Digital Leadership, AI Integration, and Cyberloafing: Pathways to Sustainable Innovation in SMEs Within Resource-Constrained Economies

Abstract

Sustainable innovation represents both a strategic priority and survival imperative for small- and medium-sized enterprises in resource-constrained economies. While digital transformation offers potential solutions, the synergistic effects of digital leadership, employee behaviors, and emerging technologies remain poorly understood. This study bridges this gap by developing and testing a behavioral-tech leadership framework grounded in the Job Demands-Resources (JD-R) model and Technology Acceptance Model. Analyzing survey data from 400 Iraqi SME employees using partial least squares structural equation modeling, we demonstrate that digital leadership directly enhances sustainable innovation while reducing counterproductive cyberloafing. Crucially, social cyberloafing, when properly managed, emerges as a positive mediator, improving employee well-being and creativity, particularly among mid-career and educated workers. Artificial Intelligence’s integration further amplifies these effects by optimizing operational efficiency and reducing human-resource strain. These findings challenge conventional perspectives by repositioning cyberloafing as a conditional resource within the JD-R framework and provide actionable insights for achieving sustainable innovation even in challenging environments. Practical implications include gender-inclusive digital leadership programs, ethical AI implementation guidelines and restorative cyberloafing policies. The study contributes to United Nations Sustainable Development Goals 8 (decent work), 9 (industry innovation) and 12 (responsible consumption) while highlighting the transformative potential of human-centric digital strategies in resource-constrained contexts.

1. Introduction

Small- and medium-sized enterprises (SMEs) serve as critical engines for achieving the United Nations Sustainable Development Goals (SDGs), particularly in resource-constrained economies where they constitute over 90% of businesses and drive local economic resilience [1]. Recent UN reports emphasize that SME digital transformation directly advances SDG 8 (decent work and economic growth), SDG 9 (industry, innovation and infrastructure) and SDG 12 (responsible consumption and production) through job creation, technological leapfrogging, and sustainable production practices [2]. However, in fragile economies like Iraq, where SMEs contribute 40% of employment yet face systemic barriers [3,4], sustainable innovation has become a survival necessity rather than merely a developmental aspiration. Defined as the creation of long-term value through innovations that advance economic efficiency, environmental stewardship, and social equity [5], sustainable innovation is severely constrained by institutional fragility, infrastructural deficits, and technological underdevelopment [6]. Digital transformation offers a promising solution. Artificial intelligence (AI) can democratize innovation by reducing operational inefficiencies and optimizing resource use [7,8], directly contributing to SDG 12 (responsible consumption and production). When strategically deployed, AI supports lean production and enhances forecasting accuracy while fostering inclusive innovation ecosystems (SDG 9) [9]. However, technology alone is insufficient; success hinges on digital leadership that aligns technological adoption with human and organizational sustainability. Since the adoption of the 2030 Agenda for Sustainable Development in 2015, SMEs have been recognized globally as central actors in advancing SDGs, particularly in fragile economies like Iraq where SMEs provide nearly 40% of employment but struggle with systemic barriers such as institutional fragility and technological underdevelopment [3,5]. These challenges make the study of digital leadership, AI, and employee behavior in SMEs both urgent and timely.

Digital leaders serve as change agents who cultivate digital literacy, promote ethical work systems, and foster psychologically safe environments [10,11]. In SMEs, where leadership is often deeply personal, they play a critical role in shaping sustainable work cultures that drive innovation (SDG 8) [12]. Yet a significant gap remains in understanding how digital leadership interacts with emerging technologies and employee behaviors in high-pressure contexts. This is particularly true for cyberloafing, non-work-related internet use [13], which has traditionally been viewed as counterproductive. Recent applications of the Job Demands-Resources (JD-R) model suggest a more nuanced reality: in resource-scarce environments, moderate cyberloafing may function as a micro-recovery mechanism [14], supporting sustainable cognitive performance when properly managed [15,16]. This potential behavioral resource could contribute to psychological safety and innovation continuity (SDG 8), but its interplay with digital leadership and AI integration remains unexplored in fragile economies. Most existing studies focus on large firms in advanced economies [17], leaving critical gaps in our understanding of how digital leadership empowers sustainable innovation through technological and behavioral pathways, (2) the mediating role of social cyberloafing in this relationship and (3) AI’s moderating effects on these dynamics in resource-constrained settings. This study addresses these gaps through a novel sustainability-driven behavioral-tech leadership model, employing mixed methods to examine Iraqi SMEs. Our contributions are threefold: First, we theoretically extend the JD-R model by repositioning cyberloafing as a conditional resource. Second, we provide methodological innovation through SDG-aligned KPIs that measure innovation frequency, AI efficiency gains, and digital well-being. Third, we offer practical strategies for fostering inclusive digital cultures in fragile economies. This study explicitly distinguishes between counterproductive cyberloafing, which digital leadership reduces, and social cyberloafing, which, when bounded and restorative, is reframed as a conditional resource that can positively mediate innovation outcomes. The central research question guiding this study is: How do digital leadership, AI integration, and social cyberloafing interact to foster sustainable innovation in SMEs operating in resource-constrained economies?

The remainder of the paper is structured as follows. Section 2 presents a comprehensive literature review and theoretical framework, elaborating on sustainable innovation in SMEs, digital leadership, AI integration, and the repositioning of cyberloafing as a behavioral resource. Section 3 details the methodology, including sampling, measurement, and analytical procedures, with an emphasis on assessing sustainability performance through behavioral and technological indicators. Section 4 reports the results, highlighting the mediating role of social cyberloafing and the moderating effect of AI. Section 5 discusses the findings in relation to SDG 8, 9, and 12, emphasizing implications for sustainable business practices, work environments, and innovation infrastructure. Section 6 concludes by affirming that sustainably innovative SMEs are achievable even in constrained environments when digital leadership is purpose-driven, AI is ethically deployed, and employee behaviors are managed with well-being and creativity in mind. It also outlines theoretical and practical implications, and acknowledges limitations such as geographic specificity and demographic homogeneity and suggests future research directions, particularly regarding gender, age, and industry-based variations in digital behavior.

2. Literature Review and Theoretical Framework

This section synthesizes current research on sustainable innovation in SMEs, digital leadership, AI integration and reinterpreted employee cyber-behaviors, positioning them within a unified theoretical framework anchored in the United Nations SDGs. The discussion culminates in the development of a sustainability-driven behavioral-tech leadership model, which informs the study’s hypotheses.

2.1. Sustainable Innovation in SMEs: A Multidimensional Imperative

Sustainable innovation in SMEs integrates environmental, social, and economic objectives into core business practices, acting as a critical driver of systemic change [4,13]. Unlike incremental innovation, it explicitly reconciles profitability with reduced ecological footprints, optimized resource efficiency, and enhanced social equity [4]. Deeply embedded in regional ecosystems, SMEs exhibit unique, context-sensitive innovation pathways that position them as pivotal actors in achieving SDGs 8, 9, and 12 [18]. Studies confirm the tangible benefits of this approach. Kumar et al., for example, found that Indian SMEs adopting circular economy principles achieved a 27% reduction in material costs and an 18% increase in customer satisfaction [19]. Similarly, Viterouli et al. demonstrated that firms incorporating green productivity metrics showed greater resilience during economic shocks, highlighting how sustainability-oriented innovation fosters long-term competitiveness [20].

