Generative AI Adoption in B2B Firms: Ethical Governance, Innovation Capabilities, and Long-Term Competitive Performance

You’re listening to “Generative AI Adoption in B2B Firms: Ethical Governance, Innovation Capabilities, and Long-Term Competitive Performance,” by Michele Alves and colleagues. Published in Systems on April 8, 2026.

Abstract.

The rapid diffusion of generative artificial intelligence (GenAI) is reshaping organisational systems and digital transformation strategies, yet it remains unclear which organisational conditions are associated with long-term competitive performance in business-to-business (B2B) contexts. This study adopts a systems-informed perspective and examines how ethical governance, environmental dynamism, exploratory and exploitative innovation, and GenAI adoption are associated with long-term competitive performance in B2B firms. Using survey data from 104 Portuguese B2B managers and Partial Least Squares Struc- tural Equation Modelling (PLS-SEM), the findings show that ethical governance is the strongest organisational correlate of long-term competitive performance, underscoring the central role of governance structures in responsible GenAI use. GenAI adoption is positively associated with performance, but its role is complementary rather than dominant. Exploratory innovation does not show a significant direct association with performance; instead, its association with performance operates through GenAI adoption in the esti- mated model, suggesting that experimentation becomes more performance-relevant when translated into digitally enabled routines. In contrast, exploitative innovation is directly associated with performance through incremental efficiency mechanisms. These findings challenge technology-deterministic assumptions and suggest that long-term competitive performance in B2B firms is more closely associated with the organisational alignment of governance structures, innovation capabilities, and GenAI adoption than with technology adoption alone.

1. Introduction.

The accelerating diffusion of generative artificial intelligence (GenAI) is fundamentally reshaping digital transformation strategies across industries.

From a systems-informed perspective, organisations can be conceptualised as socio-technical systems in which governance structures, innovation capabilities and AI-based technologies are interdependent and jointly shape performance outcomes. This view shifts the focus from isolated technology–performance links towards the configuration of organisational subsystems and their interdependencies.

Copyright: © 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.

AI-driven systems are increasingly embedded in sales processes, customer relationship management (CRM), and strategic decision-making in business-to-business (B2B) contexts, enhancing predictive accuracy, automation, and operational efficiency. As firms intensify their digital investments, GenAI is often portrayed as a transformative technology capable of redefining competitive dynamics.

In this study, GenAI is understood as a class of machine learning (ML) models capable of producing new, context-specific content—such as text, images, or code—rather than merely classifying or predicting based on existing data. GenAI applications can support sales and account management activities in B2B environments by generating tailored proposals, email drafts, product descriptions, and conversational responses that augment the capabilities of human sellers and relationship managers. Recent work on GenAI in B2B sales indicates that such tools can improve sales process effectiveness, administrative efficiency, and sales performance, thereby reshaping the role of AI in customer-facing work. In B2B settings, these applications are embedded in wider organisational systems that coordinate data, processes, people and decision rules, reinforcing the need to analyse GenAI adoption as part of an integrated socio-technical system rather than as a standalone tool.

The literature relevant to this study can be understood at three related levels. First, a broad stream on AI, analytics, and digital transformation examines how data-driven tech-nologies reshape organisational processes, capabilities, and performance outcomes. Second, a more specific body of work on organisational AI adoption focuses on the condi-tions under which firms assimilate AI-based systems and derive value from them, partic-ularly through complementary capabilities and governance mechanisms. Third, an emerging GenAI-specific literature highlights the distinctive generative, content-producing, and workflow-augmenting properties of these tools, especially in customer-facing and B2B contexts. The present study draws on all three streams, while positioning its empirical contribution primarily within the emerging literature on GenAI adoption in B2B organisational settings.

This generative capacity differentiates GenAI from traditional predictive analytics and recommendation systems typically examined in prior AI and business analytics research and implies distinct mechanisms through which AI-enabled experimentation and content creation may translate into organisational performance.

AI technologies are associated with enhanced service innovation capabilities and long-term competitive performance. However, evidence suggests that technological adoption alone does not automatically translate into sustainable organisational performance. Instead, AI’s value depends on complementary organisational capabilities, governance mechanisms, and environmental conditions that shape how technologies are integrated and deployed.

While prior research has examined AI capabilities, governance, and innovation orien-tations in broader digital transformation and AI contexts, evidence remains limited on how these organisational conditions jointly relate to GenAI adoption and long-term competitive performance in B2B settings.

Recent research on generative AI indicates its potential to stimulate both exploratory and exploitative innovation processes within firms, enabling experimentation while si-multaneously supporting efficiency-oriented improvements. Nevertheless, the mechanisms through which GenAI adoption contributes to sustainable organisational per-formance remain insufficiently understood, particularly in B2B environments characterised by high environmental dynamism and complex stakeholder relationships.

Simultaneously, growing regulatory scrutiny and ethical concerns related to algorith-mic bias, transparency, accountability, and responsible AI use underscore the importance of structured governance frameworks. Therefore, ethical governance may act not only as a compliance mechanism but also as a strategic enabler that determines whether AI adoption generates sustainable value.

Against this backdrop, this study investigates the relationships among GenAI adop-tion, organisational ambidexterity (exploratory and exploitative innovation), ethical gov-ernance, environmental dynamism, and sustainable organisational performance (here understood as long-term competitive performance) in Portuguese B2B firms. This study seeks to clarify the conditions under which GenAI contributes to long-term competitive sustainability by integrating technological, organisational and contextual perspectives into a single empirical framework.

From a systems-informed perspective, organisations can be understood as socio-technical arrangements in which governance structures, innovation capabilities, and tech-nological infrastructures are interdependent and jointly shape organisational outcomes. Rather than examining these elements in isolation, a systems-oriented approach empha-sises the interdependencies among organisational subsystems that jointly shape technology adoption and performance outcomes. In this study, GenAI adoption, ethical gov-ernance, environmental dynamism, and innovation capabilities are conceptualised as interconnected components of an organisational system that collectively shape sustainable organisational performance.

This study contributes to the literature in three ways. First, rather than proposing a new standalone theory, it offers a theory-integrative explanation of how governance, innovation capabilities, and GenAI adoption jointly relate to long-term competitive perfor-mance in B2B firms. Second, it shows that ethical governance should not be understood merely as a compliance safeguard or contextual condition, but as a core organisational capability strongly associated with long-term competitive performance. Third, it refines ambidexterity-based explanations by showing differentiated pathways to performance: exploitative innovation is directly associated with performance, whereas exploratory inno-vation is associated with performance indirectly through GenAI adoption.

Overall, the study advances systems-oriented research on digital transformation by proposing and testing a system-level model of GenAI adoption and sustainable perfor-mance, in which ethical governance, environmental dynamism and innovation capabilities are treated as interdependent organisational dimensions rather than independent predic-tors. By adopting a systems-informed perspective and modelling governance structures, innovation capabilities, and GenAI adoption as interrelated organisational components, this study explains how socio-technical configurations shape sustainable organisational performance in AI-enabled B2B contexts. In this study, long-term competitive performance refers to the firm’s perceived ability to sustain competitive effectiveness, adaptability, and resilience over time. The term is used in a strategic and organisational sense and does not refer to environmental or social sustainability outcomes.

Accordingly, the objective of this study is to examine how ethical governance, envi-ronmental dynamism, exploratory and exploitative innovation, and GenAI adoption are associated with long-term competitive performance in Portuguese B2B firms.

The remainder of this paper is structured as follows: Section 2 presents the theoretical background and hypothesis development. Section 3 describes the research methodol-ogy. Section 4 reports the empirical results. Section 5 discusses the findings. Section 6 outlines the theoretical implications, followed by managerial implications in Section 7. Section 8 presents the study’s limitations and directions for future research. Finally, Section 9 concludes.

