financial performance relationship, although this literature has so far developed largely outside the financial services context. Using Chinese A-share listed companies, Fu and Li (2023) show that ESG practices enhance corporate financial performance and that digital transformation strengthens this relationship by improving information transparency and operational efficiency. In a related vein and dealing with United Arab Emirates hospitality organizations, Shafique et al. (2026) find that AI-powered big data analytics capability enhances firms’ ESG performance, with this effect operating through an AI-enabled green learning capability that allows firms to translate analytical outputs into sustainability-oriented strategies. Both studies confirm that digital and AI-related capabilities can reinforce, rather than substitute for, the ESG–performance link, but they examine this mechanism in non-financial settings where digital transformation chiefly affects operational efficiency and information disclosure. Our study extends this emerging stream of research in three respects. First, whereas Fu and Li (2023) and Shafique et al. (2026) treat technology adoption as a continuous capability or transformation index, we measure AI adoption directly through content analysis of disclosed AI applications, capturing actual deployment rather than a general digitalization construct. Second, both studies find that technology adoption reinforces the ESG–performance link; our findings from the financial sector point in the opposite direction, suggesting that the mechanisms linking AI to ESG outcomes identified in non-financial settings—efficiency gains and improved disclosure—do not transfer straightforwardly to a heavily regulated, compliance-driven industry. Third, by focusing on financial institutions in an emerging-market context, our study complements this literature, which has so far concentrated on Chinese listed firms and service industries such as hospitality, and contributes new evidence on how AI–ESG dynamics unfold under the prudential and disclosure constraints distinctive of the financial sector. Following the integrated theoretical framework outlined in Section 2, this expected reinforcement is conditional on banks possessing sufficient dynamic capability to orchestrate AI tools and ESG objectives jointly, rather than following automatically from the presence of AI as a resource in isolation. In financial institutions, AI can also strengthen compliance with sustainability regulations, enhance risk management systems, improve resource allocation decisions, and reduce operational costs. As a result, organizations that effectively combine ESG initiatives with AI-enabled capabilities may be better positioned to translate sustainability efforts into tangible financial benefits. From a stakeholder and resource-based perspective, AI represents a complementary organizational capability that enhances the value generated from ESG investments. Figure 1 summarize the alluded relationships. Therefore, firms with higher levels of AI adoption are expected to realize stronger financial returns from their ESG practices than firms with lower levels of AI adoption. Accordingly, the following hypothesis is proposed:

H2. 

AI use positively moderates the relationship between ESG practices and financial performance, such that the positive effect of ESG practices on financial performance is stronger for firms that adopt AI technologies.

4. Sample and Research Design

4.1. Sample

Our sample consists of financial firms operating in Saudi Arabia over the period from 2020 to 2024.

Table 1. Sample selection and industry ramifications.

Description	No. of Companies	No. of Obs.
The whole dataset	62	248
minus, missing or incomplete observations	(6)	(24)
The sample dataset	56	224

outlines the sample selection process and the industry breakdown. The initial sample included 62 financial companies. However, six firms were excluded due to missing or incomplete data, resulting in a final sample of 56 companies. Observations span four years, yielding a total of 224 firm-year observations. The sample is distributed across four major financial sub-sectors: banking, diversified financials, real estate investment trusts (REITs), and insurance. The data were sourced from the Refinitiv Eikon database.

The financial sector provides a distinctive setting for examining the relationship between ESG practices and financial performance because its business model, regulatory environment, and stakeholder expectations differ fundamentally from those of non-financial industries. Unlike manufacturing and other service sectors, financial institutions rely primarily on information processing, risk assessment, and capital allocation rather than physical production activities. Consequently, ESG practices are more closely linked to governance quality, transparency, risk management, and responsible financing decisions. In addition, financial institutions operate under stringent regulatory and disclosure requirements that subject them to greater scrutiny from regulators, investors, and other stakeholders, making ESG reporting more comprehensive and consequential than in many non-financial sectors (Tang et al., 2016). These characteristics also position the financial sector as one of the earliest and most intensive adopters of AI technologies, which are increasingly embedded in core business functions such as credit assessment, fraud detection, anti-money laundering compliance, customer relationship management, and investment decision-making. As a result, AI is expected to play a more pronounced role in shaping the relationship between ESG practices and financial performance than in industries where AI is primarily used to support operational or production processes.

Within this context, Saudi Arabia’s financial sector offers a particularly relevant setting for investigating the moderating role of AI in the ESG–financial performance relationship. During the 2020–2024 period, financial institutions significantly accelerated investments in AI while simultaneously strengthening their commitment to ESG objectives in response to national regulatory initiatives. For example, the Saudi Central Bank (SAMA) introduced the Sustainable Finance Framework in 2021 to encourage environmentally and socially responsible financing practices. Similarly, the Public Investment Fund (PIF) expanded investments in sustainability-oriented projects, reinforcing the Kingdom’s commitment to sustainable development. At the same time, Saudi Arabia intensified its national AI agenda under Vision 2030 through the establishment of the Saudi Data and Artificial Intelligence Authority (SDAIA) in 2019, which has played a central role in promoting AI adoption across strategic sectors, including financial services. More recently, the Riyadh AI Declaration (2024) highlighted the potential of AI to support the United Nations Sustainable Development Goals by enhancing environmental sustainability, economic resilience, and social well-being. Collectively, these industry-specific characteristics and policy developments make Saudi Arabia’s financial sector a compelling context for examining how AI influences the relationship between ESG practices and financial performance.