The convergence of digitalization and sustainability has further unlocked innovative avenues. Digital lean practices like real-time monitoring and predictive maintenance have reduced waste and energy consumption by up to 30% in manufacturing SMEs [21]. Furthermore, IoT-enabled emissions tracking helps align operations with decarbonization targets, bridging regulatory compliance and operational efficiency [22]. Such advancements epitomize the human-centric and sustainable principles of Industry 5.0 [23]. Despite this progress, critical research gaps persist. The social dimensions of innovation such as inclusive employment, employee well-being, and equitable access to digital tools, remain underexplored compared to environmental outcomes [24,25]. The mediating role of digital leadership in technology adoption is also poorly understood [26]. Finally, few studies examine how SMEs leverage AI and automation to foster inclusive innovation ecosystems rather than merely reduce costs [27]. Addressing these gaps, this study advances a holistic framework for sustainable innovation by integrating social and behavioral dimensions with technological and environmental metrics. Employing a mixed-methods approach, we evaluate innovation through triple-bottom-line indicators (economic, environmental, social), ensuring alignment with SDG 9 and SDG 12. By examining the interplay of digital leadership, employee agency, and AI-driven optimization, we offer novel insights into cultivating inclusive, sustainable growth in SMEs within fragile economies.

2.2. Digital Leadership for Sustainable Work Environments: A Strategic Imperative for Sustainable Transformation

Digital leadership has evolved from a technical role into a strategic imperative for organizational sustainability. Modern digital leaders are defined by their ability to foster inclusive cultures, implement ethical governance, and empower employees, moving beyond mere technological adoption [28,29]. In SMEs, where leadership is often centralized, these leaders are critical agents for harmonizing digital transformation with human sustainability and operational efficiency. A core function of digital leadership is advancing digital literacy, which mitigates digital exclusion and promotes equitable participation in the digital economy [30]. Empirical evidence confirms this link: SMEs with digitally literate workforces are more successful in implementing green digital tools [31]. Leaders who prioritize upskilling also report higher employee retention and engagement, particularly among older and marginalized workers, thereby reinforcing social sustainability [32].

Effective digital leaders also design work systems that counteract the psychological toll of digital intensity. Technostress creators—such as information overload, perpetual connectivity, and role ambiguity—are pervasive challenges [13]. Proactive leaders mitigate these through policies like asynchronous communication, structured digital detox periods, and transparent AI governance [33]. For instance, Liu et al. found that SMEs with transformational digital leaders exhibited 35% lower burnout rates and 40% higher job satisfaction [15]. Within the SDG paradigm, sustainability-oriented digital leadership builds long-term resilience. Leaders who embed sustainability into digital agendas enhance organizational adaptability [26]. This is corroborated by AlHares, whose analysis of 420 firms showed that sustainability-driven digital leadership significantly increased innovation output and ESG performance, boosting stakeholder trust and underscoring a clear competitive advantage [34].

Despite its relevance, critical gaps remain. A contextual bias exists, as most studies focus on large corporations, neglecting SMEs in emerging economies [34]. The pathways through which digital leadership influences well-being and innovation are underexplored, and validated measurement tools for sustainability-oriented digital leadership are scarce. To address these gaps, this study introduces a novel measurement scale that integrates dimensions like digital inclusion, well-being advocacy, and ethical AI deployment, adapting items from established scales [35] to offer a comprehensive tool for research and practice. While research on digital leadership addresses inclusion and well-being, its specific influence on social cyberloafing is underexplored. Emerging evidence suggests a strong link: empowering leadership reduces cyberloafing by satisfying psychological needs like autonomy and competence [36]. Similarly, responsible leadership negatively impacts cyberloafing by heightening employees’ felt obligation [15]. Findings from the JD–R model also indicate that job engagement (fostered by supportive leadership) negatively correlates with cyberloafing [37]. Together, this suggests digital leadership can curb social cyberloafing through empowerment, ethical stewardship, and engagement. Thus, we propose:

H1. 

Digital leadership promotes sustainable innovation.

H2. 

Digital leadership is negatively associated with social cyberloafing among employees.

2.3. Social Cyberloafing as a Behavioral Resource and Sustainable Innovation Promoter

Cyberloafing, defined as the non-work-related use of digital tools during working hours [9], has traditionally been framed as counterproductive behavior. However, emerging research reconceptualizes it as a strategic behavioral resource that enhances digital well-being and sustainable performance [38,39]. We adopt a two-dimensional framework for cyberloafing. First, counterproductive cyberloafing is defined as excessive, unregulated internet use for activities like gaming and shopping that undermines productivity [9]. Second, restorative cyberloafing encompasses bounded, brief online breaks that replenish cognitive resources and can support innovation. A primary type of restorative cyberloafing is social cyberloafing characterized by socially interactive activities like chatting with friends on messaging apps that are aligned with organizational norms [9]. Social cyberloafing is defined as bounded, brief, socially interactive online breaks that replenish cognitive resources and can support innovation when aligned with organizational norms [9]. For consistency, the broader term restorative cyberloafing will be used throughout when discussing the general concept, while social cyberloafing will be used when specifically referring to its interactive subtype. The Job Demands-Resources (JD-R) model supports this view. In high-demand digital work environments, job demands deplete energy, while personal resources like brief, intentional social cyberloafing (e.g., social media breaks, news browsing) can act as micro-recovery mechanisms to restore cognitive resources and foster creativity [40]. This is particularly relevant in SMEs, where resource constraints and blurred work–life boundaries intensify stress.

Empirical evidence confirms the benefits of moderated social cyberloafing. Studies link it to improved mood and subsequent task performance [39], accelerated emotional recovery [41], and enhanced creative problem-solving by exposing employees to diverse stimuli that spark ideas for sustainable innovations like eco-design [37]. When bounded by organizational norms, social cyberloafing promotes psychological detachment and serendipitous knowledge exchange, sustaining long-term innovation capacity [38]. This aligns with the human-centric priorities of Industry 5.0 [23]. Forward-thinking SMEs are now implementing digital wellness policies that differentiate restorative breaks from excessive disengagement, thereby advancing SDG 8 by preventing burnout and maintaining innovation momentum [13]. However, the relationship is context-dependent; excessive use undermines productivity, while overly restrictive policies may stifle creativity, underscoring the need for ethical guidelines [42,43]. Prior research lacks longitudinal evidence on cyberloafing’s long-term impact, tests of its mediating role, and exploration of cultural moderators. This study addresses these gaps by adapting established scales [42,44] to distinguish between types of cyberloafing (passive browsing, social interaction, entertainment) and their varied impacts. We propose:

H3. 

Social cyberloafing positively influences sustainable innovation.

H4. 

Digital leadership indirectly promotes sustainable innovation.

