The digital transformation paradox: how organizational and institutional barriers neutralize leadership investment in systemic banking

The banking sector is undergoing profound change driven by digital technologies, requiring major adjustments in operating models, governance, and customer value creation. However, despite sustained investment and strong executive commitment, many incumbent banks—particularly in Greece—remain at low levels of digital maturity. This reflects a digital transformation paradox in which leadership intent and financial resources often reinforce existing organizational routines rather than accelerate change. This challenges the common assumption that technological readiness and managerial support naturally lead to successful transformation. This study examines why digital transformation stalls and why barriers persist rather than diminish over time.

Prior research identifies multiple Digital Transformation (DT) inhibitors in incumbent banks: legacy IT systems, organizational rigidity, cultural resistance, regulatory complexity, and competitive pressure from digital entrants (Amarantou et al., 2018; Rahman et al., 2023; Reis and Melão, 2023; Vargas-Halabi and Yagüe-Perales, 2024; Verhoef et al., 2021; Vial, 2019). External pressures such as regulation further complicate transformation efforts (Buttigieg and Zimmermann, 2024; Taeihagh et al., 2021). Recent studies also reveal unintended effects of digital initiatives, including technostress that weakens productivity gains (Picazo Rodríguez et al., 2024) and efficiency trade-offs associated with fintech-driven innovation (Fersi et al., 2023). Together, these findings suggest that technology adoption alone is insufficient to achieve sustained transformation.

These tensions are not just technical but deeply organizational and cultural. Vargas-Halabi and Yagüe-Perales (2024) show that cultural consistency norms suppress the adaptability that transformation demands, mirroring the inertia patterns identified here. Nor does DT automatically raise productivity: technostress can offset gains, and work engagement shapes how effectively investment translates into outcomes, a pattern documented by Picazo Rodríguez et al. (2024) and consistent with the transformation paradox. In regulated financial institutions, compliance redirects innovation away from service improvement and toward conformity (Fersi et al., 2023). Digital ecosystem formation depends critically on governance archetypes (Krasyuk et al., 2022). Taken together, these studies show that DT outcomes hinge on organizational and institutional conditions, not resource levels. This study advances that conversation by explaining why, in a highly regulated sector, those conditions systematically neutralize leadership intent — even when investment and commitment are present.

Despite this progress, much literature remains descriptive, listing barriers without clarifying their relative importance or explaining how they translate into stalled transformation and limited value creation (Abdurrahman et al., 2024; N’Dri and Su, 2024; Warner and Wäger, 2019). While organizational change theories such as dynamic capabilities (Teece, 2007) and ambidexterity (O’Reilly and Tushman, 2013) highlight tensions between innovation and stability, they offer limited insight into how institutional pressures systematically reinforce inertia in highly regulated sectors like banking.

Synthesizing these aspects, three gaps emerge. First, the Technology–Organization–Environment (TOE) framework has been widely applied but rarely extended to prioritize inhibitors, leaving its outputs largely descriptive rather than actionable (Abdurrahman et al., 2024; Bueno et al., 2024; N’Dri and Su, 2024; Warner and Wäger, 2019). Second, while the Delphi method is well established in management and information systems research (Okoli and Pawlowski, 2004), it has seldom been applied to banking DT inhibitors and almost never combined with scaling techniques to establish hierarchies of importance. Third, Greece’s systemic banks remain under-researched, despite their distinctive mix of institutional constraints and supervisory pressures (Papadopoulos, 2020). Addressing this gap is particularly interesting because Greece represents an almost archetypal post-crisis banking environment. Between 2010 and 2020, the country experienced severe economic disruption, including the sovereign debt crisis, capital controls, and heightened European Central Bank supervision, while remaining part of the European Union’s broader institutional and regulatory framework. These conditions make Greece a useful setting for examining how institutional pressures influence digital transformation in highly regulated sectors.

This study addresses the digital transformation paradox through three contributions. Theoretically, this work builds on the TOE framework by introducing Value as a key factor. Institutional Theory explains how environmental pressures shape whether digital investments lead to successful outcomes. Methodologically, it introduces a Delphi–scaling hybrid design that transforms descriptive barrier lists into prioritized hierarchies. Empirically, it examines Greece’s systemic banks as a critical case characterized by post-crisis restructuring and intense regulatory supervision where paradox dynamics are particularly visible. Together, these contributions reframe digital transformation from a problem of execution to a problem of institutional design.

The study pursues three objectives: (1) to identify technological, organizational, environmental, and value-related inhibitors of digital maturity; (2) to prioritize them; and (3) to explain stalled transformation through the interaction of legacy constraints and institutional pressures.