2. Theoretical Background, Conceptual Framework, and Hypothesis.

This study adopts a systems-oriented perspective, conceptualising organisations as interdependent organisational systems in which governance mechanisms, innovation ca-pabilities and technological adoption interact to generate performance outcomes. However, the empirical model does not capture dynamic feedback loops, reciprocal cau-sation, or non-linear system behaviour directly. Instead, it provides a structured exami-nation of how key organisational dimensions—governance, innovation capabilities, and GenAI adoption—are associated with long-term competitive performance within a systems-informed framework.

The theoretical framing developed in this study integrates insights from three re-lated literature streams: research on AI-enabled digital transformation, studies on organisational AI adoption and capability development, and the more recent literature on GenAI use in B2B contexts. This distinction is important because some of the mechanisms examined here—such as governance, absorptive capacity, and ambidexterity—originate in broader organisational and AI literature, whereas the empirical focus on generative AI reflects a more recent and still emerging stream of research.

In recent years, the relationship between digital transformation and sustainable organ-isational performance has attracted growing scholarly attention. In this study, sustainable organisational performance is regarded as long-term competitive sustainability, reflecting organisational resilience, adaptability and enduring value creation rather than environ-mental sustainability per se. In this context, sustainability refers to organisational endurance and long-term competitive viability rather than environmental sustainability outcomes. Research on AI and advanced analytics highlights their potential to en-hance decision-making, operational efficiency, and strategic responsiveness. However, whether technological adoption alone generates sustained performance improvements remains inconclusive, particularly in dynamic and uncertain environments.

Existing literature frequently examines technological capabilities, governance mech-anisms, and innovation orientations in isolation. However, the interaction among insti-tutional structures, environmental pressures, and internal strategic capabilities is more likely to result in sustainable performance outcomes, consistent with integrative capability perspectives. The role of governance frameworks in shaping legitimacy and long-term value creation is highlighted in institutional theory. Contemporary developments further emphasise how institutional logics and legitimacy dynamics shape organisational responses in sustainability-oriented contexts. The resource-based view emphasises that competitive advantage is derived from valuable, rare, and well-integrated capabilities, such as innovation and technological absorption. Complementarily, organisational am-bidexterity theory underscores the need to balance exploratory and exploitative innovation to sustain long-term competitiveness.

Building on these perspectives, this study proposes an integrative framework that positions ethical governance, environmental dynamism, and innovation capabilities as interconnected determinants of sustainable organisational performance. Rather than assum-ing a direct technological-performance link, the model examines how governance structures and innovation orientations shape both GenAI adoption and sustainable outcomes.

2.1. Ethical Governance and Sustainable Organisational Performance.

Ethical governance has emerged as a foundational pillar of sustainable organisational performance, particularly in technology-intensive and data-driven environments. In this study, sustainable organisational performance is understood as long-term competitive sustainability, emphasising enduring efficiency, adaptability and market resilience rather than environmental or social sustainability outcomes. Governance structures ensure trans- parency, accountability, compliance and responsible strategic conduct, thereby contributing to durable value creation and stakeholder legitimacy. Recent research emphasises that governance quality directly influences firm resilience and sustainability outcomes, especially when organisations rely on advanced analytics and AI-based systems.

From an institutional theory perspective, governance mechanisms play a central role in shaping organisational legitimacy, conformity, and strategic behaviour. Recent research on responsible AI governance reinforces this perspective by highlighting how formalised oversight structures institutionalise ethical norms in digital environments. Organisations operate within broader regulatory, normative, and cognitive environments that exert pressures towards compliance and responsible conduct, particularly in the context of algorithmic decision-making and data usage. In this context, structured gover-nance frameworks—such as formal ethical guidelines, oversight bodies, and accountability systems—serve not only as instruments of control but also as mechanisms through which firms secure legitimacy and reduce institutional uncertainty. In sustainability-oriented organisations, structured ethical oversight strengthens the alignment between strategic objectives and responsible business practices, thereby reducing regulatory and compliance risks while enhancing stakeholder trust and reputational capital.

Rather than acting merely as a compliance safeguard, ethical governance can be conceptualised as a strategic capability that shapes how firms translate technological and innovation initiatives into sustainable performance outcomes over time. Accordingly, we postulate the following hypothesis:

H1. Ethical governance is positively associated with sustainable organisational performance.

2.2. Environmental Dynamism and Sustainable Organisational Performance.

Environmental dynamism reflects volatility in market conditions, customer demands, and technological evolution, capturing the rate and unpredictability of change in a firm’s external context. Dynamic environments create pressure for strategic adaptation and responsiveness because firms must continuously adjust their offerings, processes, and decision-making routines to remain competitive. In such contexts, organisations that effectively sense and respond to environmental shifts are more likely to sustain competitive advantage and long-term performance.

Empirical research suggests that environmental dynamism can act both as a chal-lenge and an opportunity for performance sustainability. High volatility increases uncertainty and implementation risk; on the other hand, it can reward firms that are able to rapidly reconfigure capabilities and innovate. In the present study, we focus on sus-tainable organisational performance as long-term competitive sustainability and examine whether firms operating in more dynamic environments report stronger performance, reflecting their ability to adapt and maintain resilience. Accordingly, we formulate the following hypothesis:

H2. Environmental dynamism is positively associated with sustainable organisational performance.

2.3. Organisational Innovation Capabilities.

Organisational ambidexterity theory provides a foundational framework for under-standing how firms pursue innovation and efficiency through distinct yet complementary learning processes. March distinguishes between exploration and exploitation as fundamentally different modes of organisational learning. Subsequent research (e.g., Anzenbacher and Wagner ) further develops this distinction. In organisational contexts, these learning orientations are reflected in exploratory and exploitative innovation capabili-ties, which represent firms’ tendencies to pursue experimentation or efficiency-oriented improvements. Exploratory innovation emphasises experimentation, variation, risk-taking, technological search and the development of novel knowledge, thereby supporting long-term adaptability under conditions of uncertainty. In contrast, exploitative innovation focuses on refinement, efficiency, implementation and incremental improvement of existing competencies, enhancing reliability, control and short-term performance.

Although exploration and exploitation compete for organisational resources, ambidex-trous organisations are characterised by their capacity to sustain both orientations over time. In empirical studies, these orientations are frequently operationalised as exploratory and exploitative innovation capabilities. Prior research links such dual capability configu-rations to organisational resilience and sustained competitive performance, particularly in technology-intensive and digitally transforming sectors. This dual orientation becomes especially salient in AI-intensive environments as firms must experiment with emerging technologies while refining operational routines to extract value from them.

In this study, exploratory and exploitative innovation are conceptualised as distinct capability dimensions rather than as an interaction-based ambidexterity construct. This dual-orientation perspective allows us to examine the differentiated structural effects on both Generative AI adoption and sustainable organisational performance without assuming that ambidexterity arises from a specific balance or interaction between the two.

This approach is consistent with recent empirical studies on organisational ambidex-terity that operationalise exploration and exploitation as separate capability dimensions measured through survey instruments rather than interaction terms. Such a perspective allows researchers to examine the differentiated structural effects of each innovation orien-tation on organisational outcomes without assuming that ambidexterity necessarily arises from a specific balance between the two dimensions. Recent research adopting survey-based measurement models has followed this approach when analysing digital innovation and AI-enabled organisational capabilities.

Therefore, we investigate whether each innovation capability in B2B firms directly contributes to sustainable organisational performance. Accordingly, we hypothesise:

H3. Exploitative innovation is positively associated with sustainable organisational performance.

H4. Exploratory innovation is positively associated with sustainable organisational performance.