4.2. Measure of Variables

Table 2. Definition of Variable.

Variable	Measures	Sources
Dependent variable; Firm performance (FP)	It is measured by two measures. First, ROE evaluates a firm’s efficiency in using shareholders’ equity to generate profit, calculated as net income divided by total equity. Second, Tobin’s Q (TQ), a market-based performance indicator, is the ratio of the market value of equity plus total liabilities to total assets.	Alahdal et al. (2024); Ab Aziz et al. (2025); García-Amate et al. (2023); Alshdaifat et al. (2025)
Independent variable; (ESG)	It is a proxy of companies’ sustainable performance.	The Refinitiv Eikon database.
Moderating variable; (AI intensity)	It is measured using an index that captures variation in the scope of AI implementation across firms.	Rosa and Kubota (2025)
Control variables
Firm size (SIZE)	It is measured as the natural logarithm of total assets at yead end.	Elgharbawy and Aladwey (2025)
Leverage (LEV)	It is the ratio of total liabilities to total assets—captures financial risk.	Wintoki et al. (2012); L. Aladwey and Alsudays (2023)
Liquidity (LIQ)	It is calculated as current assets divided by current liabilities, indicating short-term financial health.	Gunawan (2023)
Firm age (FAGE)	It is representing the natural logarithm of number of years since incorporation, is used to capture organizational maturity.	Petruzzelli et al. (2018).
Board size (BS)	It is representing the natural logarithm of total number of directors.	L. Aladwey and Alsudays (2023)
Board independence (BI)	It is calculated as the proportion of independent directors.	L. Aladwey and Alsudays (2023)
Industry effect (INDUSTRY)	It is utilized by including dummies for each financial sector-REITs; diversified financials; insurance and banks.	Abu Afifa et al. (2025)
Firm size (SIZE)	It is measured as the natural logarithm of total assets at yead end.	Elgharbawy and Aladwey (2025)

shows the definition of variables utilized. In our paper, FP serves as the dependent variable and is measured through two distinct metrics to ensure robustness. First, Return on Equity (ROE) measures how efficiently a company utilizes shareholders’ equity to generate profits. It is calculated by dividing net income by total equity (Alahdal et al., 2024; Ab Aziz et al., 2025). Second, Tobin’s Q (TQ) reflects market-based performance and is derived by summing the market value of equity and total liabilities, then dividing by total assets (García-Amate et al., 2023; Alshdaifat et al., 2025). The main independent variable is the ESG score, sourced from the Refinitiv database. The ESG score serves as an indicator of a firm’s overall sustainability practices. As reported by the Refinitiv Eikon database, this composite score integrates performance across environmental, social, and governance dimensions, with values ranging from 0 to 100 (Elgharbawy & Aladwey, 2025).

The moderating variable is measured using AI intensity, which captures the breadth of AI implementation across firms. This measure is derived from the content analysis and is based on seven AI application categories adapted from Rosa and Kubota (2025): (1) text mining, (2) speech recognition, (3) natural language processing, (4) image recognition, (5) machine learning, including deep learning techniques, (6) workflow automation, and (7) autonomous machine operation. For each firm-year observation, a value of one is assigned to an application category when the annual report provides verifiable evidence that the technology has been implemented in the firm’s operations, products, services, or decision-making processes; otherwise, a value of zero is assigned. The AI implementation intensity index is computed as:

$${\mathrm{AI}}_{ikt}\mathrm{intensity}=\sum _{\mathrm{k}=1}^7\mathrm{AI}$$

where AI_it equals one if firm i discloses the implementation of AI application category k in year t, and zero otherwise. The resulting index ranges from 0 to 7, with higher values representing a broader deployment of AI technologies across organizational activities.

Consistent with Rosa and Kubota (2025), AI applications were identified based on disclosures related to one or more of the following categories: (1) text mining, where AI is used to analyze and extract insights from written language; (2) speech recognition, involving the conversion of spoken language into machine-readable formats; (3) natural language processing (NLP), which enables the understanding and processing of written or spoken language; (4) image recognition, allowing the identification and interpretation of visual content; (5) machine learning, including advanced techniques such as deep learning for pattern recognition, predictive analytics, and data-driven decision-making; (6) workflow automation, where AI is employed to automate routine operational activities and business processes; and (7) autonomous machine operation, whereby AI enables machines or systems to make decisions and perform tasks with minimal human intervention.