2.4. AI Integration and Sustainable Business Practices

The integration of AI has evolved from a competitive advantage to a critical enabler of sustainable transformation for SMEs. Advances in cloud-based platforms and low-code tools have democratized access, allowing resource-constrained firms to leverage AI for significant sustainability gains [27]. This section synthesizes evidence on AI’s dual role in enhancing operational efficiency and human-centric innovation, while addressing key governance challenges.

AI directly contributes to SDG 12 by optimizing resource use and minimizing waste. For example, AI-powered demand forecasting reduces overproduction by 22% in retail SMEs [6], while predictive maintenance cuts equipment downtime by 30% in manufacturing [7]. These operational improvements are bolstered by data-driven decision-making; AI adoption enhances resource allocation and decision efficiency, validating the Technology Acceptance Model in these contexts [45]. For SMEs with flat hierarchies, AI-augmented analytics are particularly valuable for facilitating cross-functional collaboration [46]. AI also enables circular economy practices. It promotes supply chain transparency and environmentally friendly production methods [47], helping SMEs align with regulations like the EU Green Deal while maintaining profitability. This is supported by AI-driven life cycle assessment tools that enable real-time carbon footprint monitoring, which is critical for compliance [48].

Contrary to job displacement fears, strategic AI implementation can enhance human capital. AI capabilities drive product innovation through improved business-IT collaboration [49]. Furthermore, AI reduces cognitive load in decision-making; AI-assisted workflows significantly reduce executive cognitive strain, freeing mental capacity for innovation [50]. These benefits extend to frontline employees, with AI-augmented workers showing 50% greater participation in innovation initiatives due to improved access to data-driven insights [51]. Ethically implemented, AI can thus support SDG 8 by reducing work intensification while enhancing creative problem-solving [52]. Significant risks, however, accompany AI adoption. Algorithmic bias in HR applications can undermine diversity efforts [53], while excessive AI monitoring has been linked to increased compensatory cyberloafing [54]. Transparency is a critical mitigating factor; firms disclosing more AI-related information demonstrate higher liquidity ratios, suggesting stakeholder trust benefits from transparency practices [55]. Implementation gaps persist. SMEs often lack formal AI ethics guidelines, exacerbating risks of unintended consequences [56]. This governance deficit is concerning, as firm size (but not industry) predicts AI disclosure levels, indicating smaller enterprises require targeted support [55]. Three critical gaps motivate this study: (1) the underexplored synergy between digital leadership and AI capabilities, (2) limited evidence on how AI-mediated cognitive load reduction translates to innovation outcomes, and (3) contradictory findings regarding AI’s impact on behaviors like cyberloafing. Addressing these, we propose:

H5. 

AI integration positively moderates the digital leadership–sustainable innovation relationship.

Our methodology extends prior work by developing a multidimensional AI integration scale assessing automation, analytics, and governance, employing NASA-TLX to quantify cognitive load effects, and controlling for firm size effects identified in disclosure studies [55]. This approach provides both theoretical advancement by integrating TAM and JD-R perspectives and practical guidance for SMEs navigating AI adoption for sustainability.

2.5. Conceptual Framework

Building on the integrated review of sustainable innovation, digital leadership, AI integration, and cyberloafing, this study proposes a sustainability-driven behavioral-tech leadership model that positions digital leadership as the catalyst for sustainable innovation in SMEs, mediated by balanced digital behaviors and moderated by AI integration (see Figure 1). Grounded in the JD-R model, the TAM and aligned with the SDG agenda, the model posits that digital leadership fosters sustainable innovation not only through technological empowerment but also through the cultivation of balanced employee behaviors. In our conceptual framework, AI integration is positioned as a moderator, not a mediator. Following Baron and Kenny’s framework [57], a moderator alters the strength or direction of the relationship between an independent variable (digital leadership) and a dependent variable (sustainable innovation) without being part of the causal sequence itself [58]. AI integration functions in this way by conditioning the effect of digital leadership on outcomes (H5), rather than transmitting it. This model advances a novel perspective: sustainable innovation in SMEs is an emergent property of the interaction between technology, leadership, and human behaviors.

By reframing cyberloafing as a resource and AI as an enabler, the framework challenges traditional views of digital distraction. The study’s next section details the methodological procedures for testing the proposed model using Partial Least Squares Structural Equation Modeling (PLS-SEM).

2.6. Theoretical and Practical Contributions

Firstly, this study provides a critical reconceptualization of cyberloafing and extends the JD-R model by repositioning social cyberloafing from a purely negative “demand” or deviant behavior to a conditional “personal resource” that can mediate the path to sustainable innovation. This challenges the deficit-based view prevalent in much of the literature and offers a more nuanced understanding of employee behavior in the digital workplace. Secondly, we introduce a novel sustainability-driven behavioral-tech leadership framework that integrates the JD-R model and the TAM. This unified model provides a comprehensive lens for understanding the complex interplay between leadership (antecedent), employee behavior (mediator), technology (moderator) and sustainable innovation (outcome). Thirdly, by focusing on Iraqi SMEs, a severely under-researched context characterized by institutional fragility and resource scarcity, we enrich the global understanding of digital transformation. Our findings demonstrate that theories developed in advanced economies must be adapted to account for such contextual factors, thereby providing a more nuanced and geographically inclusive view of digital sustainability. Fourthly, this study contributes a validated, multidimensional scale for digital leadership that incorporates critical sustainability dimensions (digital inclusion, well-being advocacy, and ethical AI deployment), filling a gap identified in the literature [34].

The findings translate into actionable strategies for organizations. For practitioners, particularly within SME digital leadership programs, this research provides an evidence-based blueprint for cultivating leaders who can effectively manage technological adoption while safeguarding employee well-being and fostering innovation. Furthermore, the study offers a foundation for developing robust AI ethics guidelines. By empirically linking leadership practices to ethical AI deployment, it equips organizations with the rationale and framework to create policies that ensure technology is used responsibly and inclusively. Finally, our reconceptualization of cyberloafing directly informs the creation of more nuanced and restorative cyberloafing policies. Instead of universally restrictive measures, organizations can develop guidelines that differentiate between detrimental and restorative cyberloafing, allowing for brief, recovery-oriented digital breaks that can ultimately support, rather than hinder, sustainable performance and innovation.

3. Materials and Methods

This study employs a quantitative, cross-sectional research design to examine the behavioral drivers of sustainable innovation in SMEs operating in resource-constrained economies. Grounded in the JD-R model, TAM and anchored in the United Nations SDGs, the methodology is designed to assess how digital leadership, AI integration and employee cyber-behaviors interact to support sustainable innovation.