2.4. Innovation Capabilities and Generative AI Adoption.

The adoption of emerging technologies, such as GenAI, is not merely a function of external technological availability but fundamentally depends on internal organisational capabilities, particularly absorptive capacity and innovation orientation. Absorptive capacity enables firms to recognise the value of new knowledge, assimilate it, and apply it effectively within existing processes, while innovation capabilities shape the willingness to experiment with and integrate novel digital tools into core workflows. In the case of GenAI, this implies the ability to deploy new algorithms and redesign content-intensive ac-tivities, such as proposal writing, opportunity qualification, and customer communication, around generative outputs.

Firms characterised by exploratory innovation tend to emphasise experimentation, technological search and openness to novel solutions, which increases their propensity to adopt and test advanced digital technologies, such as GenAI. In B2B sales and account management contexts, exploratory-oriented firms are particularly well positioned to pilot GenAI use cases that involve AI-generated sales pitches, tailored proposals, creative campaign ideas, or co-created solution concepts with customers, where the outputs are novel and their value is initially uncertain. These organisations are more willing to tolerate ambiguity and iteration as they learn how to embed generative capabilities in their sales and decision-making routines.

In contrast, exploitative-oriented firms focus on the efficiency, refinement, and in-cremental improvement of existing capabilities. GenAI may be deployed primarily to standardise and streamline repetitive tasks, such as generating first drafts of routine commu-nications, summarising customer interactions into CRM systems, or producing consistent documentation across accounts. Rather than prioritising experimentation for its own sake, exploitative innovation supports GenAI implementation when clear efficiency gains, cycle-time reductions or quality improvements can be anticipated and when generative tools can be embedded into existing processes with limited disruption.

Taken together, these arguments suggest that both exploratory and exploitative inno-vation orientations may influence GenAI adoption, albeit through distinct mechanisms, one through experimentation and knowledge expansion in novel generative use cases, the other through efficiency-seeking and process optimisation in existing sales and decision workflows. Accordingly, we propose the following hypotheses:

H5. Exploratory innovation is positively associated with generative AI adoption.

H6. Exploitative innovation is positively associated with generative AI adoption.

2.5. Generative AI Adoption and Sustainable Organisational Performance.

GenAI systems provide advanced decision-support capabilities that go beyond tra-ditional analytics by producing new textual, visual, or conversational outputs that can augment human judgement in sales, marketing, and strategic decision-making. In B2B set-tings, GenAI can assist sales and account teams in preparing tailored proposals, generating personalised communications, drafting responses to customer queries, and synthesising complex information into persuasive narratives for key accounts. Initial empirical findings suggest that GenAI usage can improve sales process effectiveness, administrative efficiency, and sales performance in B2B environments, highlighting its potential to reshape customer-facing activities.

From a resource-based view, GenAI adoption can therefore be conceptualised as a strategic organisational capability rather than merely a technological tool. Its contri-bution to sustainable organisational performance depends on effective integration within organisational routines, the development of complementary skills and governance mecha-nisms and alignment with broader innovation and digital transformation strategies. When generative capabilities are embedded into core B2B sales and decision processes and supported by appropriate oversight, they may be associated with long-term competi-tive sustainability by improving efficiency, responsiveness and relationship quality with key customers.

Empirical evidence regarding AI’s direct performance impact remains mixed, and systematic research on GenAI in B2B contexts is limited. Existing studies on AI and advanced analytics suggest that performance benefits often materialise only when AI deployment is strategically integrated and supported by dynamic capabilities, rather than when technologies are adopted in isolation. This perspective implies that GenAI adoption should be examined along with ethical governance and innovation capabilities, which may determine whether generative tools translate into enduring performance ad-vantages. Nevertheless, given its potential to enhance efficiency, adaptability, and value creation in B2B environments, we posit a direct positive association between GenAI adoption and sustainable organisational performance. Accordingly, we formulate the following hypothesis:

H7. Generative AI adoption is positively associated with sustainable organisational performance.

2.6. Ethical Governance and Generative AI Adoption.

Governance mechanisms can significantly shape how organisations approach the adoption and use of AI technologies, including GenAI. From an institutional and strategic management perspective, governance structures define decision-making boundaries, risk tolerance and accountability mechanisms that influence how technological initiatives are evaluated, prioritised and monitored. Firms with structured ethical oversight frameworks are more likely to assess AI technologies through the lenses of compliance, transparency and long-term strategic alignment, rather than purely short-term experimentation or cost reduction.

Prior research suggests that effective IT and data governance enhances business–IT alignment and channels technology deployment towards core organisational objectives. In sustainability-oriented contexts, ethical governance does not merely constrain technolog-ical initiatives; it can also direct them towards responsible, stakeholder-aligned applications that reinforce legitimacy and trust. In the specific case of GenAI, ethical governance may play a dual role: mitigating risks related to bias, opacity, and misuse, while also enabling the scaling of AI solutions that are compatible with regulatory expectations and stakeholder values.

However, the relationship between governance and GenAI adoption is not necessarily straightforward. Strong governance may accelerate adoption by reducing perceived risk and clarifying acceptable use cases. However, stringent controls may slow down experimen-tation and limit the breadth of applications pursued. In this study, we conceptualise ethical governance as a structural condition that can facilitate responsible GenAI implemen-tation by providing clear principles, accountability structures, and alignment with strategic goals. Accordingly, we propose the following hypothesis:

H8. Ethical governance is positively associated with generative AI adoption.

2.7. Environmental Dynamism and AI Adoption.

Dynamic environments increase decision complexity, market volatility, and strategic uncertainty, intensifying the need for timely information and adaptive capabilities. Firms face rapidly changing customer demands, competitive moves, and technological trajectories that may challenge traditional decision-making routines under conditions of high environmental dynamism. However, the relationship between environmental dynamism and AI adoption is likely to be more complex than a simple direct effect. While a “pressure-driven” perspective suggests that greater volatility directly stimulates technology adoption, a “capability-driven” view argues that when firms possess sufficient absorptive capacity, digital maturity, and complementary resources, environmental pressures only translate into AI implementation. In B2B contexts characterised by long sales cycles and relationship-based competition, dynamism may increase the perceived need for AI, but internal innovation capabilities may be more influential in actual adoption than external pressures alone. This study explores whether Portuguese B2B firms operating in more turbulent contexts are more likely to integrate GenAI into sales processes. Accordingly, the following exploratory hypothesis is formulated:

H9. Environmental dynamism is positively associated with generative AI adoption.

2.8. Conceptual Framework of the Study.

Figure 1 presents the integrative conceptual framework. The model assumes that governance mechanisms and environmental dynamism are associated with sustainable or- ganisational performance while also shaping the conditions under which GenAI is adopted. Innovation capabilities operate as key organisational dimensions of AI implementation, reflecting the ambidextrous balance between experimentation and operational refinement. Instead of assuming a linear relationship between technology and performance, the frame-work treats governance, innovation capabilities, and GenAI adoption as interdependent organisational dimensions. However, it does not model interaction terms, reciprocal effects, or dynamic feedback processes.

3. Research Methodology.

3.1. Research Design.

This study adopts a quantitative, cross-sectional research design to examine the re-lationships among GenAI adoption, organisational ambidexterity, ethical governance, environmental dynamism, and sustainable organisational performance in Portuguese B2B firms.

A survey-based approach was selected to measure latent constructs and test the pro-posed hypotheses. Survey research is particularly appropriate for examining managerial perceptions and organisational practices across firms, especially when analysing latent con-structs within explanatory models that also assess predictive relevance. Given the model’s predictive orientation, the exploratory nature of GenAI adoption in B2B contexts, and the moderate sample size, Partial Least Squares Structural Equation Modeling (PLS-SEM) was considered methodologically appropriate. PLS-SEM was selected because the study aims to estimate structural relationships among multiple latent constructs measured by multi-item scales, rather than to reduce variables into component scores. In this context, PCA was not appropriate because it is primarily a data reduction technique and does not test structural relationships among latent variables within an integrated explanatory model. Because the study relies on cross-sectional survey data, the estimated relationships should be interpreted as theory-guided directional associations rather than definitive causal effects.