The remainder of the paper is structured as follows. Section 2 reviews the literature and outlines research gaps. Section 3 presents the theoretical framework. Section 4 describes the methodology. Section 5 reports the results. Section 6 discusses the findings and implications. Section 7 outlines limitations and future research directions, and Section 8 concludes.

Scholars have approached digital transformation in banking from diverse angles, highlighting its complex nature. Some studies emphasize strategic alignment and technology adoption as transformation foundations (Gregory et al., 2019; Krasyuk et al., 2022), while others highlight the role of digital resources and internal design in enabling change (Verhoef et al., 2021). A parallel line adopts a customer-centric view, focusing on evolving consumer demands and continuous service innovation (Loonam et al., 2018). Collectively, this literature highlights that DT is not a linear or single-dimensional process but the product of interactions among technology, strategy, organizational structures, and customer engagement.

The TOE model is particularly relevant for capturing these dynamics (Tornatzky and Fleischer, 1990; Vial, 2019). Yet many studies remain descriptive, identifying inhibitors without evaluating their relative weight or strategic significance. For instance, while legacy IT and regulatory complexity are consistently cited (Bueno et al., 2024; Irani et al., 2023; Lyons and Zhu, 2024; Ulrich-Diener et al., 2025), their prioritization across contexts is rarely explored. This has produced “checklists” of barriers that lack managerial clarity. Other domains have advanced by integrating Delphi consensus with determinant ranking to establish factor hierarchies (Abdul et al., 2024; Aromal and Ma, 2023; Han et al., 2023; Passarelli et al., 2024), but in banking such methodological integration is still limited (Cech and Tellioglu, 2019).

Within the environmental dimension of TOE, competitive pressures have intensified dramatically. Fintechs, Neobanks (Palos-Sanchez et al., 2025), BigTech entrants, Embedded finance and Banking-as-a-service (BaaS) platforms, have accelerated innovation in payments, lending, and wealth management, forcing incumbents to rethink their positioning (Xu et al., 2025). Building on this, research shows transformation should be read through value creation, delivery, and capture—not counts of adopted technologies. Platform studies map success factors to that triad, showing how governance and data orchestration translate into value for users and capture for the focal firm—aligning with our Value dimension and moving assessment from inputs to outcomes (Pathak et al., 2022).

Regulation constitutes a parallel force, shaping DT both positively and negatively. Revised Payment Services Directive (PSD2) and General Data Protection Regulation (GDPR) have simultaneously opened markets and imposed heavy compliance obligations (McNulty et al., 2023; Taeihagh et al., 2021). The Digital Operational Resilience Act (DORA) adds risk management and resilience testing requirements that impose costs but also create opportunities for harmonized standards across the EU (Buttigieg and Zimmermann, 2024).

Despite these external forces, internal inhibitors remain the most persistent barriers to DT. Legacy IT infrastructures create integration challenges and lock banks into costly maintenance cycles (Osei et al., 2023; Reis and Melão, 2023). Cultural resistance and organizational inertia undermine adoption even when leadership is supportive (Amarantou et al., 2018; Antoniou and Papalamprou, 2020). Skills’ shortages, particularly in AI, data analytics, and cybersecurity, further constrain capability to use new tech. This reinforces the “Organizational” aspect of the TOE framework. It shows how internal capabilities turn technology adoption into actual value, justifying our TOE–V extension.

The Greek banking sector presents a particularly relevant context for such research. Since the early 2000s, systemic banks have pursued digital reforms (Challoumis and Eriotis, 2024), but the financial crisis and subsequent European Central Bank (ECB) supervision meant initiatives were fragmented, compliance-driven, and underfunded (Mavroudi et al., 2022). The result has been slower progress relative to more digitally mature markets (Edo, 2025; Manta et al., 2024). Comparative research confirms that Greek banks lag their European peers, though many have adopted hybrid “phygital” approaches to balance digital and physical channels (Bueno et al., 2024; Manta et al., 2024). The combination of crisis legacies, capital constraints, and regulatory intensity calls for context-specific research (Challoumis and Eriotis, 2024). The fact that these institutions survived the crisis, restored stability, and moved into a period of expansion makes the persistence of transformation barriers all the more worth examining.

While the Delphi method achieves expert consensus effectively, a recognized limitation is producing validated but flat lists without indicating relative importance, limiting actionable insight (Prokopenko et al., 2025; Strasser, 2019). To address this, methodologies from Multi-Criteria Decision-Making (MCDM) have been integrated with Delphi in fields like urban planning and sustainability (Abdul et al., 2024; Aromal and Ma, 2023; Han et al., 2023). Techniques such as the Analytic Hierarchy Process (AHP) are powerful but can impose a high cognitive burden on experts with pairwise comparisons for a large set of factors (Riabacke et al., 2012). A simple scaling survey, conversely, lacks the iterative consensus-building that ensures a shared understanding of complex constructs.