3.1. Research Design and Rationale

A quantitative research design was employed to empirically examine the hypothesized relationships with statistical rigor while ensuring broader generalizability. The study utilized a structured questionnaire to collect primary data from employees of service-sector SMEs in Northern Iraq. Partial Least Squares Structural Equation Modeling (PLS-SEM). PLS-SEM was selected as the primary analytical technique due to its specific suitability for this study’s objectives and context. Firstly, PLS-SEM is robust in evaluating complex models featuring multiple latent variables, indirect effects (mediation), and interaction effects (moderation), which are central to our proposed behavioral-tech leadership framework [59]. Secondly, the technique emphasizes predictive relevance and theory development, making it ideal for exploratory research in emerging fields like sustainability-driven digital transformation [60]. Thirdly, PLS-SEM produces reliable estimates with non-normally distributed data and is effective with moderate sample sizes [59,60], conditions often encountered in SME-based research in developing economies [61]. Methodologically, this approach supports SDG 9 (Industry, Innovation, and Infrastructure) by generating data-driven insights into the digital innovation ecosystems within SMEs.

3.2. Sample and Data Collection

The study sample comprised 400 employees from service sector SMEs in Northern Iraq, a region characterized by economic fragility, institutional volatility, limited financial access, skilled labor shortages, and underdeveloped technological infrastructure. A priori power analysis (G*Power 3.1) indicated that a minimum sample size of 129 was required for detecting medium effects (f² = 0.15) with power = 0.95 at α = 0.05. The achieved sample of 400 thus exceeds the recommended threshold, ensuring statistical robustness. This study’s context provides an appropriate setting for investigating sustainable innovation under constrained conditions [5]. SME selection was based on two key criteria: active utilization of digital tools including cloud platforms, communication software, and AI-powered customer service solutions, along with demonstrated relevance to local economic sustainability. The service sector was specifically targeted due to its dominant position in the regional SME landscape and its recognized potential for fostering intrapreneurial behavior, which serves as a significant driver of sustainable innovation [5]. The service-sector SMEs included IT services, retail, finance, hospitality, and education subsectors. Inclusion criteria required SMEs with 10–250 employees actively using digital tools such as cloud platforms and AI-powered customer engagement. Micro-enterprises and informal businesses were excluded. The participant pool included employees across diverse functional roles encompassing administrative, managerial and technical positions, ensuring a comprehensive representation of organizational dynamics. Data collection employed a mixed-method approach utilizing both online and paper-based structured questionnaires, distributed through established local business networks including chambers of commerce. The survey was distributed to 600 employees and 400 questionnaires were completed, yielding a 66.67% response rate. Potential bias may arise from overrepresentation of digitally literate and higher-educated respondents, which may limit generalizability to traditional SMEs.

3.3. Measurement Instruments

This study relied on structured survey data collected from SME employees in Northern Iraq, ensuring both digital tool adoption and sustainability relevance. Data collection was conducted using paper-based questionnaires distributed via local chambers of commerce (see Supplementary Materials—Questionnaire), followed by rigorous data cleaning, pilot testing, and validation procedures. All constructs were operationalized using multi-item Likert-scale measures (1 = Strongly Disagree to 5 = Strongly Agree) adapted from established scales in prior literature. All constructs were operationalized using multi-item Likert-scale measures adapted from established scales in prior literature. Responses were captured on a 5-point Likert scale (1 = Strongly Disagree to 5 = Strongly Agree), ensuring consistency in measurement. The selected instruments align with performance indicators relevant to the SDGs, reinforcing the study’s theoretical grounding in sustainability. Digital Leadership was assessed using a 12-item scale adapted from Zeike et al., measuring key dimensions such as digital competency, transformational support, and ethical technology adoption [35]. A sample item includes: “My supervisor encourages the use of digital tools to improve work processes.” This construct aligns with SDG 8, as it captures leadership practices that promote inclusive and resilient work environments. AI Integration was evaluated via an 8-item scale developed by Shaikh and Afshan [62]. Four additional items were further included to, examine the extent to which artificial intelligence is embedded in organizational processes (e.g., automation, data analytics and customer engagement). A representative item states: “Our organization uses AI systems to automate routine tasks.” This measure serves as a proxy for SDG 9 and SDG 12, reflecting technological advancement and operational sustainability.

Social Cyberloafing was measured using a 9-item scale derived from Andreassen, Torsheim and Pallesen [42] and Wu et al. [44], distinguishing between recreational and socially motivated non-work internet use. An example item reads: “I occasionally browse social media during work hours to stay connected with friends or family.” This construct addresses the psychosocial aspects of digital sustainability, highlighting how micro-break behaviors may influence long-term employee well-being (SDG 8).

Sustainable innovation was measured using the 15-item Sustainable Innovation Scale developed by Hansen, Grosse-Dunker and Reichwald [63], which assesses three key dimensions: process innovation (e.g., “Our company improves operational efficiency while reducing environmental impact.”), product and service innovation (e.g., “We develop offerings that address social or environmental needs.”) and strategic innovation (e.g., “Our long-term business model integrates sustainability principles.”). This scale was chosen because it explicitly links innovation to sustainability outcomes, aligning with SDG 9. It has been validated in multiple organizational contexts [63,64], ensuring reliability. Moreover, it captures both incremental and radical sustainable innovations, providing a holistic assessment.

To mitigate potential confounding effects, control variables, including industry type, firm size, employee tenure, and education level were incorporated into the analysis.

The survey instrument was professionally translated into Arabic and Kurdish to ensure linguistic accuracy and cultural appropriateness. A pilot test was conducted with 30 SME employees to assess the clarity, reliability, and validity of the scales. The pilot study yielded Cronbach’s alpha and composite reliability values exceeding the 0.7 threshold, and AVE values greater than 0.50, confirming satisfactory scale reliability [59]. Minor adjustments were made to wording based on pilot feedback. The research protocol received formal ethical approval from the Institutional Review Board of Cyprus International University (Protocol Code: EKK23-24/010/04, Date: 16 June 2025). Written informed consent was obtained from all participants prior to their involvement in the study. Data anonymity and confidentiality were guaranteed throughout the research process.

3.4. Data Analysis Procedure

Data cleaning included screening for incomplete responses, leading to the removal of 13 cases. Outliers were detected using Mahalanobis distance (p < 0.001 threshold) and robust z-scores. The remaining dataset contained less than 2% missing values, which were imputed using mean substitution. The analysis was then conducted in SmartPLS 4.0. A pilot test was conducted with 30 SME employees to assess the clarity, reliability, and validity of the scales. The data analysis was conducted in two phases using SmartPLS 4.0:

First, the measurement model was assessed to ensure reliability, validity, and unidimensionality. Internal consistency reliability was verified through Cronbach’s alpha (α) and composite reliability (CR), with thresholds exceeding 0.70 to confirm scale robustness [59]. Convergent validity was established using the average variance extracted (AVE), requiring a minimum value of 0.50 [59], while discriminant validity was tested via the Fornell–Larcker criterion and the heterotrait–monotrait (HTMT) ratio, with values below 0.85 indicating sufficient discriminant validity [59]. Additionally, unidimensionality and the absence of multicollinearity were confirmed through variance inflation factor (VIF) analysis, where values below 3.0 suggested no significant collinearity issues.