3.2. Sample and Data Collection.

The final sample comprised 104 Portuguese managers and professionals from various sectors with direct experience in the use of AI-based systems employed in B2B companies. The inclusion criteria required respondents to have direct professional involvement in strategic or operational roles in B2B contexts and demonstrable exposure to AI-related tools, processes, or decision-making activities. Eligibility was verified through filter questions at the beginning of the questionnaire. The variable “years of professional experience” reported in Table 1 refers to respondents’ overall professional experience and should not be interpreted as AI-related experience. Data were collected through a self-administered online questionnaire developed in Google Forms, pretested with 20 professionals (Cronbach’s alpha > 0.846 in the main scales).

The questionnaire was distributed via targeted email between November 2025 and January 2026 (n = 647 invitations, response rate = 16.1%). Participation was voluntary and anonymous, with informed consent obtained at the beginning of the study and prior approval from the Ethics Committee. Incomplete answers (n = 3) were excluded, following standard criteria to mitigate attention bias and self-selection.

Although efficient in reaching qualified professionals in a digital context, this ap-proach has inherent limitations associated with online questionnaires, namely potential self-selection bias (respondents more interested in AI) and lower representation of SMEs or traditional sectors with lower digital literacy. The generalisation of the results should consider the Portuguese B2B context and technological landscape prevailing at the time of data collection. In addition, because the study relies on a non-probabilistic sample and one respondent per firm, the findings should be interpreted as context-specific evidence rather than as broadly generalisable relationships.

The demographic statistics were analysed using IBM SPSS Statistics v31 (Table 1).

Sample adequacy was assessed through an a priori power analysis using G*Power 3.1. Assuming a multiple regression model with a maximum of five predictors— corresponding to the largest number of structural paths directed at any endogenous con-struct in the model—a medium effect size (f2 = 0.15), and a significance level of α = 0.05, the minimum required sample size to achieve a statistical power of 0.80 was 92 observations, following Cohen’s effect size conventions. The final sample of 104 valid responses exceeded this threshold, ensuring adequate statistical power to detect medium-sized effects.

Furthermore, the sample size satisfies the established guidelines for PLS-SEM, which recommend that the minimum sample size should exceed 10 times the maximum number of structural paths directed at any endogenous construct. In addition, recent method-ological advances suggest that power analysis provides a more rigorous assessment of sample adequacy in PLS-SEM applications.

3.3. Research Instrument.

A structured questionnaire was developed to collect perceptual data from B2B firm managers. All constructs were measured using multi-item scales on a five-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”).

Measurement items were adapted from validated scales widely used in prior research. Environmental dynamism items were adapted from Dubey et al., Exploratory and exploitative innovation from Jansen et al. and ethical governance from recent studies on responsible AI governance. Sustainable organisational performance items were adapted from research on AI-enabled and analytics-driven performance outcomes. In line with the conceptual perspective adopted in this study, sustainable organisational performance refers to long-term competitive performance, reflecting the organisation’s ability to maintain competitive effectiveness and value creation over time rather than environmental or social sustainability outcomes.

GenAI adoption was measured using perceptual items adapted from prior research on IT assimilation, AI capability, and digital workflow integration at the organisational level. The items capture the extent to which GenAI tools are embedded in B2B sales processes and decision-making routines. All items were contextualised to reflect the GenAI and B2B sales setting of this study (see Appendix A). However, the measurement captures the perceived extent of GenAI use and integration rather than distinguishing between exploratory pilot initiatives and fully embedded organisational applications, which should be considered when interpreting the results.

Survey-based measures are widely used in research on digital transformation and artificial intelligence adoption, particularly when assessing managerial perceptions of technology usage, organisational readiness, and capability development. Recent empirical studies have successfully applied multi-item questionnaire scales to examine firms’ AI capabilities, adoption patterns, and perceived performance outcomes.

Common method bias (CMB) was assessed using Harman’s single-factor test. The results indicated that no single factor accounted for the majority of variance. In addition, collinearity diagnostics were examined at both the indicator and structural levels. All inner VIF values remained below the critical threshold of 5 and below the conservative threshold of 3.3, indicating that full collinearity bias and common method variance are unlikely to significantly affect the structural estimates. Taken together with the procedural remedies applied during survey design (e.g., respondent anonymity and careful question-naire design), these results suggest that common method variance does not appear to be a serious concern in this study. Nevertheless, because all constructs were measured from a single respondent at one point in time, the possibility of residual common method variance cannot be completely ruled out. Accordingly, the structural relationships should be inter- preted with appropriate caution, particularly where perceptual and conceptually proximate constructs are involved. This may have inflated some of the observed associations.

3.4. Data Analysis Procedure.

Data were analysed using PLS-SEM implemented in SmartPLS 4. PLS-SEM is appropriate for research models aimed at examining structural relationships and assess-ing predictive relevance, theory development, and moderate sample sizes. The recommended two-step approach was followed for the analysis: (i) measurement model assessment and (ii) structural model assessment.

3.4.1. Measurement Model Assessment.

The measurement model was evaluated in terms of indicator reliability, internal consistency reliability, convergent validity, discriminant validity, and collinearity.

• Indicator reliability. Outer loadings above 0.70 indicate that the construct explains >50% of the indicator’s variance, ensuring satisfactory reliability.

• Internal consistency reliability. Evaluated using Cronbach’s alpha (CA), rho_A, and composite reliability (ρc). In SmartPLS, ρc corresponds to the traditional composite reliability measure proposed by Werts et al.. Values above 0.70 indicate acceptable reliability.

• Convergent validity. Assessed via the Average Variance Extracted (AVE). AVE values above 0.50 demonstrate adequate convergence.

• Discriminant validity. Evaluated using the heterotrait–monotrait ratio and Fornell– Larcker criterion. HTMT values below 0.90 indicate adequate discriminant valid-ity. According to the Fornell–Larcker criterion, the square root of the AVE (diago-nal values) is greater than the correlations between constructs.

• Collinearity. Examined using the Variance Inflation Factor (VIF). VIF values below 5 indicate that multicollinearity is not a concern.

3.4.2. Structural Model Assessment.

After confirming the measurement model’s adequacy, the structural model was evalu-ated. At the structural level, collinearity among predictor constructs was further assessed using inner VIF values.

• Path coefficients (β). The strength and direction of the relationships between the constructs are represented. Statistical significance was assessed using bootstrapping with 5000 resamples. A two-tailed test was performed at the 5% significance level (t ≥ 1.96; p < 0.05).

Coefficient of determination (R2). Evaluates the explanatory power. The values of • 0.25, 0.50, and 0.75 indicate weak, moderate, and substantial explanatory power, respectively.

Effect size (f2). The impact of each exogenous construct on endogenous variables • According to Cohen’s guidelines, values of 0.02, 0.15, and 0.35 indicate small, medium, and large effects, respectively.

• Predictive relevance. Out-of-sample predictive relevance was assessed using the Cross-Validated Predictive Ability Test (CVPAT) and PLSpredict. CVPAT compares prediction errors against a linear benchmark at the construct level, where a significant negative average loss difference (p < 0.05) indicates superior predictive relevance. PLSpredict evaluates indicator-level predictive relevance using Q2 predict values and RMSE comparisons with a linear model benchmark; Q2 predict values above zero and lower prediction errors indicate predictive relevance.

This procedure ensures robustness in the evaluation of both the measurement quality and structural relationships.