To ensure robustness, a variety of control variables were incorporated into the analysis. First, firm-specific characteristics were considered. Firm size (SIZE), measured as the natural logarithm of total assets, captures the scale of a company’s operations (Elgharbawy & Aladwey, 2025). In addition, leverage (LEV), defined as the ratio of total liabilities to total assets, serves as an indicator of financial risk (Wintoki et al., 2012). Liquidity (LIQ), calculated by dividing current assets by current liabilities, reflects a firm’s ability to meet short-term obligations (Gunawan, 2023). Moreover, firm age (FAGE), measured as the natural logarithm of the number of years since the company’s incorporation, is used to capture organizational maturity and longevity in the market (Petruzzelli et al., 2018). Alongside these, governance-related variables were also integrated into the model. Specifically, board size (BS) is represented by the natural logarithm of the total number of directors, offering insights into the board’s structural composition. Board independence (BI), which denotes the proportion of independent or non-executive directors, is included to reflect the board’s monitoring capacity. Finally, to control for sector-specific influences, an industry dummy variable (IND) was introduced, distinguishing firms across different financial sectors (Abu Afifa et al., 2025).

4.3. Research Model

According to Hayes and Rockwood (2017), moderation analysis examines the conditions under which an independent variable (IV) influences a dependent variable (DV) by introducing a third variable—the moderator—that alters the strength or direction of this relationship. In this study, we investigate whether the effect of ESG performance (IV) on FP (DV) varies according to firms’ adoption of AI, which serves as a binary moderator.

AI adoption is measured as a dichotomous variable, where a value of 1 indicates that a firm has adopted and disclosed the implementation of AI-related applications and technologies, and 0 otherwise. Following Rosa and Kubota (2025), AI adoption is identified through a systematic content analysis of annual reports based on evidence of specific AI applications, including text mining, speech recognition, natural language processing, image recognition, machine learning, workflow automation, and autonomous machine operations. While this measure does not capture differences in the intensity or sophistication of AI implementation, it provides a practical and transparent approach for identifying organizational AI adoption in the absence of standardized quantitative measures.

To evaluate the conditional effect of AI adoption on the ESG–FP relationship, moderated multiple regression is employed. This approach is appropriate for models involving a continuous independent variable (ESG), a dichotomous moderator (AI), and a continuous dependent variable (FP) (Shieh, 2009; Sathyanarayana & Mohanasundaram, 2025). Following Hayes and Rockwood (2017) and Hayes (2018), the analysis proceeds in two stages. First, the direct effect of ESG performance on firm performance is estimated without the interaction term to establish the baseline relationship. Second, AI adoption is introduced as an additional predictor together with the interaction term (ESG × AI) to assess whether the relationship between ESG performance and firm performance differs between AI-adopting and non-AI-adopting firms. Accordingly, the following models are estimated:

• FP as measured by ROE: The direct effect of ESG

  ROE_it = β₀ + β₁ ESG_it + β₂ FS_it + β₃ LEV_it + β₄ LIQ_it + β₅ BS_it + β₆ BI_it + β₇ FAGE_it + B₈ INDUSTRY + FIRM
  and YEAR Fixed effect + ε_it ----------Model A                            

The moderation effect model, AI

ROE_it = β₀ + β₁ ESG_it + β₂ AI + β₃ AI_it × ESG_it + β₄ FS_it + β₅ LEV_it + β₆ LIQ_it + β₇ BS_it + β₈ BI_it + β₉ FAGE_it
+ B₁₀ INDUSTRY + FIRM and YEAR Fixed effect + ε_it -------Model B                

• FP as measured by TQ: The direct effect of ESG

  TQ_it = β₀ + β₁ ESG_it + β₂ FS_it + β₃ LEV_it + β₄ LIQ_it + β₅ BS_it + β₆ BI_it + β₇ FAGE_it + B₈ INDUSTRY + FIRM and
  YEAR Fixed effect + ε_it ---------Model C                               

The moderation effect model, AI

TQ_it = β₀ + β₁ ESG_it + β₂ AI + β₃ AI_it × ESG_it + β₄ FS_it + β₅ LEV_it + β₆ LIQ_it + β₇ BS_it + β₈ BI_it + β₉ FAGE_it +
B₁₀ INDUSTRY + FIRM and YEAR Fixed effect + ε_it -------Model D                

where i represents the firm index and t represents the year index.

5. Discussion

5.1. Descriptive Statistics

Table 3. Summary statistics for utilized variables.

Panel A: Descriptive Analysis for Utilized Variables
Variable		Obs	Mean	Std. Dev.	Min	Max
ROE		224	0.079	0.363	−0.934	2.002
TQ		224	0.207	0.306	0	0.85
ESG		224	12.691	20.301	0	72.95
FS		224	19.935	2.284	12.89	23.412
LEV		224	5.595	17.311	0.212	114.642
LIQ		224	2.908	2.433	0.011	9.875
FAGE		224	1.219	0.343	0.477	1.839
BS		224	1.99	0.355	1.099	2.639
BI		224	3.065	1.341	0	7
Panel B: The number of AI’s observation per industry
AI	REITs	Diversified Financials		Insurance	Banks	Total
0	30	16		36	68	150
1	14	8		16	36	74
total	44	24		52	104	224