Subsequently, the structural model was evaluated to test hypothesized relationships. Direct, indirect (mediation), and moderating (interaction) effects were examined using a bootstrapping procedure with 5000 subsamples to ensure statistical robustness. Model fit was assessed through multiple indices: the standardized root mean square residual (SRMR) (<0.08) [59,60,65,66], normed fit index (NFI) (>0.95), and a significant chi-square (p < 0.01). Predictive relevance was further confirmed using R² (explanatory power) and f² (effect size) metrics. To reinforce fit assessment, d_ULS (squared Euclidean distance) and d_G (geodesic distance) were employed, with satisfactory fit indicated by values below their respective upper confidence intervals [59,66]. Common method bias was mitigated procedurally and confirmed statistically through Harman’s test (32.7% variance) and marker variable analysis (r = 0.08, p > 0.10).

This study extends conventional mediation analysis by incorporating AI integration as a moderator within the leadership–cyberloafing–innovation pathway (H4), thereby advancing a multidimensional sustainability perspective. Apart from this analysis contributing to a human-centric framework for digital transformation in SMEs in fragile economies, the findings offer theoretical advancements in sustainable organizational behavior while providing practically actionable insights for policymakers and business leaders seeking to balance technological adoption with workforce well-being.

4. Results

4.1. Demographic Analysis

The demographic data reveal a workforce characterized by relative youth, a pronounced male majority, and high educational attainment factors that position respondents favorably to engage with digital transformation and sustainable innovation initiatives. As illustrated in

Table 1. Demographic analysis.

Item	Category	Frequency	Percentage
Gender	Male	259	64.75%
	Female	141	35.25%
	Total	400	100%
Age	18–25	76	19.00%
	26–35	94	23.35%
	36–45	125	31.25%
	46–55	55	13.75%
	56–65	36	9.00%
	66+	14	3.50%
	Total	400	100%
Educational Level	Diploma	86	21.50%
	Associate’s degree	64	16.00%
	Bachelor’s Degree	102	25.50%
	Master’s Degree	122	30.50%
	Doctorate Degree	26	6.50%
	Total	400	100%

, male respondents significantly outnumber their female counterparts (64.75% vs. 35.25%). Given that digital leadership and AI integration are often shaped by organizational culture and access to technical roles, this gender disparity underscores the need for targeted policies and further research to foster inclusive digital upskilling and leadership development for women in SMEs. Such efforts would align with SDG 5, which advocates for gender equality in professional and technological domains.

Regarding age distribution, the majority of respondents (74.6%) fall within the 26–45 age range, with the largest subgroup concentrated in the 36–45 bracket (31.25%). This suggests a workforce that is not only digitally literate but also sufficiently experienced to assume core operational and managerial responsibilities within SMEs. However, the underrepresentation of younger employees (18–25: 19%) and older workers (56+: 12.5%; 66+: 3.5%) may constrain the study’s ability to capture intergenerational variations in digital behavior, AI adoption, and cyberloafing tendencies. The marginal presence of employees aged 66 and above further highlights the persistent exclusion of older workers from digital transformation initiatives.

Educational attainment among respondents is notably high, with over half holding postgraduate qualifications: 30.5% possess a master’s degree and 6.5% have obtained a Doctorate. An additional 25.5% hold bachelor’s degrees, reinforcing the presence of a cognitively skilled workforce capable of navigating complex digital tools, AI systems, and innovation processes. This trend implies that the surveyed SMEs may predominantly operate within knowledge-intensive or technology-driven sectors, thereby enhancing the relevance of the study’s findings on AI-driven efficiency and digital leadership. Meanwhile, the inclusion of employees with diplomas (21.5%) and associate degrees (16%) reflects workforce diversity, particularly in technical and operational roles where digital literacy and well-being practices like cyberloafing as a micro-recovery mechanism remain critical.

These findings emphasize the necessity of designing digital leadership strategies and AI policies that are inclusive, age-sensitive, and gender-responsive, ensuring that technological advancements in SMEs contribute to equitable and sustainable development rather than exacerbating existing disparities. While the demographic profile strengthens the study’s focus on balanced digital behaviors and ethical leadership as catalysts for inclusive sustainability, the underrepresentation of women and older employees raises concerns regarding digital equity. Addressing these gaps is imperative for fostering sustainable work environments that align with the broader objectives of the SDGs.

4.2. Factor Analysis

The exploratory factor analysis (EFA) with varimax rotation validated the measurement model’s construct validity, revealing four distinct factors corresponding to the study’s key constructs: AI Integration, Digital Leadership, Intrapreneurial Behavior, and Social Cyberloafing. All items demonstrated strong convergent validity as shown in

Table 2. Outer loading of variables.

	Artificial Intelligence	Digital Leadership	Sustainable Innovation	Social Cyberloafing
AI3	0.789
AI5	0.794
AI7	0.766
AI8	0.776
AI9	0.795
AI10	0.792
AI11	0.747
AI12	0.759
DL1		0.853
DL2		0.831
DL4		0.809
DL6		0.843
DL9		0.850
DL11		0.589
SI1			0.854
SIB2			0.715
SI3			0.700
SI4			0.788
SI5			0.798
SI7			0.718
SI9			0.741
SI13			0.817
SI15			0.799
SCL1				0.818
SCL2				0.785
SCL3				0.797
SCL4				0.766
SCL5				0.827
SCL8				0.711
SCL9				0.816

, with loadings exceeding the 0.50 threshold [59,60]. The AI Integration items (AI) loaded exclusively on Factor 1 (0.747–0.795), confirming robust dimensionality. Digital Leadership items (DL) formed Factor 2, with loadings from 0.589–0.853; DL11’s lower loading (0.589) was retained due to theoretical importance and scale reliability (α > 0.85). Sustainable Innovation items (SI) loaded cleanly on Factor 3 (0.700–0.854), confirming their unidimensionality and robustness in capturing employee-driven in-novation within organizations, while Social Cyberloafing items (SC) formed Factor 4 (0.711–0.827). The absence of cross-loadings demonstrates excellent discriminant validity, supporting the theoretical distinction among constructs in sustainable innovation research for SMEs. These results justify the scales’ use in subsequent structural equation modeling and confirm the appropriateness of the measurement instrument for testing the hypothesized relationships in the study.

4.3. Reliability and Validity

The measurement model assessment, presented in

Table 3. Reliability and validity.

	Cronbach’s Alpha	Composite Reliability	Rho-c	AVE
Artificial intelligence	0.773	0.810	0.835	0.703
Digital leadership	0.884	0.885	0.914	0.642
Sustainable innovation	0.733	0.763	0.807	0.828
Social cyberloafing	0.880	0.896	0.908	0.588

, demonstrates that all constructs satisfy or surpass established psychometric thresholds for PLS-SEM (Hair et al., 2022) [59,60]. Reliability analyses reveal strong internal consistency, with Cronbach’s alpha coefficients ranging from 0.733 (Sustainable innovation, SI) to 0.884 (digital leadership, DL), all exceeding the conventional 0.70 benchmark [59,60]. Composite reliability (CR) scores, which are preferred for reflective constructs due to their lower sensitivity to item number, range from 0.763 (SI) to 0.896 (self-control, SC), comfortably surpassing the 0.70 threshold [59,60].