4. Data Analysis and Results.

4.1. Results of the Measurement Model Assessment.

The internal consistency reliability, convergent validity, discriminant validity, and collinearity of the measurement model were evaluated. As shown in Table 2, all con-structs demonstrate strong reliability, with Cronbach’s alpha (CA), rho_a, and rho_c values exceeding the recommended threshold of 0.70.

Note: VIF < 5; CA ≥ 0.70; rho_a;rho_c ≥ 0.70 and AVE ≥ 0.50.

Convergent validity is also established, as all Average Variance Extracted (AVE) values are above 0.50, indicating that each construct explains a substantial proportion of variance in its indicators.

Collinearity diagnostics confirmed that multicollinearity was not a concern because all VIF values remained below the recommended threshold of 5.

Overall, the measurement model exhibits robust psychometric properties, supporting its suitability for structural model evaluation (Table 2).

The discriminant validity was assessed using the HTMT ratio of correlations. Most HTMT values were below the conservative threshold of 0.90, indicating adequate discrimi-nant validity across the constructs (Table 3). The HTMT value between ethical governance (EG) and long-term competitive performance (SOP) was 0.909, slightly above this conser-vative threshold, which suggests a degree of conceptual proximity that warrants caution in interpretation. This pattern is substantively plausible, as ethical governance captures formalised oversight and accountability mechanisms, while long-term competitive perfor-mance reflects managerial perceptions of enduring organisational effectiveness, adaptabil-ity, and resilience. Accordingly, the observed association may reflect partial conceptual proximity rather than construct redundancy.

Discriminant validity was further examined using HTMT inference based on bias-corrected bootstrapping (95% confidence interval). None of the confidence intervals in-cluded the value of 1.000, providing additional evidence that the constructs remain empiri-cally distinguishable despite their proximity.

The Fornell–Larcker criterion further supports discriminant validity because the AVE square root of each construct exceeds its inter-construct correlations (Table 4).

4.2. Results of the Structural Model Assessment.

After confirming the adequacy of the measurement model, the structural model was evaluated using bootstrapping with 5000 resamples to assess the hypothesised relationships’ significance and magnitude. The analysis examined the standardised path coefficients (β), their statistical significance, effect sizes (f2), and the explanatory power of the model (R2).

The model suggests substantial explanatory capability, accounting for 65.2% and 82.3% of the variance in GenAI adoption (GAIA) and sustainable organisational performance (SOP) (Figure 2). Although the R2 value for sustainable organisational performance (SOP) (0.823) is substantial, this result should be interpreted with some caution because sev-eral predictors, particularly ethical governance (EG), exploitative innovation (ExyI), and GenAI adoption (GAIA), show relatively strong associations with SOP. Consistent with the Fornell–Larcker results, the correlation between ethical governance (EG) and sustainable organisational performance (SOP) is 0.826 and the correlation between exploitative innova-tion (ExeI) and sustainable organisational performance (SOP) is 0.803. When squared, these correlations correspond to approximately 68% and 64% of shared variance respectively, indicating that a considerable portion of the explanatory power arises from conceptually related organisational capability constructs.

This pattern raises the possibility that part of the explanatory power reflects a certain conceptual proximity among capability-related organisational constructs rather than ex- clusively independent predictive effects. However, the discriminant validity assessments provide evidence that the constructs remain empirically distinct. The HTMT inference results with 95% confidence intervals (BCa), Fornell–Larcker criterion, and indicator-level validity checks consistently support discriminant validity across all constructs.

Consequently, the high explanatory power observed for sustainable organisational per-formance is more plausibly explained by the model’s integrative nature—where governance mechanisms, innovation capabilities, and digitally embedded GenAI adoption operate as interdependent organisational dimensions—rather than by full construct redundancy.

As an additional robustness check, firm size was introduced as a control variable affecting the endogenous constructs. The inclusion of firm size did not alter the overall pat-tern of significance for the strongest relationships in the model, although some coefficients changed in magnitude and some previously significant associations became non-significant. In particular, the positive association between ethical governance and sustainable organ-isational performance remained strong and statistically significant (β = 0.378, p < 0.001), and the association between exploratory innovation and GenAI adoption also remained strong (β = 0.697, p < 0.001). However, environmental dynamism no longer showed a statistically significant association with sustainable organisational performance (β = 0.134, p = 0.112), and exploratory innovation remained non-significant in its direct association with sustainable organisational performance (β = −0.201, p = 0.063). Overall, these results suggest that the main findings are broadly robust to firm-size controls, while indicating that some coefficient estimates are sensitive to model specification. The full robustness results are reported in Appendix B.

Table 5 shows the standardised path coefficients, t-values, p-values and effect sizes. Ethical governance (EG) shows the strongest positive association with sustainable organisa-tional performance (SOP) (β = 0.366, p < 0.001), with a large effect size (f2 = 0.438).

Environmental dynamism (ED) and exploitative innovation (ExeI) are also positively associated with performance, although with smaller effect sizes. In contrast, exploratory innovation (ExyI) does not show a significant direct association with sustainable organisa-tional performance.

Exploratory innovation is positively associated with GenAI adoption (GAIA), whereas ethical governance, environmental dynamism, and exploitative innovation do not show statistically significant associations with GenAI adoption.

GenAI adoption is positively associated with sustainable organisational performance, suggesting a complementary role in the organisational system. This suggests that AI adop- tion is associated with a complementary performance-enhancing role, while governance structures remain the strongest organisational correlate of sustainability outcomes.

Structural collinearity was assessed using inner VIF values for the predictors of the endogenous constructs. The inner VIF values for the predictors of sustainable organ-isational performance remain below the recommended threshold of 5, indicating that multicollinearity does not threaten the stability of the structural estimates.

To further examine the underlying mechanisms linking organisational factors and sustainable organisational performance, mediation effects were assessed using bootstrap-ping procedures (Table 6). The results indicate that the association between exploratory innovation and sustainable organisational performance operates through GenAI adoption (β = 0.221, p < 0.01), suggesting that exploratory innovation is indirectly associated with performance via its relationship with GenAI adoption. No statistically significant indirect effects were observed for the remaining constructs (Table 6).

Note: Bootstrapping with 5000 resamples; two-tailed test.

4.3. Predictive Relevance.

A cross-validated predictive ability test (CVPAT) was conducted by comparing the PLS-SEM model with a linear benchmark model (LM) to assess the predictive relevance of the model. The PLS model demonstrated superior predictive relevance (average loss difference = −0.111, t = 2.507, p = 0.014). At the construct level, the model shows signifi-cantly better prediction accuracy for sustainable organisational performance (SOP) (average loss difference = −0.089, t = 2.106, p = 0.038). Although the PLS model also exhibited lower prediction loss for GenAI adoption (GAIA), the difference was not statistically significant (average loss difference = −0.132, t = 1.700, p = 0.092). (Table 7).

These findings provide partial evidence of predictive relevance, particularly for sus-tainable organisational performance.

Table 8 presents the PLSpredict results at the indicator level. All Q2_predict values are above zero, confirming out-of-sample predictive relevance. The PLS-SEM model outperformed the linear benchmark in nine out of ten indicators, indicating indicator-level predictive relevance, particularly for sustainable organisational performance.

Overall, the PLS-SEM model outperformed the linear benchmark at the global level and for sustainable organisational performance, indicating partial out-of-sample predictive relevance. However, the predictive results are mixed, as GenAI adoption did not reach statistical significance in the CVPAT assessment and one indicator was better predicted by the linear benchmark. Accordingly, the model’s predictive relevance should be interpreted with caution.