This study adopts and extends the TOE framework which conceptualizes adoption as influenced by three domains: technological, organizational, and environmental contexts. TOE is particularly suitable for banking, where transformation rarely stems from a single factor but rather emerges from interactions between internal capacities, external pressures, and institutional legacies (Irani et al., 2023; Vial, 2019).

Despite its utility, the TOE framework has a recurring limitation: it pays limited attention to post-adoption outcomes such as customer value, legitimacy, and competitive performance. Many applications classify adoption drivers but overlook whether transformation produces improved satisfaction or credibility (Kraus et al., 2022; N’Dri and Su, 2024). Adoption is too often treated as a static event rather than a dynamic process reshaping trust and legitimacy. To address this, scholars have extended TOE with dimensions such as sustainability and dynamic capabilities (Chatterjee et al., 2021; Chaudhuri et al., 2024; Muhic et al., 2023).

As stated, this study follows that path by incorporating a Value dimension, creating a TOE–V framework. The Value dimension emphasizes customer experience, personalization, accessibility, cost efficiency, and stakeholder outcomes such as regulatory legitimacy and investor confidence (Pathak et al., 2022). Similar extensions exist in related domains: Religia et al. (2025) added digital leadership to TOE for CRM adoption, while Abdurrahman et al. (2024) embedded dynamic capabilities to link transformation to performance outcomes in banking. Our TOE–V extension differs by focusing specifically on stakeholder value outcomes (rather than leadership or generic capabilities) and by integrating Institutional Theory to explain how regulatory and legitimacy pressures condition the relationship between adoption and value realization in highly regulated contexts. This explains the paradox’s persistence: banks may adopt technology primarily to satisfy coercive pressures and legitimacy expectations rather than improve efficiency, resulting in visible adoption without deep reconfiguration (Gegenhuber et al., 2022).

The Value extension ensures transformation is understood not only as adoption, but also as a process that generates demonstrable stakeholder value. Our TOE–V extension addresses this gap by reframing DT as a dual process of adoption and value realization, allowing inhibitors to be evaluated not only by their impact on implementation but also by their effect on stakeholder-level influence. This extension improves the explanatory scope of TOE in regulated contexts such as banking. We expand the established TOE framework by adding value outcomes. This approach allows for a more comprehensive assessment of a bank’s digital maturity, shifting the focus from the initial adoption of technology to the realization of success (Figure 1).

Institutional Theory complements and extends the TOE–V framework by explaining how external pressures shape organizational behaviour in complex, highly regulated environments (DiMaggio and Powell, 1983; Gegenhuber et al., 2022; Scott, 2001). It captures the coercive, normative, and mimetic forces that drive organizations to conform for legitimacy. These forces interact with inhibitors in non-linear ways; for instance, regulatory ambiguity can act as a barrier, while clear mandates become enablers. Institutional Theory further contextualizes TOE–V by placing inhibitors within broader institutional dynamics, highlighting how legitimacy concerns, regulatory oversight, and field-level norms shape both the trajectory and the pace of DT. In post-crisis Greece, strong regulatory oversight and European standards increase pressure on banks to show resilience and modernization, even when efficiency gains are unclear.

Existing frameworks each explain part of the picture but not all of it. Dynamic capabilities theory (Teece, 2007) assumes managers can freely choose which capacities to build — an assumption that breaks down under coercive regulatory oversight, where some capabilities are simply not permissible. Ambidexterity theory (O’Reilly and Tushman, 2013) captures the tension between exploiting existing routines and exploring new ones but says little about how institutional logics redefine what “exploration” even means for a bank operating under ECB supervision. Standard TOE goes further by mapping adoption antecedents across multiple levels, yet it stops short of asking whether adoption actually delivers value to customers, regulators, or investors. The Value dimension in TOE–V addresses this directly, shifting the measure of maturity from what is implemented to what is demonstrably delivered. Institutional Theory then supplies the missing explanation for why that delivery so often falls short: where legitimacy-seeking dominates efficiency-seeking, adoption becomes symbolic rather than substantive. Together, these elements explain why transformation paradoxes persist even when resources and leadership are in place.

This integrated TOE–V and Institutional Theory model (Figure 1) provides the basis for the empirical design and the operationalization of propositions.

Drawing on the TOE–V framework and Institutional Theory, the study proposes a set of research propositions to explain how technological, organizational, environmental, and value-related factors act as inhibitors or catalysts of DT in incumbent banks. These propositions examine not only the direct drivers of adoption but also the institutional conditions under which adoption leads (or fails to lead) to stakeholder value, addressing the core tension between legitimacy-seeking and efficiency-seeking behaviours presented above.