presents the summary statistics for all variables included in the analysis. Panel A, Table 3 displays the descriptive statistics for the main variables employed in this paper, based on a sample of 224 firm-year observations from financial companies operating in Saudi Arabia. The mean ROE is 7.9%, with a relatively high standard deviation of 36.3%, highlighting considerable variability in the profitability levels among these firms. TQ has an average value of 0.207 and a standard deviation of 0.306. The observed range (0 to 0.85) suggests that the market perception of firm value varies notably across the financial sector in Saudi Arabia. ESG score averages 12.691, with a substantial standard deviation of 20.301, reflecting a wide disparity in firms’ sustainability practices. The minimum ESG score is 0, while the maximum reaches 72.95, suggesting that while some financial firms are deeply engaged with ESG initiatives, others report no sustainability activities at all.

For control variables, FS records a mean value of 19.935, with a range between 12.89 and 23.412. This indicates that the sample includes both smaller and very large financial institutions within the Saudi market. LEV shows an average of 5.595 and a very high standard deviation of 17.311, pointing to significant differences in the capital structures and financial strategies employed by the sampled firms. LIQ, ranging from 0.011 to 9.875, also illustrates substantial variation in liquidity management practices among Saudi financial firms. The mean of FAGE is 1.219, indicating that most firms in the sample are relatively mature. BS exhibits an average of 1.99, and BI reports a mean of 3.065, highlighting variation in the extent of size and independence across boards in the Saudi financial sector. Following Elgharbawy and Aladwey (2025), the data were winsorized at the 1st percentile.

Panel B, Table 3 presents the distribution of AI adoption across financial industries. These patterns suggest that AI adoption within the Saudi financial sector is uneven, with certain industries—particularly banks—showing a relatively higher engagement with emerging technologies compared to other sectors.

5.2. Correlation Matrix and Multicollinearity Diagnosis

Table 4. Pearson correlation and multicollinearity statistics.

Variables	Panel A: Pairwise Correlation									Panel B: VIF Diagnosis
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	VIF	1/VIF
(1) ROE	1.000									1.12	0.893
(2)TQ	0.055	1.000								1.86	0.536
	(0.417)
(3) ESG	0.183	0.585	1.000							1.45	0.680
	(0.006)	(0.000)
(4) FS	−0.044	−0.262	0.001	1.000						1.30	0.767
	(0.516)	(0.000)	(0.991)
(5) LEV	−0.003	−0.151	−0.150	−0.313	1.000					1.26	0.794
	(0.967)	(0.024)	(0.025)	(0.000)
(6) LIQ	0.032	−0.474	−0.177	0.233	0.217	1.000				1.29	0.778
	(0.629)	(0.000)	(0.008)	(0.000)	(0.001)
(7) FAGE	0.063	0.380	0.453	−0.137	−0.070	−0.128	1.000			1.53
	(0.350)	(0.000)	(0.000)	(0.040)	(0.297)	(0.056)
(8) BS	0.126	0.485	0.488	−0.146	−0.092	−0.065	0.234	1.000		2.21	0.452
	(0.060)	(0.000)	(0.000)	(0.029)	(0.169)	(0.332)	(0.100)
(9) BI	−0.156	0.077	0.246	0.009	−0.020	0.185	0.561	0.285	1.000	1.59	0.630
	(0.019)	(0.252)	(0.000)	(0.895)	(0.766)	(0.005)	(0.000)	(0.015)

demonstrates the findings from the Pearson correlation analysis and the multicollinearity diagnostics. The Pearson correlation matrix is commonly applied to assess the strength and direction of relationships among continuous variables (Schober et al., 2018). As demonstrated in Table 4, panel A, the correlation coefficients among the independent variables are all below 0.6, which suggests that multicollinearity is not a serious issue (Dormann et al., 2013). All Variance Inflation Factor (VIF) scores fall below the threshold of 10, suggesting that multicollinearity does not pose an issue in the current regression analysis as reported in Table 4, Panel B. This finding aligns with Hair et al. (2019), who consider VIF values under 10 to indicate an acceptable level of correlation among explanatory variables.

5.3. Regression Results

Table 5. The Regression results of direct effect and the moderation effect.

Variables	Panel A: Direct Effect		Panel B: Moderation Effect
	Model A: ROE Coefficient/Sig.	Model B: TQ Coefficient/Sig	Model C: ROE Coefficient/Sig.	Model D: TQ Coefficient/Sig
ESG	0.005 ***	0.004 **	0.007 ***	0.001
FS	−0.008	−0.006	−0.007	0.008 *
LEV	0.030	0.000	0.000	−0.002 ***
LIQ	0.015	0.015	0.016	−0.001
FAGE	0.024	−0.031	0.058	0.002
BS	0.315 ***	0.306 ***	0.312 ***	0.036
BI	−0.122 ***	−0.103 ***	−0.102 ***	0.001
INDUSTRY	0.331 *	0.130 *	0.137 *	−0.588 ***
AI Intensity		0.301		0.041
AI Intensity × ESG		−0.014 *		−0.021
Constant	−0.369 *	−0.242 *	0.368 *	0.496 ***
Overall R-squared	0.146	0.152	0.156	0.202
Number of Obs.	224	224	224	224
Hausman Test	0.8783	0.7420

*** p < 0.01, ** p < 0.05, * p < 0.1.

reports the results of the direct-effect models (Model A and Model B) and the moderation models (Model C and Model D). The Hausman test statistics yield p-values of 0.8783 and 0.7420 for Models A and B, respectively, indicating that the differences between the fixed-effects and random-effects estimators are not statistically significant. Therefore, the null hypothesis of the Hausman test cannot be rejected, supporting the use of the random-effects estimator, which is both consistent and efficient for the current dataset.