5. Discussion.

The purpose of this study was to examine how ethical governance, innovation capabil-ities, environmental dynamism, and Generative AI adoption jointly relate to sustainable organisational performance in B2B firms. The results provide empirical support for several of the proposed relationships and offer a nuanced understanding of how technological and organisational factors interact in AI-intensive contexts. Taken together, these find-ings reinforce a systems-informed view of organisations as socio-technical arrangements in which governance structures, innovation capabilities, and AI-enabled technologies in-teract as interdependent organisational subsystems. In this perspective, governance mechanisms, innovation capabilities, and AI-enabled technological infrastructures can be understood as interdependent organisational subsystems that jointly shape the outcomes of digital transformation.

The results further indicate that sustainable organisational performance—understood here as long-term competitive sustainability—emerges from the structured alignment between governance mechanisms, innovation orientations, and digitally embedded capa-bilities rather than from a simple linear relationship between technology adoption and performance outcomes. Consistent with this systemic perspective, the relatively high explanatory power observed for sustainable organisational performance reflects the com-bined contribution of multiple organisational capabilities that together support enduring competitive effectiveness. At the same time, the robustness analysis suggests that not all performance-related associations show the same degree of stability across model specifica-tions. Accordingly, the evidence appears strongest for the core relationships in the model, whereas some weaker direct associations with performance should be interpreted more cautiously. Each set of relationships is discussed in detail in the following subsections, and the results are interpreted considering the theoretical foundations presented in Section 2.

5.1. Ethical Governance as a Driver of Sustainable Organisational Performance.

The results indicate that ethical governance shows the strongest positive association with sustainable organisational performance, supporting H1. This finding is consistent with institutional theory, which highlights the role of governance structures in shaping legitimacy and stability. It also aligns with contemporary AI governance research, which underscores the institutionalisation of ethical oversight as a stabilising mechanism in technology-intensive environments. In AI-intensive environments characterised by algorithmic opacity and ethical risks, structured oversight mechanisms are associated with prior research linking governance quality to organisational resilience and performance stability.

The positive association between governance and performance aligns with prior re-search highlighting the strategic relevance of responsible AI frameworks. Rather than merely serving as compliance safeguards, ethical governance mechanisms—such as formal AI policies, transparency protocols, and accountability systems—may help reduce uncertainty and foster organisational coherence. This stability may facilitate consistent decision-making and strengthen stakeholder trust, thereby contributing to durable value creation.

However, ethical governance does not show a statistically significant association with GenAI adoption (p = 0.126). Although the estimated coefficient is positive (β = 0.142), the effect is not statistically reliable. In this context, governance functions as a structural foundation for sustainable performance rather than as a direct catalyst for GenAI implementation.

5.2. GenAI Adoption and Sustainable Organisational Performance.

GenAI is positively associated with sustainable organisational performance, support-ing H7. This result is consistent with the resource-based view, which posits that when effectively integrated into operational routines, valuable organisational capabilities contribute to sustained competitive advantage.

Empirical research on AI and advanced analytics has demonstrated that digital tech-nologies enhance decision quality, efficiency, and responsiveness when embedded within organisational processes. In B2B contexts, generative AI tools support sales processes, content generation, and CRM-related activities. The present findings are consistent with the argument that such integration is associated with long-term competitive sustainability.

Nevertheless, GenAI’s effect is smaller than that of ethical governance. This indi-cates that technological adoption alone is insufficient to guarantee a sustainable advan-tage. Instead, performance gains depend on the alignment of GenAI capabilities with organisational structures and innovation orientations, consistent with dynamic capability perspectives.

5.3. Innovation Capabilities and Differentiated Performance Pathways.

The findings reveal distinct pathways through which exploratory and exploitative innovation are associated with sustainable organisational performance.

Exploitative innovation is directly associated with improvements in performance, thereby supporting H3. This finding is consistent with March’s conceptualisation of ex-ploitation as refinement, efficiency, and incremental improvement, which in organisational contexts is reflected in exploitative innovation capabilities.

Exploitative learning strengthens reliability and operational consistency. The results indicate that incremental optimisation is associated with long-term competitive sustainability in B2B environments.

In contrast, exploratory innovation does not directly show a statistically significant as-sociation with sustainable organisational performance (H4 not supported). The negative (al-though non-significant) coefficient suggests that exploratory initiatives may involve short-term trade-offs between experimentation and operational efficiency. Exploratory activities typically require investments in search, experimentation, and learning processes directed to-wards uncertain technological opportunities. In resource-constrained organisations—such as the SME-dominated sample examined in this study—these activities may temporarily divert managerial attention and organisational resources away from efficiency-oriented operations and short-term performance outcomes. Consequently, the immediate perfor-mance implications of exploratory innovation may appear neutral or even slightly negative despite the potential long-term strategic benefits.

This interpretation is consistent with the literature on organisational ambidexterity, which highlights the inherent tension between exploration and exploitation activities and the resource trade-offs involved in balancing these innovation orientations. Within this framework, exploratory innovation is associated with long-term renewal and adaptability, whereas exploitative innovation is more directly associated with operational efficiency and measurable performance outcomes.

5.4. Exploratory Innovation and GenAI Adoption.

Although exploratory innovation does not directly show a statistically significant association with sustainable performance, it is strongly is associated with GenAI adoption, supporting H5. This result aligns with the literature on absorptive capacity and innova-tion orientation, which suggests that firms characterised by experimentation and technological openness are more likely to adopt emerging digital tools.

In B2B sales and decision-making contexts, exploratory-oriented firms are more willing to pilot generative AI applications, experiment with AI-generated content, and redesign workflows around novel technologies. The mediation analysis further indicates that GenAI adoption operates as the mechanism through which the association between exploratory innovation and sustainable organisational performance is observed in the estimated model.

This pattern suggests that when translated into operational digital systems, experi-mentation becomes associated with sustainability. GenAI may function as a mechanism that embeds exploratory learning within organisational routines, thereby helping translate experimentation into performance-relevant practices.

5.5. Environmental Dynamism and Organisational Responses.

A weak positive association was observed between environmental dynamism and sustainable organisational performance (β = 0.155, p = 0.045). Although statistically sig-nificant, the effect size is small (f2 = 0.053), suggesting that environmental turbulence may play a limited role in shaping performance outcomes compared with organisational capabilities such as governance and innovation. However, this association should be interpreted with caution, as it becomes non-significant when firm size is included as a control variable in the robustness analysis. This result suggests that the relationship is comparatively weaker and more specification-sensitive than the central paths in the model.

Dynamic environments may reward organisations that maintain flexibility and re-sponsiveness. However, environmental dynamism does not show a statistically significant association with GenAI adoption (H9 not supported). Contrary to pressure-driven assump-tions, contextual turbulence does not automatically trigger technological experimentation, at least within the sampled Portuguese B2B context. This result suggests that internal capabilities—particularly exploratory innovation—play a more decisive role than environ-mental volatility in their association with GenAI adoption decisions. While environmental dynamism may increase the need for adaptation, the ability to recognise, experiment with and operationalise emerging AI technologies primarily depends on the organisation’s exploratory capabilities. From a dynamic capability perspective, this finding indicates that firms require internally developed innovation capacities to translate technological opportunities into concrete digital initiatives. This explanation may be particularly relevant in the SME-dominated sample of the present study, where resource constraints can limit the extent to which environmental turbulence translates into technology adoption.