The TOE–V framework, enriched with Institutional Theory, provides a comprehensive basis for analysing DT in incumbent banks. TOE captures multi-level antecedents of adoption; the Value dimension emphasizes stakeholder outcomes; and Institutional Theory highlights the institutional pressures shaping organizational decisions. Together, these perspectives underpin the study’s empirical design and the prioritization of inhibitors in Greek systemic banks.

The following table (Table 1) provides a concise overview of the research propositions, linking each proposition to its theoretical foundation, the related constructs, and the expected effects.

This study adopted a sequential exploratory design combining expert consensus with quantitative prioritization. The research used a two-round Delphi process followed by a determinant-scaling survey. The Delphi method enabled structured expert judgment through anonymity and controlled feedback, reducing dominance effects and supporting convergence across rounds (Hsu and Sandford, 2007; Okoli and Pawlowski, 2004; Strasser, 2019). Two rounds typically suffice in bounded domains (Keeney et al., 2001).

The initial pool of items was deductively derived from an a priori TOE–V construct map developed from the DT and banking literature (see Section 2). Factors included technical debts and integration capability, leadership and governance, cultural and skill-related aspects, competitive and market pressures, and value outcomes such as usability and personalization. In the first round, experts rated these items using a five-point agreement scale (1 = fully agree, 5 = fully disagree), with higher disagreement indicating that the factor functioned as an inhibitor. Between rounds, items were refined using transparent criteria: those with interquartile range (IQR) greater than 1.0 or standard deviation (SD) greater than 1.5 were treated as ambiguous and reworded, split, or merged to reduce interpretive overlap. Cronbach’s α was applied as a heuristic indicator of construct coherence within TOE–V groupings, consistent with its diagnostic use in Delphi refinement rather than as a psychometric test (Cronbach, 1951; Tavakol and Dennick, 2011).

In the second round, participants were provided with aggregated medians, IQRs, and anonymized qualitative comments from Round One. This feedback let experts reconsider their assessments while preserving independence of judgment. Items reaching IQR ≤ 1.0 or SD ≤ 0.8 were interpreted as convergent, while those failing to reach this threshold were retained for qualitative interpretation rather than forced into unreliable quantitative consensus (Warner, 2014). Two rounds were sufficient to achieve stability (Barrios et al., 2021), as no major new inhibitors were introduced after Round One and consensus tightened significantly on the refined items.

Following the Delphi phase, inhibitors were prioritized using the determinant-scaling approach described in the methodology section. This approach maintained accuracy while minimizing fatigue (Passarelli et al., 2024; Riabacke et al., 2012; Strasser, 2019). Similar Delphi–scaling designs have been successfully applied in previous management, information systems, and sustainability studies (Abdul et al., 2024; Aromal and Ma, 2023; Han et al., 2023). The design thus moved from identifying inhibitors to establishing their relative strategic significance, generating outputs that are both rigorous and actionable.

Fifteen experts were selected, a size consistent with established Delphi guidelines prioritizing expert knowledge over large samples, indicating that panels of 10–20 experts (Hsu and Sandford, 2007; Okoli and Pawlowski, 2004) typically achieve saturation in bounded domains, (Strasser, 2019). Participants brought ≥10 years of direct accountability for digital-transformation initiatives (e.g. Chief Information Officers, Heads of Transformation, product and operations managers, senior consultants). Sampling rigor was ensured by balancing institutional types (systemic banks, fintechs, advisory firms) and functional domains (technology, operations, risk/compliance, strategy, product). To mitigate bias, participation was anonymous, feedback was aggregated, and selection targeted cross-functional diversity to limit dominance effects. Panel demographics (gender, age, institution, expertise) are detailed in Table A.1 of the Online Supplementary Material.

Potential sources of bias were addressed through several design features. An executive-heavy panel can understate cultural frictions encountered at ground level, while advisors may emphasize feasibility constraints and fintech practitioners may highlight customer-facing agility. To mitigate these risks, different constituencies were combined, anonymity preserved throughout, and only aggregated distributions and anonymized comments were presented as feedback between rounds. Using the same panel for both Delphi and scaling could introduce anchoring effects; this was addressed by emphasizing in the instructions that the scaling survey was a distinct judgment task focused on relative importance rather than repetition of earlier agreement ratings. While this limits statistical generalizability, the design prioritizes analytically grounded insight appropriate for exploratory research in institutionally dense contexts. Nonetheless, the reliance on a single-country sample and self-selection remain limitations, as acknowledged in Section 7.

The methodological progression is illustrated in Figure 2, which depicts the structured sequence from initial item generation and Round One ratings, through refinement and Round Two convergence, to the scaling survey and final analysis.