5.3.1. Direct Effects of ESG on Financial Performance

The results of Table 5, Model A indicate a positive and statistically significant association between ESG performance and firm profitability (β = 0.005, p < 0.01). These results support the growing body of literature asserting that proactive ESG strategies contribute positively to internal financial performance, especially through improved operational efficiency, better stakeholder relationships, and reduced regulatory risk (Eccles et al., 2014; Friede et al., 2015). Companies with strong ESG profiles are often more transparent and better governed, which enhances investor confidence and reduces the cost of capital (Lins et al., 2017). However, Model B shows that ESG’s effect on market valuation is positive but not statistically significant (β = 0.006). These findings suggest that while ESG may improve accounting-based performance (ROE), it may not yet be fully recognized by the market in terms of valuation. This aligns with recent findings by H. Li et al. (2022) and García-Meca et al. (2015), who report that investors in some markets may remain skeptical about the financial relevance of ESG activities, especially if they perceive this as non-core or symbolic rather than substantive. In Saudi Arabia, ESG practices remain non-mandatory for companies, reflecting the early stage of institutionalization of sustainability within the national corporate governance framework. Although the Capital Market Authority (CMA) and the Saudi Stock Exchange (Tadawul) have introduced voluntary guidelines and reporting frameworks to encourage firms to integrate ESG considerations, these initiatives remain recommendatory rather than legally binding (Saudi Exchange, 2021; CMA, 2019). As a result, corporate adoption of ESG practices varies significantly. Leading firms—particularly those listed on Tadawul and multinational enterprises—are more likely to disclose ESG-related information, while many others engage only symbolically or abstain entirely (Alotaibi & Hussainey, 2016; Hussain et al., 2024).

This voluntary approach contrasts with emerging global standards, where mandatory ESG disclosure is increasingly embedded in regulatory frameworks, such as the European Union’s Corporate Sustainability Reporting Directive (CSRD) and the United Kingdom’s mandatory climate-related disclosures (Kotsantonis & Pinney, 2022). By comparison, the Saudi framework reflects a gradualist regulatory philosophy aligned with Vision 2030, which positions sustainability and responsible corporate governance as long-term national priorities (Vision 2030, 2021; Al-Shammari, 2022). However, in practice, the absence of compulsory requirements means that ESG adoption in Saudi firms is largely driven by institutional investors, international stakeholder pressures, and reputational incentives rather than domestic legal obligations (Alotaibi & Hussainey, 2016; Hussain et al., 2024). This aligns with stakeholder theory, the perspective that suggests various stakeholder groups’ interests or goals often differ from one another, implying that the company needs to satisfy the needs of its ‘key’ stakeholders to maintain access to the resources necessary for its survival (Freeman, 1999; Brenner & Cochran, 1991).

The voluntary nature of ESG practices in Saudi Arabia raises several critical implications. First, it weakens comparability and transparency across firms, limiting investors’ ability to assess corporate sustainability performance effectively (Alotaibi & Hussainey, 2016). Second, it creates risks of “greenwashing,” where firms adopt ESG discourse without meaningful implementation (Ioannou & Serafeim, 2019). Finally, it may slow the cultural embedding of ESG principles, particularly in family-owned or state-affiliated firms where shareholder primacy continues to dominate governance logics (Al-Shammari, 2022). Nevertheless, ongoing initiatives including Tadawul’s ESG guidelines, the launch of its sustainability index, and CMA’s growing emphasis on sustainable finance indicate a gradual institutional shift that could ultimately pave the way for the formalization of mandatory ESG requirements in the Saudi corporate sector (Saudi Exchange, 2021; Hussain et al., 2024).

Consequently, what has already been said unequivocally indicates that ESG practices positively influence company performance (ROE), aligning with existing literature and preliminary data in Saudi Arabia. Consequently, Hypothesis 1 (H1): Bank performance is positively correlated with ESG policies is supported.

5.3.2. Role of AI and the ESG–AI Interaction

The inclusion of AI in Models C and D, shown in Table 5, Panel B, enables an examination of whether AI strengthens or weakens the ESG–performance relationship. The main effect of AI on ROE and TQ is positive but statistically insignificant (Model C: β = 0. 301; Model D: β = 0.041), suggesting that AI intensity alone does not directly improve financial performance. This finding may reflect the relatively early stage of AI implementation within Saudi financial institutions, where AI applications remain concentrated in specific functions such as fraud detection, risk assessment, and customer service rather than being fully embedded across organizational processes (Kokina & Davenport, 2017; Alhassan & Asutay, 2021). Although Saudi Arabia has made substantial progress in promoting digital transformation through Vision 2030 and the National Strategy for Data and Artificial Intelligence (NSDAI), the development of organizational capabilities required to generate value from AI remains an ongoing process (SDAIA, 2020).