5.6. Integrated Interpretation of the Findings.

Taken together, the findings support a systemic interpretation of AI-enabled sustain-ability. The model explains a substantial proportion of variance in sustainable organi-sational performance (R2 = 0.823) and GenAI adoption (R2 = 0.652), indicating strong explanatory power. An important qualification emerges from the robustness analysis. Although the overall structural pattern remains broadly stable after the inclusion of firm size as a control, some coefficients vary in magnitude and significance. In particular, the association between environmental dynamism and sustainable organisational perfor-mance becomes non-significant, whereas the association between executive involvement and performance increases in magnitude in the controlled model. These results do not overturn the main interpretation of the model, but they indicate that some weaker direct associations with performance are sensitive to model specification and should therefore be interpreted more cautiously. These findings should also be interpreted in light of the study’s methodological boundaries, including the cross-sectional design, the single-informant data structure, and the Portuguese SME-dominated B2B context. At the same time, these results should be interpreted in light of the study’s methodological boundaries, including the cross-sectional design, the single-informant data structure, and the Portuguese SME-dominated B2B context. These findings should therefore be interpreted as context-bound evidence from Portuguese B2B firms rather than as universally generalisable relationships. Within these boundaries, the findings suggest that sustainable competitiveness in AI-intensive B2B contexts is associated with the alignment of the following structural components:

• Governance stability enhances legitimacy and organisational coherence, par-ticularly in digitally regulated environments where ethical AI governance becomes institutionalised.

• Innovation capabilities shape learning dynamics and experimentation.

• Digitally embedded GenAI capabilities, which operationalise exploratory learning and enhance efficiency.

The results indicate that generative AI is associated with sustainable performance when embedded within governance structures and supported by appropriate innovation orientations rather than reflecting a technology-deterministic perspective. This config-uration highlights the interdependence of institutional stability, organisational learning, and digital implementation in shaping long-term competitive sustainability. Overall, the evidence is strongest for the more central and stable relationships in the model, while some weaker direct performance links should be regarded as indicative rather than uniformly ro-bust across specifications. This systemic configuration extends prior research on AI-enabled digital transformation by showing that sustainable competitive outcomes are associated not only with technology adoption but with the coordinated interaction of governance mechanisms, innovation capabilities, and digitally embedded AI systems.

6. Theoretical Implications.

The findings generate several theoretical implications for research on AI-enabled sustainability, innovation capabilities, and governance structures. By conceptualising governance structures, innovation capabilities and GenAI adoption as interdependent organisational dimensions, this study extends systems-oriented research on AI-enabled digital transformation. These findings also suggest that innovation capabilities and GenAI adoption can be understood as complementary organisational capabilities within a broader system configuration, where exploratory learning mechanisms enable digital experimenta-tion while exploitative capabilities support efficiency-driven performance improvements.

Ethical governance as the strongest organisational correlate of performance. The result offers empirical elaboration of institutional theory in AI-intensive contexts by suggesting that governance mechanisms in AI-intensive environments operate not only as legitimacy-preserving structures but also as strategic organisational conditions associated with long-term competitive sustainability. In line with emerging research on responsible AI governance, these mechanisms may be interpreted as institutionalised organi-sational capabilities that translate normative expectations into performance-enhancing routines. While prior studies often conceptualise governance primarily as a compliance or moderating mechanism, the present findings position ethical governance as a core organisational capability that is strongly associated with sustained performance outcomes.

GenAI as a complementary capability. The findings extend resource-based argu-ments in the context of GenAI. Although GenAI adoption is positively associated with sustainable organisational performance (H7: β = 0.329, p = 0.007, f2 = 0.147), its effect is more moderate than that of governance. This suggests that, as generative AI tools become increasingly accessible, competitive differentiation shifts toward organisational configu-ration rather than technological possession. Thus, the findings contribute to emerging AI capability research by clarifying that technological adoption alone does not guarantee sustained competitive advantage.

Differentiated innovation pathways. Mediation analysis contributes to the theory of ambidexterity and organisational learning. Exploratory innovation does not directly show a statistically significant association with sustainable performance (H4: β = −0.172, p = 0.439) but is indirectly associated with performance through GenAI adoption (indirect effect: β = 0.221, p = 0.007). This full mediation pattern indicates that when translated into technology-enabled implementation mechanisms, exploratory capabilities generate strate-gic value. Conversely, exploitative innovation is directly associated with improvements in sustainable performance (H3: β = 0.142, p = 0.021) without significantly influencing AI adoption (H6: β = 0.001, p = 0.386). Exploration appears to require operational em-bodiment through AI systems to produce measurable sustainability outcomes in digitally transforming environments, whereas exploitation operates through more direct efficiency mechanisms. This interpretation extends ambidexterity theory by suggesting that exploration and exploitation contribute through differentiated pathways rather than sym-metrical performance effects.

Capability-driven vs. environment-driven adoption of AI. The absence of a signif-icant effect of environmental dynamism on GenAI adoption (H9: β = 0.044, p = 0.984) contrasts with the strong influence of exploratory innovation (H5: β = 0.672, p < 0.001). Al-though environmental dynamism exerts a modest direct effect on sustainable performance (H2: β = 0.155, p = 0.045), it does not appear to be a primary driver of technology adoption decisions in the Portuguese B2B context examined in this study. These findings suggest that GenAI implementation in this sample is more closely associated with internal innovation capabilities than with environmental turbulence. However, this interpretation should be considered with caution because the relative homogeneity of the SME-dominated sample may limit the observable variation in environmental pressures.

Several factors may explain this result. First, the sample’s predominance of mi-cro/small enterprises (84.6%) may indicate resource constraints that limit rapid environ-mental responsiveness through technology investments. Second, the low entry barriers and widespread availability of GenAI tools may have created a relatively uniform baseline of experimentation across firms, reducing variance attributable to environmental pres- sures. These findings contribute to digital transformation research by emphasising internal capabilities over purely environmental triggers in early-stage AI adoption.

The integrative theoretical configuration. Overall, the study advances an integrative theoretical perspective in which governance structures constitute foundational stability mechanisms, innovation orientations shape digital experimentation, and GenAI adoption operates as a mediating capability that converts exploratory learning into sustainable perfor-mance outcomes. This systemic configuration explains 82.3% of the variance in sustainable organisational performance (R2 = 0.823) and 65.2% of the variance in GenAI adoption (R2 = 0.652), thereby enriching theoretical debates at the intersection of institutional theory, resource-based perspectives and organisational ambidexterity in AI-enabled environments.

7. Managerial Implications.

The findings offer several actionable implications for managers operating in digitally transforming B2B environments.

Ethical governance should be treated as a strategic investment rather than merely a regulatory obligation. The strong direct association between governance and sustainable organisational performance suggests that firms with structured AI oversight, transpar-ent decision-making processes, and accountability mechanisms are better positioned to sustain competitive performance over time. Therefore, managers should formalise AI gov-ernance frameworks, establish oversight committees and integrate ethical evaluation into digital transformation strategies. Governance stability appears to provide the structural foundation upon which AI-enabled value creation can be sustained.

GenAI adoption is positively associated with performance, but its impact is com-plementary rather than dominant. Managers should avoid assuming that AI tools alone will automatically generate a competitive advantage. Instead, AI deployment should be strategically aligned with organisational capabilities, routines, and governance systems. Effective integration, employee training, and workflow embedding determine whether GenAI initiatives translate into measurable long-term benefits.

The differentiated roles of exploratory and exploitative innovation suggest that firms should consciously manage dual innovation orientations. Exploratory initiatives, such as experimentation with emerging digital tools, are critical for driving the adoption of AI. However, exploration becomes associated with measurable performance outcomes when operationalised through technological implementation. In contrast, exploitative innovation directly enhances performance through improvements in incremental efficiency. Therefore, managers should balance experimentation with operational refinement rather than privileging one orientation at the expense of the other.

Environmental turbulence alone does not appear to be sufficient to trigger AI adoption. Rather than reacting solely to competitive or technological pressures, managers should focus on strengthening internal innovation culture, absorptive capacity, and digital compe-tencies. In shaping AI implementation decisions, capability development appears more decisive than contextual volatility.