By integrating consensus-building with comparative weighting, the research design provides both definitional clarity and a prioritized hierarchy of inhibitors. This sequential approach enhances the explanatory strength of TOE–V, ensures methodological rigour through transparent thresholds and reliability checks, and generates insights that are both theoretically relevant and managerially actionable.

To move from consensus-building to actionable prioritization, the study employed a two-step analytic strategy: Delphi responses were analysed using dispersion and reliability measures to ensure definitional clarity, then the same panel completed a determinant-scaling survey to rank inhibitors by relative importance.

Delphi responses were initially analysed using descriptive statistics, interquartile ranges (IQR), and standard deviations (SD) to evaluate convergence. High consensus was operationalized as IQR ≤ 1.0 or SD ≤ 0.8, moderate consensus as IQR up to 1.5, and low consensus as values above 1.5 (Barrios et al., 2021; Warner, 2014). Movement from low or moderate toward high consensus across rounds was interpreted as evidence that iteration clarified expert views, consistent with the principle that changes between rounds represent refinement rather than instability (Hsu and Sandford, 2007). Internal consistency within the TOE–V domains was assessed using Cronbach’s α, used as a diagnostic indicator of interpretive consistency rather than as a formal psychometric test (Cronbach, 1951; Okoli and Pawlowski, 2004; Tavakol and Dennick, 2011).

Round One produced low α values in some domains (e.g. Technology, which was interpreted as signals of ambiguous wording or divergent interpretation). These results guided the rewording, splitting, or merging of items between rounds. By Round Two, α coefficients improved, approaching the conventional 0.7 benchmark, reflecting increased clarity and shared interpretation. Median and IQR values remained the primary criteria for consensus at the item level, while α provided a practical check on interpretive coherence. Raw data and questionnaire context for both rounds are available in Tables A.2 and A.4 of the Online Supplementary Material.

Particular attention was paid to the organizational culture construct. Delphi results indicated low internal consistency (α = 0.50), confirming that culture is complex and context-dependent. Yet qualitative comments and the subsequent scaling exercise demonstrated strong consensus on culture’s overall importance as an inhibitor. Accordingly, culture was treated as a hidden influence: its significance was captured through expert judgment rather than forced into a fragile quantitative scale. This approach aligns with calls to combine survey-based consensus with interpretive methods for constructs resisting single-metric reduction (Hsu and Sandford, 2007; Powell et al., 2021).

Consensus thresholds and reliability benchmarks are reported alongside domain-level outcomes; the construct-to-item mapping is in Table 2.

In Round One, fifteen positively worded items were formulated and allocated to the four domains of the TOE–V framework (Technological, Organizational, Environmental, and Value). To ensure conceptual precision and methodological consistency, all items were reviewed and reformulated to maintain a uniform Likert direction, removing the need for any reverse coding. During Round Two, three additional items were proposed, and three initial items were removed from analysis due to either low expert consensus, conceptual overlap, or insufficient construct clarity. Expert evaluations (median agreement levels and dispersion measures) guided the iterative refinement of the questionnaire, resulting in a final, validated instrument consisting of fifteen stable items (Q1–Q15). These items retained identical codes across both Delphi rounds and were aligned with their respective TOE–V domains, as presented in Table 2. This coding structure ensured traceability across analytical stages and provided a stable reference framework for subsequent data analysis.

The second analytic stage introduced a determinant-scaling survey to move from validated constructs to comparative prioritization (Table A.6 of the Online Supplementary Material). Each expert rated the importance of the 14 refined inhibitors on a seven-point scale (1 = “lowest importance,” 7 = “highest importance”). This procedure is conceptually similar to AHP (Abdul et al., 2024; Han et al., 2023) but avoids the complexity of pairwise comparisons, which impose significant cognitive burden (Strasser, 2019). Direct ratings were appropriate given the bounded item set, the seniority of respondents, and the need to minimize fatigue (Cech and Tellioglu, 2019). Comparable Delphi–scaling hybrids have been applied in urban planning, healthcare, and sustainability (Abdul et al., 2024; Aromal and Ma, 2023; Han et al., 2023); Direct rating approaches have been validated in similar Delphi–MCDM hybrids, offering reliable prioritization while avoiding the cognitive overload associated with full pairwise comparison matrices (Riabacke et al., 2012). This study extends the approach to DT in banking.

Analysis of the scaling data included computation of mean and median importance scores, supported by SD and IQR to assess dispersion. Inter-rater agreement was assessed with Kendall’s coefficient of concordance (W) (tie-corrected), using the χ² approximation for significance (χ² = m(n−1) W, df = n−1). Pearson correlation coefficients across items were also examined to evaluate distinctiveness; low-to-moderate correlations confirmed that inhibitors contributed unique information rather than duplicating each other. Results were visualized in a correlation heatmap (Figure 3) to illustrate interdependencies and potential clustering among inhibitors.