More importantly, the interaction term AI*ESG is negative and weakly significant in Model C (β = −0.014, p < 0.1), while remaining insignificant in Model D. This finding suggests that AI does not strengthen the positive impact of ESG on financial performance and may, in the case of accounting-based performance, weaken the extent to which ESG initiatives translate into improved financial outcomes. Rather than contradicting prior studies that highlight the benefits of AI, this result can be understood through the complementary lenses of the RBV and Institutional Theory.

From an RBV perspective, both AI intensity and ESG initiatives are resource-intensive strategic investments that require substantial financial capital, managerial attention, technological infrastructure, and specialized expertise (Barney, 1991; Peteraf, 1993). While AI has the potential to improve ESG analytics, monitoring, and decision-making, these benefits depend on the development of complementary organizational capabilities, including robust data governance systems, sustainability-oriented digital infrastructure, and ethical AI frameworks (Singh et al., 2023). This argument is consistent with recent Saudi evidence suggesting that AI technologies can strengthen ESG disclosure quality and improve sustainability outcomes through enhanced data processing and monitoring capabilities (Ebnaoof, 2026; Hamdouni, 2025). However, the realization of these benefits depends on the development of complementary organizational capabilities and effective integration between AI and sustainability strategies. Consequently, when firms simultaneously pursue AI-driven digital transformation and ESG enhancement, resource allocation trade-offs may emerge. During the capability-building phase of AI adoption, organizations may prioritize investments in digital infrastructure, employee training, and technological integration, thereby reducing the resources available to convert ESG initiatives into immediate financial returns. In this context, the negative moderating effect reflects not a failure of ESG or AI individually, but rather the challenges associated with managing two resource-intensive strategic priorities simultaneously.

This explanation is particularly relevant in the Saudi Arabian context. Under Vision 2030, firms face concurrent pressures to advance both digital transformation and sustainability agendas. However, the institutional environment surrounding AI intensity is currently stronger and more mature than that surrounding ESG. Through initiatives led by the Saudi Data and Artificial Intelligence Authority (SDAIA) and the National Strategy for Data and Artificial Intelligence, AI has become a central component of Saudi Arabia’s economic transformation agenda, creating strong incentives for firms to invest in digital capabilities and technological innovation. By contrast, although ESG disclosure and sustainability practices have gained momentum through initiatives promoted by the Capital Market Authority (CMA) and Tadawul, ESG implementation remains largely voluntary and continues to evolve within a relatively emerging institutional framework. Consistent with Institutional Theory (DiMaggio & Powell, 1983; Scott, 2014), firms are likely to prioritize initiatives associated with stronger legitimacy and regulatory pressures. Consequently, managerial attention and organizational resources may be disproportionately directed toward AI capability development, limiting firms’ ability to fully capture the financial benefits of ESG activities during the transition period.

The findings therefore suggest that the relationship between AI, ESG, and financial performance is shaped by organizational capability development and institutional priorities. The negative moderating effect should not be interpreted as evidence that AI undermines ESG value creation. Rather, it reflects a transitional stage in which Saudi financial institutions are still developing the complementary resources and dynamic capabilities required to align AI investments with sustainability objectives (Teece et al., 1997; Eisenhardt & Martin, 2000). As these capabilities mature and AI becomes more deeply integrated with ESG-related processes, the relationship may become increasingly complementary. Therefore, Hypothesis H2 is not supported.

Within this industry context, Saudi Arabia’s financial sector offers a particularly suitable setting for investigating the moderating role of AI in the ESG–financial performance relationship. The sector has experienced significant regulatory and technological transformation driven by the objectives of Vision 2030, which prioritizes digital innovation and sustainable economic development. Financial institutions have been encouraged to strengthen ESG practices while simultaneously accelerating AI intensity to improve operational efficiency, risk management, and service quality. For example, the Saudi Central Bank (SAMA) introduced the Sustainable Finance Framework in 2021 to promote environmentally and socially responsible financing practices, while the establishment of the Saudi Data and Artificial Intelligence Authority (SDAIA) in 2019 accelerated the integration of AI across strategic sectors, including financial services. In addition, the Riyadh AI Declaration (2024) recognizes AI as a key enabler of the United Nations Sustainable Development Goals, reinforcing the strategic convergence of AI innovation and sustainability within the Kingdom. Therefore, the Saudi financial sector during the 2020–2024 period provides a timely and policy-relevant environment for examining how AI influences the relationship between ESG practices and financial performance.