Additional confirmation comes from Dijkstra–Henseler’s rho_A values, which range from 0.807 (SI) to 0.914 (DL), further substantiating the measurement model’s reliability [59]. Convergent validity assessments show AVE values between 0.588 (SC) and 0.828 (SI). While SC’s AVE (0.588) slightly falls below the ideal 0.60 threshold, it remains above the acceptable 0.50 minimum [59,60], with all other constructs demonstrating strong convergent validity by explaining over 50% of their indicators’ variance. The exceptional AVE for SI (0.828) indicates particularly robust construct clarity and indicator coherence. These comprehensive results confirm that all constructs meet rigorous psychometric standards for reliability and validity, thereby supporting their use in subsequent hypothesis testing. The findings significantly strengthen confidence in the measurement instrument’s robustness for examining sustainable innovation dynamics in SMEs, particularly in digital transformation contexts.

4.4. Discriminant Validity

As demonstrated in

Table 4. Fornell–Larcker criterion.

	Artificial Intelligence	Digital Leadership	Sustainable Innovation	Social Cyberloafing
Artificial intelligence	0.634
Digital leadership	0.567	0.801
Sustainable innovation	0.693	0.574	0.573
Social cyberloafing	0.624	0.753	0.554	0.767

Note: Diagonal values represent the square root of AVE from previous analysis: AI = √0.7026 ≈ 0.838; IB = √0.8281 ≈ 0.910; DL = √0.6420 ≈ 0.801; SC = √0.5884 ≈ 0.767.

, all constructs satisfy the Fornell–Larcker criterion, with the square root of the average variance extracted (√AVE) for each construct exceeding its highest correlation with other constructs [59]. This provides robust evidence of discriminant validity, confirming that each construct represents a distinct latent variable within the theoretical framework. Of particular interest is the strong correlation between digital leadership and social cyberloafing (r = 0.753), which aligns with the theoretical proposition that supportive digital leadership fosters balanced digital behaviors, including restorative cyberloafing [38].

While this correlation is substantial, discriminant validity remains intact as it falls below the √AVE of SC (0.767), underscoring the importance of maintaining theoretical distinctions between closely related constructs. These results not only validate the psychometric properties of the measurement model but also support its application in subsequent PLS-SEM analysis. The findings are particularly relevant for studies examining the complex interplay between digital leadership, artificial intelligence adoption, employee behavior and sustainable innovation in SMEs, contributing to both academic discourse and practical applications in organizational sustainability.

4.5. Model Fit

As presented in

Table 5. Model Fit.

Fit index	Acceptable Values	Estimated Model	Reference
SRMR	≤0.080	0.061	[53,54]
NFI	≥0.900	0.972
χ2/df	≥0.9 p < 0.01 ***	1.989 p < 0.01 ***
d_ULS	≤UCI	2.018
d_G	≤UCI	1.345

χ² means Chi-square and *** = significant at 0.01 level.

, the model demonstrates strong goodness-of-fit across multiple indices, aligning with established benchmarks for SEM. The SRMR falls below the conservative threshold of 0.08 [59], suggesting excellent model fit. Further supporting this, the normed fit index (NFI = 0.972) approaches the ideal value of 1, reinforcing the model’s robustness [60]. The chi-square statistic (χ² = 1.989, p < 0.01) further corroborates the model’s statistical adequacy, with significance at the 1% level. Additionally, the discrepancy measures, d_ULS and d_G lie beneath their respective 95% upper confidence interval (UCI) bounds, as prescribed by Hair et al. [59,60]. Collectively, these indices provide convergent evidence that the hypothesized model exhibits a good fit to the empirical data. The consistency of these fit indices not only validates the model’s statistical reliability but also underscores its alignment with the underlying theoretical framework. Such robustness is critical for ensuring the credibility of subsequent hypothesis testing and interpretive analyses, particularly in sustainability research where model misspecification could obscure key relationships. Building upon these empirical insights, the subsequent section critically examines the path analysis results to elucidate their theoretical and practical implications within the context of sustainability.

4.6. Path Analysis

The path analysis results obtained from the Smart PLS-SEM analysis are shown in

Table 6. Path analysis without demographic moderation effects.

Hypothesis	Path	β	T Statistics	p Val.	R2	f2	95% Bootstrap CI (L,U)	Decision
H1	DL → SI	0.269	4.180	0.000	0.072	0.101	[0.182, 0.354]	Supported
H2	DL → SC	−0.412	6.732	0.000	0.170	0.205	[−0.523, 0.301]	Supported
H3	SC → SI	0.148	2.987	0.003	0.022	0.132	[0.052, 0.224]	Supported
	Mediation-Moderation Analysis
H4	DL → SC → SI	0.061	2.455	0.000	0.088	0.202	[0.108, 0.104]	Partial Med.
H5	AI → DL → SI	0.183	3.621	0.000	0.134	0.198	[0.082, 0.284]	Supported

. According to the findings, DL exhibited a significant positive effect on sustainable innovation (β = 0.269, p < 0.001), explaining 7.2% of variance (R² = 0.072). The effect size (f² = 0.101) indicates a moderate impact, supporting H₁. Contrary to traditional views, DL reduced counterproductive cyberloafing (β = −0.412, p < 0.001), supporting H₂ and aligning with JD-R theory. This suggests DL fosters self-regulated digital behaviors. The results also show that SC positively influenced sustainable innovation (β = 0.148, p = 0.003), supporting its reframing as a recovery mechanism (H₃). The indirect effect of DL→SC→SI (β = 0.061, p = 0.014) confirmed partial mediation (H₄), per Preacher and Hayes criteria [58]. AI strengthened the DL–Innovation relationship (β = 0.183, p < 0.001), with a significant interaction effect (ΔR² = 0.062), supporting H₅. This aligns with the TAM, showing that AI enhances efficiency and reduces cognitive strain.

The R² values indicate varying levels of explanatory power across the models. Digital leadership accounts for a small but meaningful share of variance in sustainable innovation (R² = 0.072), while it explains a larger portion of variance in cyberloafing behaviors (R² = 0.205). Cyberloafing alone contributes only modestly to innovation outcomes (R² = 0.022), though the mediation model slightly improves explanatory power (R² = 0.088). Importantly, AI integration strengthens the model further, with a notable increase in explained variance (ΔR² = 0.062). As shown in Table 6, effect sizes (Cohen’s f²) range from small to moderate (0.101–0.205), indicating that digital leadership has a stronger impact on reducing cyberloafing (f² = 0.205) than directly on sustainable innovation (f² = 0.101). These values reinforce the significance of cyberloafing as a behavioral mediator. After incorporating demographic moderation effects,

Table 7. Demographic moderation effects.