All analyses were performed in Python using Pandas, NumPy, and SciPy. Given the modest but expert-intensive panel, emphasis was placed on practical significance and prioritized rankings rather than on null-hypothesis significance testing. This approach emphasizes expert judgment and delivers a transparent prioritization of inhibitors without overstating statistical inference.

Two methodological choices deserve clarification. The use of the same expert panel in both the Delphi and scaling phases introduces a potential risk of anchoring bias. This approach was retained, however, because it ensures continuity: the shared understanding of sector-specific constructs developed during the Delphi phase strengthens the internal validity of the subsequent prioritization. To mitigate anchoring effects, the scaling exercise was explicitly framed as a separate task focused on assessing relative importance for resource allocation. Anonymity was preserved throughout, and only aggregated feedback was provided between rounds.

Second, the determinant scaling approach avoids pairwise comparisons but may limit fine-grained differentiation among closely related inhibitors. The findings are analytically rather than statistically generalizable, specific to post-crisis, highly regulated banking. Finally, the expert panel (n = 15) reflects the concentrated structure of the Greek banking sector and represents depth of expertise rather than broad population coverage. However, the analytical implications deserve explicit reflection. An executive-heavy panel assesses strategic and governance inhibitors with high validity but may underweight operational frictions—cultural resistance, skills gaps—most acutely felt at ground level. Consultant panellists may frame barriers in feasibility terms; fintech practitioners may foreground customer-agility constraints that differ from legacy incumbents. These professional lenses introduce perspectival bias that anonymity mitigates but does not eliminate: the inhibitor hierarchy captures senior decision-maker judgment and may compress ground-level barriers. Future research should triangulate senior, middle-management, and operational perspectives, supplemented by qualitative methods—ethnographic observation or practitioner interviews—to surface the lived experience of transformation.

The Delphi study proceeded in two rounds to identify and prioritize DT inhibitors. Round One revealed divergence in technological readiness and cultural assessments, contrasting with strong consensus around leadership commitment. Technological baselines ranged from modular to monolithic architectures, reflecting uneven post-crisis investment. Cultural variation was equally marked: agile practices coexisted with hierarchical, risk-averse norms (Warner and Wäger, 2019). Cronbach’s α and dispersion measures reflected this heterogeneity, indicating weaker reliability in the technological and cultural domains compared to organizational and environmental dimensions.

The second Delphi round achieved markedly higher consensus and reliability. Item rewording, construct consolidation, and clearer definitions improved interpretive alignment across the panel. Technological and organizational domains displayed significant gains in internal consistency, with legacy IT, limited integration capabilities, and cultural inertia consistently identified as the most critical inhibitors. In particular, culture emerged not as an isolated construct but as a latent moderator, amplifying or dampening the influence of other inhibitors. These findings align with research on organizational culture as a “black box” influencing innovation (Vargas-Halabi and Yagüe-Perales, 2024), where technological rigidities and cultural norms are cited as structural barriers to DT (Reis and Melão, 2023; Vial, 2019). Detailed data on means, medians, SDs, IQRs are available in Tables A.3 and A.5 of the Online Supplementary Material.

The iterative Delphi process clarified constructs and strengthened expert consensus, as detailed in Table 3. The results demonstrate clear progression from Round One to Round Two following feedback, rewording, and item consolidation. This strengthening is evidenced by the rise in inter-rater agreement from moderate (Kendall’s W = 0.329, χ²(14) = 69.03, p < 0.001) in Round One to stronger consensus (W = 0.434, χ²(17) = 110.70, p < 0.001) in Round Two. Reliability also improved across all TOE-V domains, most markedly in the Technological construct (α = 0.62 → 0.91). Organizational reliability rose (α = 0.79 → 0.86), while Environmental (α = 0.80 → 0.84) and Value (α = 0.76 → 0.82) stabilized at strong levels. Cultural indicators, showing greater dispersion, are interpreted as qualitative moderators rather than standalone constructs. Taken together, the refined 15-item TOE-V instrument demonstrates stronger internal consistency and clearer construct separation, providing a sound statistical basis for the subsequent prioritization analysis.

With consensus achieved, the second stage of analysis focused on prioritizing inhibitors using a determinant-scaling survey. This method assigned relative importance to the validated constructs (Table 4), transforming the findings from a descriptive list into a ranked hierarchy that supports strategic decision-making (Tornatzky and Fleischer, 1990).

Leadership emerged as the most decisive enabler: top management support received the highest rating (mean = 6.13), followed by adequate financial resources (mean = 5.40), though experts noted budgets are often consumed by routine maintenance—“keeping the lights on”—rather than innovation (Ulrich-Diener et al., 2025).