5.3.3. Insights from Control Variables

Several control variables offer additional insights. Firm size (FS) shows no significant effect on ROE but is positively associated with TQ (Model D, β = 0.008, p < 0.05), suggesting that larger firms are more favorably valued by the market, possibly due to perceived stability or resource availability for innovation and ESG initiatives. Board size (BS) has a consistently positive and highly significant effect on ROE (Models A and C, p < 0.01), reinforcing prior evidence that larger boards may provide broader expertise, enhanced oversight, and diverse perspectives that support strategic decision-making and organizational performance (Adams & Ferreira, 2007). In contrast, board independence (BI) exhibits a negative and statistically significant association with ROE in both models (β = −0.122 and −0.102, p < 0.01). This finding contrasts with the conventional corporate governance literature, which generally associates independent directors with stronger monitoring, improved transparency, and reduced agency conflicts (Fama & Jensen, 1983; Dalton et al., 1998). One possible explanation is that formal independence alone may not necessarily translate into effective oversight of increasingly complex strategic initiatives, such as AI intensity and ESG integration. Independent directors may face information asymmetries, possess limited firm-specific knowledge, or encounter difficulties in evaluating highly specialized technological and sustainability-related decisions. Consequently, the effectiveness of independent directors may depend not only on their formal independence but also on their expertise, engagement, and ability to contribute to strategic decision-making. Therefore, the negative coefficient should not be interpreted as evidence that board independence is inherently detrimental to firm performance. Rather, it may reflect contextual factors influencing the effectiveness of board oversight within Saudi financial institutions, where governance structures, digital transformation initiatives, and sustainability practices continue to evolve.

The industry variable exhibits divergent effects across performance measures, showing a positive association with ROE (Models A and C, significant at p < 0.1) but a negative and significant association with TQ (Models B and D, p < 0.01). These findings suggest that sector-specific characteristics may influence accounting-based and market-based performance differently. However, the underlying mechanisms cannot be directly observed from the present analysis. The observed differences may reflect variations in competitive conditions, regulatory environments, risk profiles, technological intensity, or stakeholder expectations across industries. Future research could further investigate how industry-specific factors shape the effectiveness of ESG initiatives and AI intensity in influencing firm performance.

5.4. Endogeneity Analysis: GMM Robustness Test

To further address potential endogeneity arising from reverse causality, simultaneity, and omitted-variable bias, we performed an additional robustness analysis using the two-step System GMM estimator proposed by Arellano and Bover (1995). This method is particularly appropriate for investigating the ESG–financial performance nexus because firms’ financial performance may influence their ESG engagement, while ESG initiatives may simultaneously affect financial outcomes. By incorporating lagged values of potentially endogenous variables as internal instruments and controlling for unobserved firm-specific heterogeneity, the System GMM approach helps mitigate these endogeneity concerns (Eugster et al., 2024; L. M. Aladwey, 2026). As reported in

Table 6. GMM Endogeneity test.

Panel A. GMM Model results
Variable	Model E: L.ROE		Model F: L.TQ
	Coef	p-value	Coef	p-value
L.ROE/L.TQ	0.964	0.000	0.080	0.01
ESG	0.109	0.090	0.091	0.024
FS	−0.006	0.796	−0.002	0.032
LEV	0.050	0.912	0.001	0.001
LIQ	0.020	0.174	−0.032	−0.032
FAGE	0.605	0.059	0.013	0.013
BS	0.216	0.000	0.233	0.03
BI	−0.132	0.000	−0.546	0.03
AI	0.343	0.06	−0.254	0.000
ESG × AI	−0.011	0.048	−0.382	0.021
Panel B. GMM Model Diagnostics
Diagnostic Test	Statistic	p-Value	Statistic	p-Value
Arellano-Bond AR(1)	z = −0.89	0.373	z = −0.92	0.113
Arellano-Bond AR(2)	z = −0.24	0.810	z = −0.24	0.310
Sargan Test	χ2(16) = 23.41	0.103	χ2(16) = 11.41	0.110
Hansen J Test (Robust)	χ2(16) = 51.44	0.200	χ2(16) = 31.44	0.110

, the diagnostic tests indicate that the System GMM models are statistically valid. For Model E (ROE), the Arellano-Bond AR(2) test is insignificant (z = −0.24, p = 0.810), suggesting the absence of second-order serial correlation in the differenced residuals. A similar result is obtained for Model F (Tobin’s Q), where the AR(2) statistic is also insignificant (z = −0.24, p = 0.310). In addition, the Hansen J test does not reject the null hypothesis of instrument validity for either the ROE model (χ² = 51.44, p = 0.200) or Tobin’s Q model (χ² = 31.44, p = 0.110). The insignificant Sargan test results provide further evidence that the selected instruments are appropriate. Overall, these diagnostics support the reliability of the System GMM specifications and indicate that the reported findings are unlikely to be driven by endogeneity. The results also reveal strong persistence in financial performance. The coefficients on the lagged dependent variables are positive and highly significant for both L.ROE (β = 0.964, p < 0.01) and L.TQ (β = 0.080, p < 0.05). This evidence confirms the dynamic nature of firm performance and supports the use of a dynamic panel-data estimator.