Moderator	Category	DL → SI (H1)	DL → SC (H2)	SC → SI (H3)	DL → SC → SI (H4 Mediation)	AI Moderation (H5)
Gender	Male (64.75%)	0.225 * [0.131, 0.319]	−0.388 *** [−0.497, −0.279]	0.159 * [0.022, 0.124]	0.061 * [0.018, 0.104]	0.171 ** [0.062, 0.280]
	Female (35.25%)	0.317 *** [0.211, 0.423]	−0.451 *** [−0.562, −0.340]	0.092 [−0.012, 0.054]	0.042 [−0.005, 0.089]	0.201 *** [0.097, 0.305]
Age	18–25 (19.00%)	0.187 * [0.041, 0.333]	−0.352 ** [−0.498, −0.206]	0.121 [−0.035, 0.277]	0.043 [−0.012, 0.098]	0.142 * [0.018, 0.266]
	26–35 (23.35%)	0.254 ** [0.122, 0.386]	−0.401 *** [−0.533, −0.269]	0.148 * [0.026, 0.270]	0.059 * [0.011, 0.107]	0.241 *** [0.132, 0.350]
	36–45 (31.25%)	0.283 *** [0.175, 0.391]	−0.423 *** [−0.531, −0.315]	0.181 ** [0.075, 0.287]	0.077 ** [0.025, 0.129]	0.219 *** [0.123, 0.315]
	46+ (26.25%)	0.210 * [0.082, 0.338]	−0.372 *** [−0.500, −0.244]	0.097 [−0.031, 0.225]	0.036 [−0.019, 0.091]	0.082 [−0.035, 0.199]
Education	Diploma (21.50%)	0.198 * [0.042, 0.354]	−0.361 ** [−0.517, −0.205]	−0.112 * [−0.208, −0.016]	−0.040 * [−0.078, −0.002]	0.148 * [0.042, 0.254]
	Bachelor’s (25.50%)	0.243 ** [0.115, 0.371]	−0.392 *** [−0.520, −0.264]	0.133 * [0.005, 0.261]	0.052 * [0.003, 0.101]	0.195 ** [0.083, 0.307]
	Master’s+ (37.00%)	0.302 *** [0.206, 0.398]	−0.441 *** [−0.537, −0.345]	0.167 ** [0.071, 0.263]	0.074 ** [0.028, 0.120]	0.261 *** [0.153, 0.369]

Key: β coefficients shown with 95% bias-corrected bootstrap CIs in brackets. Significance: *= p < 0.05, ** = p < 0.01, *** = p < 0.001.

reveals that digital leadership’s impact on innovation (H1) is 28.7% stronger for women (β = 0.317) than men (β = 0.225), with a significant group difference (Δβ = +0.092, p = 0.038). Conversely, cyberloafing’s mediation (H3) is only significant for males and mid-career employees (36–45 age group), supporting its context-dependent role as a recovery mechanism. Age and education further shaped these relationships. Among age groups, mid-career employees (36–45) showed the strongest leadership and mediation effects, while younger employees (18–25) benefited more from AI support than from cyberloafing. The 26–35 group displayed balanced gains across DL, SC, and AI, whereas older employees (46+) showed weaker effects overall. In terms of education, diploma holders experienced cyberloafing as a hindrance rather than a resource, while bachelor’s degree holders demonstrated balanced positive effects. Employees with a master’s or higher benefitted the most, as both SC and AI strongly amplified the innovation outcomes of digital leadership.

5. Discussion

This study provides critical insights into how digital leadership, AI integration, and social cyberloafing collectively drive sustainable innovation in resource-constrained SMEs, with significant implications for achieving the United Nations SDGs. Grounded in the JD-R model and TAM, our findings highlight both the transformative potential and contextual nuances of digital transformation in fragile economies like Iraq. Our findings must be interpreted within Iraq’s unique post-conflict context, where SMEs face institutional fragility, infrastructural deficits, and skilled labor shortages [5].

Unlike studies in stable economies [6], our analysis confirms H1, demonstrating that digital leadership significantly enhances sustainable innovation (β = 0.269, p < 0.001, R² = 0.072, f² = 0.101). This aligns with Ossiannilsson’s [29] research positioning digital leaders as catalysts for innovation by promoting digital literacy, ethical AI use, and inclusive decision-making. The effect was particularly pronounced among women (β = 0.317 vs. β = 0.225 for men), supporting Nambisan et al.’s [26] findings on inclusive leadership’s role in bridging gender disparities in technology adoption. This gender differential suggests that digital leadership fosters psychological safety and participatory decision-making, confirming JD-R theory’s emphasis on leadership as a resource provider. Beyond individual benefits, these results indicate that gender-inclusive digital leadership can act as a structural enabler of SDG 5 (gender equality), while reinforcing SDG 8 through fairer work environments, as highlighted by AlHares [34]. In fragile economies, where leadership is often personalized and hierarchical, this finding highlights the theoretical need to adapt leadership frameworks to contexts of institutional fragility.

Contrary to traditional views of cyberloafing as purely counterproductive, our results support H2 by showing digital leadership significantly reduces social cyberloafing (β = −0.412, p < 0.001, R² = 0.205, f² = 0.205). This aligns with Bakker et al.’s [40] JD-R model emphasis on resource conservation and qualifies prior work by Abbas et al. [39] and Grobelny et al. [41], who linked moderate cyberloafing to improved task performance and emotional recovery. However, unlike earlier research in stable contexts where cyberloafing was treated uniformly [12], our findings reveal that leadership practices can actively reframe digital behaviors in fragile economies. We found that cyberloafing operates as a conditional resource, enhancing innovation for highly educated employees (β = 0.167) through micro-recovery and serendipitous learning [37], while showing negative effects for less-educated groups (β = −0.112) likely due to unstructured use. These results refine the JD-R model by demonstrating that cyberloafing’s benefits depend on digital literacy [38], suggesting SMEs should implement SDG 8-aligned “restorative cyberloafing” policies to balance mental health (SDG 3) and creativity. These findings clarify that while digital leadership effectively reduces counterproductive cyberloafing, restorative social cyberloafing can act as a positive mediator of innovation under specific conditions, thereby reconciling the dual perspectives on cyber-loafing.

The analysis confirms H3, with social cyberloafing positively influencing sustainable innovation (β = 0.148, p = 0.003, R² = 0.022, f² = 0.132). This supports Abbas et al.’s [39] and Guo et al.’s [37] reconceptualization of cyberloafing as a strategic micro-recovery mechanism. However, multigroup analysis reveals critical demographic contingencies: employees with advanced degrees (Master’s+) showed substantial innovation gains (β = 0.167, 95% CI [0.071, 0.263]), while vocationally trained workers exhibited negative effects (β = −0.112, 95% CI [−0.208, −0.016]). This pattern extends Guo et al.’s [37] cognitive load theory by showing that innovation outcomes from cyberloafing depend not only on recovery but also on structured digital literacy. Theoretically, this advances JD-R by embedding skill level as a determinant of how resources are utilized. Practically, these findings underscore the need to integrate digital wellness training into vocational curricula (SDG 4) while addressing the environmental costs of unproductive digital behaviors (SDG 12) [11].