Customer-centric outcomes were also central. Ease of use (mean = 5.73) was identified as a critical benchmark for maturity, underscoring the shift from technology adoption metrics to value delivery for customers and stakeholders (Verhoef et al., 2021). Workforce-related factors, such as employee engagement (mean = 5.08), were also significant, reflecting the importance of ability to learn and adapt and skills in scaling transformation initiatives (Loonam et al., 2018). Organizational culture (mean = 4.87) remained a key determinant of agility and execution speed.

Technological constraints persisted as major barriers. Technology readiness (mean = 3.00) and integration (mean = 3.87) continued to impede progress, demonstrating how legacy infrastructures and disconnected systems undermine transformation outcomes. Regulatory requirements (mean = 3.47) ranked lower, indicating that while compliance imposes complexity, internal capability gaps pose a greater challenge. This hierarchy reveals a fundamental imbalance: leadership commitment and financial resources are present, but their impact is offset by persistent technological fragmentation and cultural inertia. Table 4 presents a summary while the detailed, full-scale data and measurements are presented in Table A.7 of the Online Supplementary Material.

Kendall’s coefficient of concordance (tie-corrected) shows moderate-to-strong agreement in the final scaling stage across 14 items and 15 experts (W = 0.580; χ² = 113.09; df = 13; p < 0.001). Because the Friedman test statistic equals χ² = m(n−1)W, this significance indicates that between-item differences are systematic, providing good evidence that the prioritized ranking reflects genuine consensus rather than noise.

To preserve methodological traceability, Table A.8 of the Online Supplementary Material provides a Ranking–Delphi code crosswalk linking the 14 ranked items to the validated 15-item Delphi instrument. Non-one-to-one mappings (e.g. Top Management Support) are flagged to prevent code drift when comparing the results of § 5.3 with those of § 5.1 and § 5.4.

Correlation analysis was conducted to explore the relationships between inhibitors and enablers, revealing systemic interdependencies that explain why many transformation initiatives fail to scale effectively.

Regulatory requirements correlated positively with technology readiness and integration: clear mandates—PSD2, GDPR, DORA—accelerate modernization and Application Programming Interface (API) development (Buttigieg and Zimmermann, 2024). Within the organizational domain, resources were positively associated with employee engagement, reflecting how investment in training and incentives fosters workforce activation.

Leadership support, however, showed only weak direct associations with other variables, underscoring that strategic intent alone is insufficient without complementary governance, incentives, and execution mechanisms. Organizational culture displayed a strong negative correlation with ease of use, indicating that entrenched routines and siloed structures significantly degrade customer experience. External stakeholder pressure also correlated negatively with customer-centric outcomes, suggesting that competing institutional demands may introduce operational complexity.

These findings confirm that inhibitors are interconnected. Progress in one area cannot offset stagnation in others. A coordinated approach addressing infrastructure renewal, cultural change, and value creation is therefore essential.

The propositions (Section 3.2) were evaluated using three evidence sources: Delphi Round Two consensus, scaling-survey prioritization scores, and inter-factor correlation patterns. Propositions were classified as confirmed, partially confirmed, or not confirmed based on consistency of empirical support (Table 5). In what follows, empirical evidence is presented first for each proposition outcome, followed by its analytical interpretation. This sequencing maintains a clear boundary between what the data show and what they mean within the TOE–V framework.

Overall, six propositions are fully confirmed and two are partially confirmed; none are disproved. Technological and organizational propositions received the strongest empirical support. Legacy IT constraints (P1), integration capability (P2), top-management support (P3), and digital governance structures (P4) were consistently identified as decisive determinants of DT maturity. Environmental influences exhibited conditional effects. Competitive pressure from fintech and BigTech entrants (P6) was confirmed as a significant external driver of transformation investment. Value-related capabilities were also central. Personalization capability (P7) showed a strong positive association with DT maturity, reinforcing the relevance of the Value dimension within TOE–V. Proposition P5 (regulatory influence) was partially confirmed because regulation exhibited a dual effect: ambiguous or evolving mandates delayed modernization, while clear and enforceable frameworks (e.g. PSD2, DORA) accelerated integration and infrastructure renewal. Proposition P8 (social influence) was also partially supported, as peer imitation and industry benchmarking were present but significantly weaker than coercive regulatory pressures and direct competitive forces (N’Dri and Su, 2024), suggesting limited normative diffusion in highly regulated institutional environments.

The results collectively reveal a deeper paradox within incumbent banking: despite leadership commitment, financial resources, and formal governance, transformation maturity remains limited. This challenges the linear assumption that resource input directly produces digital capability. The study reframes DT as an institutional-equilibrium challenge: a self-regulating state where control, compliance, and legitimacy mechanisms neutralize strategic intent, turning innovation investment into stability maintenance.