5.5. Robustness Test: Alternative AI Implementation Intensity Measure

To further assess the robustness of the findings and address potential concerns regarding the use of a binary AI adoption measure, we employ an alternative proxy. AI adoption is measured as a binary indicator, where a value of 1 denotes that a firm has adopted and disclosed the implementation of AI-related tools and 0 otherwise. AI adoption was identified through a systematic content analysis of firms’ annual reports; that is, firms were not classified as AI adopters based solely on generic references to AI. Rather, subject to data availability and consistency with prior literature (e.g., Rosa & Kubota, 2025; L. Aladwey, 2026), classification required evidence of the implementation or use of specific AI applications within organizational operations, products, services, or decision-making processes. While this binary measure does not capture differences in the intensity, sophistication, or maturity of AI implementation across firms, it provides a practical and transparent approach for identifying organizational AI adoption in the absence of standardized quantitative measures of AI capability. This approach is particularly suitable for studies of emerging markets, where detailed firm-level data on AI investments and deployment remain limited (L. Aladwey, 2026). The robustness results are reported in

Table 7. Moderation Effect (AI binary effect).

Variables	Model G: ROE Coefficient/Sig.	Model H: TQ Coefficient/Sig
ESG	0.007 ***	0.002
FS	−0.007	0.011 **
LEV	0.000	−0.002 ***
LIQ	0.016	0.001
FAGE	0.058	−0.032
BS	0.312 ***	0.013
BI	−0.102 ***	0.000
INDUSTRY	0.137 *	−0.454 ***
Constant	0.368 *	0.582 ***
Overall R-squared	0.156	0.315
Number of Obs.	224	224
AI	0.037	0.008
AI×ESG	−0.005 *	−0.030

*** p < 0.01, ** p < 0.05, * p < 0.1.

. The findings remain broadly consistent with those presented in Table 5, indicating that the principal conclusions are not driven by the choice of AI measure.

6. Conclusions

This study examined the relationship between ESG practices, AI adoption, and financial performance in Saudi Arabian banks. The findings show that ESG initiatives improve accounting-based performance, as measured by ROE, but have no significant impact on market-based performance (Tobin’s Q). This suggests that while ESG practices enhance operational profitability, investors in an emerging market such as Saudi Arabia may not yet fully incorporate ESG information into firm valuation.

The results also highlight the importance of corporate governance. Larger boards are associated with higher profitability, whereas greater board independence has a negative effect on financial performance. These findings indicate that board structure remains an important determinant of bank performance within the Saudi banking sector.

The role of AI is more complex. Although AI intensity shows a marginally positive direct association with performance, it negatively moderates the relationship between ESG and financial performance. Rather than strengthening the benefits of ESG, AI appears to reduce them during the current stage of adoption. This finding suggests that the integration of AI and ESG is still evolving. In the short term, implementation costs, organizational adjustments, and capability gaps may offset the expected benefits. From a resource-based perspective, firms must develop complementary capabilities, including effective governance, ethical AI frameworks, and managerial expertise, before AI can enhance the value created through ESG initiatives.

These findings should be interpreted within the Saudi institutional context. ESG reporting remains largely voluntary, while AI intensity is still developing across the financial sector. As a result, many banks may lack the technological infrastructure, skilled workforce, and governance mechanisms needed to create synergies between ESG and AI. This highlights the importance of gradually building organizational capabilities while strengthening regulatory support in line with Saudi Vision 2030.

Overall, the findings convey two important messages. First, ESG adoption generates clear benefits for accounting-based financial performance. Second, AI has not yet enhanced these benefits because its implementation remains at an early stage. As AI capabilities mature and become more closely aligned with sustainability strategies, stronger complementarities may emerge.

This study has several limitations. First, it focuses exclusively on Saudi Arabian banks, which limits the generalizability of the findings to other industries and countries. Second, the analysis relies on quantitative secondary data and therefore cannot capture managerial perspectives or organizational processes that influence ESG and AI adoption.

Future research could address these limitations by examining non-financial sectors, conducting cross-country comparisons, and employing mixed-method or longitudinal research designs. Developing more detailed measures of AI maturity, such as AI investment intensity, technological infrastructure, or AI-related expenditures, would provide a more comprehensive understanding of how AI influences the ESG–performance relationship. Future studies may also examine the moderating roles of regulatory quality, digital maturity, ownership structure, or board characteristics.

This study contributes to the literature by providing evidence from the Saudi banking sector on the interaction between ESG practices, AI adoption, and financial performance. It extends stakeholder theory and the resource-based view by showing that ESG creates financial value, but that AI does not automatically strengthen this relationship. Instead, realizing the combined benefits of ESG and AI depends on the development of complementary organizational capabilities and supportive institutional conditions. For practitioners and policymakers, the findings suggest that AI should be implemented as part of a broader sustainability strategy rather than as a standalone technological investment. Strengthening governance, investing in organizational capabilities, and improving ESG and AI regulatory frameworks will be essential for achieving sustainable long-term value creation and supporting the objectives of Saudi Vision 2030.