Supporting H4, we found that digital leadership indirectly enhances innovation through cyberloafing (β = 0.061, p = 0.000, R² = 0.088, f² = 0.202). This partial mediation integrates Peng et al.’s [36] work on empowering leadership with Wu et al.’s [44] findings about cyberloafing as a coping strategy, while addressing gaps in AlHares’s [34] GMM analysis. By revealing that restorative breaks explain part of the leadership–innovation link, our results extend JD-R’s resource pathway logic [14]. The mediating role varied demographically, serving as an innovation catalyst for males and mid-career employees (36–45 years: β = 0.061–0.077). This suggests that recovery-oriented behaviors are not equally effective across groups but instead interact with career stage and gender, adding nuance to theories of resource utilization. For SMEs, this highlights the importance of tailoring digital leadership strategies to workforce composition, ensuring restorative mechanisms benefit diverse employee groups.

Finally, AI integration significantly moderates the leadership–innovation relationship (β = 0.183, p < 0.001, ΔR² = 0.062), supporting H5. This aligns with Gao et al.’s [49] findings on AI’s role in reducing cognitive load but reveals disparities: employees aged 26–35 showed the strongest benefits (β = 0.241), while those over 46 derived minimal gains (β = 0.082), echoing Möller et al.’s [8] concerns about age-based digital exclusion. Similarly, AI’s impact was 37% stronger for advanced degree holders (Master’s+: β = 0.261) than vocational-trained employees (Diploma: β = 0.148), reinforcing Gao et al.’s [49] argument that human capital determines AI’s value. These findings advance the TAM by demonstrating that perceived usefulness and ease of use are not uniform but conditional upon employee demographics. Theoretically, this underscores the importance of embedding human capital considerations into models of technology acceptance. Practically, while AI adoption can reduce inefficiencies and overproduction (22% [3]), achieving SDG 12 requires inclusive upskilling and training initiatives to prevent digital exclusion in fragile economies. Nonetheless, the generalizability of these findings must be interpreted with caution given sampling limitations. While the sample was balanced across age and education groups, gender distribution was uneven, with women slightly overrepresented in some subsamples. This may have accentuated gender-specific effects, such as the stronger influence of digital leadership among female employees. In addition, the geographic focus on SMEs in Iraq means that results may be shaped by post-conflict institutional fragility, cultural norms, and sectoral structures unique to this setting. Future research should test the model across more diverse samples and regional contexts to assess the robustness of these patterns. With these caveats in mind, the study nonetheless offers several important theoretical and practical contributions.

Taken together, these findings extend prior research by demonstrating that digital leadership not only reduces counterproductive behaviors but also redefines social cyberloafing as a conditional resource that enhances innovation under specific conditions. This challenges Lim’s deviance-based framing [12] and aligns more closely with Abbas, Dabic, and Halaszovich [39], who emphasize its restorative potential. By positioning AI as a moderator, this study advances TAM-based perspectives by illustrating how technology alleviates cognitive load and amplifies leadership’s impact on innovation. More broadly, the results show that both JD-R and the TAM must be contextualized when applied in fragile economies, where institutional fragility and skill disparities shape outcomes in unique ways. For SMEs, these insights provide actionable guidance: formalize restorative cyberloafing practices, invest in targeted digital training, and tailor AI adoption strategies to different workforce demographics. These contributions strengthen the alignment of SME practices with SDGs 8, 9 and 12.

6. Conclusions

This study explored how digital leadership, AI integration, and social cyberloafing interact to influence sustainable innovation in SMEs operating under fragile economic conditions. The analysis confirmed five hypotheses: digital leadership significantly promotes innovation, reduces counterproductive cyberloafing, and indirectly enhances creativity through restorative cyberloafing; social cyberloafing positively influences innovation under certain conditions; and AI integration amplifies the leadership–innovation relationship. Digital leadership is not merely about technological adoption but about creating a culture of empowerment and well-being and AI is not a replacement for human capital but a tool to augment it, reducing strain and enhancing efficiency. Finally, employee behaviors like cyberloafing should not be policed as deviance but understood as potential recovery mechanisms that, when guided by supportive leadership, can fuel creativity and innovation. These findings affirm that sustainable innovation is an emergent property of the interaction between human-centric leadership, enabling technology and balanced work practices.

6.1. Policy Implications

The findings of this study carry practical implications for multiple stakeholder groups involved in SME development in fragile economies.

For SME managers, the results suggest that investing in digital leadership training can directly enhance innovation capacity. Managers should adopt inclusive leadership styles that promote digital literacy and psychological safety, especially for women and vocationally trained employees who may face digital skill gaps. Concrete steps include introducing structured “ethical cyberloafing” and “restorative cyberloafing” practices such as short, scheduled online breaks and providing targeted digital wellness workshops. These measures can reduce burnout, increase creativity, and align workplace practices with SDG 8 (decent work) and SDG 5 (gender equality). When adopting AI, pair technology investment with comprehensive upskilling programs to ensure all employees, regardless of age or education level, can benefit.

For government agencies, the evidence highlights the need for national policies that support digital upskilling programs and vocational education reform. Governments could, for example, integrate digital wellness and AI literacy modules into vocational curricula, ensuring that less-educated workers can benefit from technological change. In addition, offering tax incentives or subsidies for SMEs adopting responsible AI tools would encourage equitable and sustainable digital transformation, directly supporting SDGs 4, 8, and 9.

For international donors and development partners, the findings underscore the importance of designing funding programs that address both technological infrastructure and human capacity building. Donors could support initiatives that provide SMEs with affordable access to AI tools while also financing leadership and digital literacy training tailored to fragile contexts. Such programs would ensure that investments in technology also yield inclusive innovation outcomes, thereby advancing SDG 12 (responsible consumption and production) alongside broader development goals.

By tailoring these implications to the needs of SME managers, policymakers, and international donors, this study moves beyond abstract recommendations to provide actionable strategies for building resilient, innovative, and sustainable SME ecosystems in fragile economies.

6.2. Limitations and Future Directions

While this study advances understanding of digital leadership in resource-constrained SMEs, several limitations warrant caution. First, the cross-sectional design precludes causal claims; longitudinal tracking could reveal how AI’s moderating role evolves with prolonged use. Second, despite Iraq’s relevance as a fragile economy, cultural specificities like centralized decision-making may limit generalizability to contexts with flatter hierarchies. Third, demographic underrepresentation of women and older workers risks sampling bias, potentially overstating digital literacy effects. Fourth, self-reported data, especially for cyberloafing, may suffer from common method bias; future studies could integrate passive digital traces. Finally, macro-level barriers, especially Iraq’s infrastructural deficits, were unmeasured but likely interact with leadership and AI adoption. Future research should address these limitations through longitudinal or experimental designs to establish causality, while cross-cultural comparisons like Iraq vs. Scandinavia could test the model’s boundary conditions. Demographic diversity must be prioritized, with intersectional analyses of how age, gender, and education interact with digital behaviors. Objective measures from digital phenotyping for cyberloafing to ESG metrics for sustainability would reduce reliance on self-reports. Finally, macro-level studies should examine how infrastructure, policies, and AI subtypes, such as generative vs. analytical, shape these dynamics.