Collectively, these implications suggest that sustainable competitiveness in AI-intensive B2B contexts requires structured governance, disciplined capability development, and deliberate integration of technological experimentation into operational systems. Man-agers who align governance, innovation orientation, and digital implementation are more likely to achieve enduring performance outcomes.

Although exploratory innovation shows the largest effect size in relation to GenAI adoption, ethical governance is prioritised due to its direct association with long-term competitive performance and its central role in shaping responsible AI deployment. Ac- cordingly, the prioritisation reflects managerial relevance rather than a strict ranking based on f2 values alone. Table 9 summarises these managerial priorities.

Note: Priorities reflect a combination of empirical effect sizes (f2), proximity to long-term competitive performance, and managerial relevance. Cohen’s thresholds for f2 are 0.02 (small), 0.15 (medium), and 0.35 (large).

8. Limitations and Future Research.

Several limitations should be acknowledged, which also create opportunities for future research. Additionally, although demographic variables such as firm size, sector, and managerial experience were collected, they were not included as control variables in the structural model. Given the study’s predictive orientation and the moderate sample size, theoretical parsimony was prioritised to preserve statistical power. Future research should incorporate control variables and multi-group analyses to further examine potential heterogeneity effects across organisational contexts.

First, the study relies on cross-sectional survey data collected from 104 managers and professionals in Portuguese B2B firms. Accordingly, reverse causality and alternative temporal orderings cannot be ruled out. Although the sample size exceeds the mini-mum threshold indicated by the a priori G*Power analysis for the estimated PLS-SEM model, the non-probabilistic convenience sampling strategy and the single-country context limit the external validity and generalisability of the findings. Although discrim-inant validity diagnostics support empirical distinctiveness across constructs, the high HTMT value between ethical governance and long-term competitive performance, together with the strong correlations involving SOP, suggests some degree of conceptual proxim-ity. Accordingly, the explanatory power of the model should be interpreted with caution. Furthermore, the sample composition may influence some relationships observed in the structural model. Specifically, 84.6% of the participating organisations are micro or small enterprises. While this distribution broadly reflects the structure of the Portuguese business landscape, smaller firms typically face stronger resource constraints, including limited financial capacity, specialised technical expertise and organisational slack for experiment-ing with emerging digital technologies. Consequently, the non-significant relationship between environmental dynamism and GenAI adoption (H9) may partly reflect the struc-tural characteristics of the sampled firms rather than a purely theoretical capability-based explanation. In highly resource-constrained organisations, the perception of environmental turbulence may not necessarily translate into technology adoption decisions. This model should be replicated and extended in future research using larger, probabilistic samples across multiple countries and industries, as well as longitudinal designs that allow stronger causal inferences. Future research should further examine whether the substantial ex-planatory power observed for sustainable organisational performance reflects, in part, the proximity between performance-related and capability-related constructs, using alternative model specifications and confirmatory factor approaches. Accordingly, reverse causality and alternative temporal orderings cannot be ruled out.

Second, all focal constructs were measured using self-reported perceptual scales, which raises the possibility of common method variance despite the use of procedural remedies and diagnostic tests. Although Harman’s single-factor test and collinearity diagnostics suggest that common method bias is unlikely to be a severe threat, the exclusive reliance on a single informant per firm remains a limitation. Because the survey instrument did not include a theoretically unrelated marker construct, marker-variable approaches such as the Lindell and Whitney procedure cannot be implemented post hoc in a fully defensible manner. Future studies could combine survey measures with archival or behavioural data (e.g., objective performance indicators, system usage logs) and adopt multi-respondent designs to triangulate perceptions across different managerial roles.

Third, Generative AI adoption is measured using perceptual items that capture the extent to which GenAI tools are used in B2B sales and decision-making processes. While this approach is consistent with prior research on IT assimilation and AI capa-bility, it does not distinguish between experimental pilots and deeply embedded, mission-critical applications, nor does it capture the organisation-wide scope of GenAI deployment (e.g., across multiple functions or business units). As a result, the construct reflects GenAI’s perceived presence in sales-related workflows rather than the organisa-tional level’s maturity or breadth of generative AI integration. Future research should develop more granular measures of GenAI adoption that differentiate the depth of integration (from pilots to core processes) and organisational scope, potentially combining perceptual scales with objective indicators of system usage and coverage.

Finally, exploratory and exploitative innovation are conceptualised as separate capa-bility dimensions rather than as a full-fledged ambidexterity configuration. Although this dual-orientation approach enables the examination of distinct structural effects, it does not capture the potential interaction or balance effects between exploration and ex-ploitation. Future research should investigate configurational and interaction-based models of ambidexterity in AI-intensive environments and explore whether the relation-ships identified here hold when sustainable organisational performance is conceptualised more broadly to include environmental and social dimensions in addition to long-term competitive sustainability.

9. Conclusions.

The findings advance understanding of how GenAI relates to sustainable organisa-tional performance in B2B contexts by integrating governance mechanisms, environmental dynamism, and dual innovation orientations within a unified explanatory framework.

The results suggest that long-term competitive sustainability in AI-intensive envi-ronments depends less on technological adoption alone than on the structured alignment of governance systems, innovation capabilities, and digital implementation mechanisms. Ethical governance emerges as the most influential direct organisational correlate of sus-tainable organisational performance, underlining the strategic importance of responsible AI oversight and institutional alignment.

Although GenAI adoption is positively related to performance, its role appears com-plementary rather than dominant. Technological implementation seems to generate greater value when embedded in governance stability and supported by appropriate innovation ca-pabilities. Through AI adoption, exploratory innovation shows an indirect relationship with sustainable performance, suggesting that experimentation becomes performance-relevant when translated into operational digital systems. By contrast, exploitative innovation relates to performance through incremental refinement mechanisms that are less dependent on technological experimentation.

These results challenge technology-deterministic assumptions and support a systemic perspective on AI-enabled sustainability. However, the robustness analysis also indicates that not all direct performance relationships are equally stable across specifications. While the strongest findings remain substantively consistent, some weaker links are sensitive to the inclusion of control variables and should therefore be interpreted with caution.

This study contributes to research at the intersection of institutional theory, the resource-based view, and organisational innovation in digitally transforming B2B environ-ments by clarifying these structural and mediating mechanisms. Overall, the study provides theory-guided evidence that firms can leverage GenAI responsibly and strategically in support of long-term competitive performance, while also showing that some performance-related relationships require further testing in alternative samples, with additional controls, and through longitudinal designs.

D.M. and M.A.; validation, D.M., M.A., R.M. and P.S.; formal analysis and investigation, D.M. and M.A.; writing, review and editing, D.M., M.A., R.M. and P.S. All authors have read and agreed to the published version of the manuscript.

Funding: The APC was funded by ISLA Santarém—Polytechnic University.

Institutional Review Board Statement: This study was approved by the Ethics Committee of the ISLA Santarém—Polytechnic University (code: 002/2025).

Informed Consent Statement: Informed consent was obtained from all subjects involved in the study via an online consent form prior to survey participation.

Data Availability Statement: The survey instrument is provided in the Appendix A. The anonymised dataset and supporting materials are available from the corresponding author upon reasonable request for academic purposes, subject to confidentiality and data protection considerations.

Conflicts of Interest: The authors declare no conflicts of interest.

Appendix A

Research Questionnaire

Table A1. All measurement items were assessed using a five-point Likert scale ranging from 1 (“strongly disagree”) to 5 (“strongly agree”).

Table A1. Cont.

Appendix B

Robustness Check Including Firm Size as a Control Variable

Table A2. PLS-SEM results for the robustness check including firm size as a control variable.

Note: The control model includes firm size as an exogenous variable affecting both endogenous constructs (GAIA and SOP). Coefficients are based on bootstrapped estimates.