Three findings constitute the core contribution of this study. First, DT stalls because the institutional mechanisms that ensure compliance and continuity absorb innovation capacity — not for lack of resources or leadership commitment. This reframes transformation failure as a structural property of regulated sectors, not a managerial deficit, with direct implications for how executives and policymakers diagnose stalled programmes. Second, the TOE–V framework shows that the Value dimension is not merely additive: legacy systems and cultural inertia do not only slow adoption, they degrade the customer-facing and stakeholder outcomes that constitute genuine maturity. Organizations measuring DT success in input terms will consistently overestimate progress. Third, regulatory clarity functions as a modernization lever: specific, enforceable mandates—PSD2, DORA—accelerate integration and infrastructure renewal, a finding with direct policy relevance. Together, these insights shift the conversation from “how do we invest more?” to “how do we recalibrate the institutional equilibrium?”—a question relevant to any post-crisis or highly regulated banking environment.

The transformation trap can also be seen as a failure of structural ambidexterity. Banks try to introduce new digital solutions at the front end, but they remain tied to old IT systems and compliance routines that consume resources and limit flexibility. Institutional pressures further strengthen this imbalance by favouring stability and legitimacy over experimentation. Banks therefore struggle to balance innovation with reliance on existing systems in highly regulated environments (O’Reilly and Tushman, 2013; Teece, 2007).

DT does not fail because of missing resources but because the mechanisms of control, compliance, and continuity that ensure legitimacy also suppress adaptation. Leadership commitment signals strategic intent, but its impact is neutralized by technical debts and entrenched culture (Vargas-Halabi and Yagüe-Perales, 2024). Core modernization remains the dominant bottleneck: front-end upgrades layered onto outdated architectures create costly workarounds and suppress organizational learning (Reis and Melão, 2023; Słoniec and González Rodriguez, 2018). Budgets absorbed by legacy maintenance divert investment away from innovation—consistent with path-dependent resource allocation theory (Teece, 2007; Warner and Wäger, 2019). In this context, modernizing core platforms—via phased renewal, API-first decoupling, or cloud migration—is a critical enabler. This reveals a key contradiction in transformation theory: governance structures that ensure reliability and compliance also limit the process and technology changes needed for deep renewal.

Culture influences all domains. Despite executive sponsorship, resistance in middle management and among less digitally literate staff creates a persistent gap between top-down intent and bottom-up execution (Amarantou et al., 2018; Rahman et al., 2023). Qualitative evidence echoes the phenomenon of ceremonial adoption, where pilot projects fail to become embedded organizational routines (Das et al., 2018; Ulrich-Diener et al., 2025; Vial, 2019). Persistently weak reliability scores for cultural indicators support treating culture not as a standalone construct but as a contextual moderator influencing adoption speed, scaling, and customer outcomes (Barrios et al., 2021; Warner, 2014).

Environmental forces are conditional rather than deterministic. Regulation partially functions as an inhibitor (P5): ambiguity delays progress, while clear mandates—PSD2, GDPR, DORA—prompt upgrades and resilience investments (Bueno et al., 2024; Buttigieg and Zimmermann, 2024; Loonam et al., 2018; Taeihagh et al., 2021). This duality aligns with Institutional Theory: coercive and normative pressures can constrain or induce based on clarity (DiMaggio and Powell, 1983; Gegenhuber et al., 2022). Competitive pressure from fintechs and BigTech (P6) remains a significant external driver, but without internal readiness — including core renewal, integration capability, and cultural alignment — responses risk remaining reactive and defensive.

The findings therefore suggest that DT maturity depends less on adding resources and more on how organizational practices interact with institutional pressures. This reinterpretation shows the role of inhibitors as structural features of balance rather than temporary barriers. The TOE–V framework, when viewed through this lens, becomes not only a diagnostic model but also a tool for understanding how equilibrium forms and persists.

Extending the TOE framework with a Value dimension (TOE–V) proves decisive. Proposition P7 is supported: ease of use, personalization, and service quality emerge as critical maturity benchmarks, shifting evaluation from adoption counts to measurable stakeholder outcomes (Das et al., 2018; Palos-Sanchez et al., 2025; Rahman et al., 2023). This shift aligns with service-dominant logic and contemporary digital customer-experience theory (Kraus et al., 2022; N’Dri and Su, 2024; Vial, 2021). Social influence (P8) is partially supported: peer imitation exists but is weaker than coercive regulation and customer expectations, diverging from individual-level models like the Unified Theory of Acceptance and Use of Technology (UTAUT) (Venkatesh et al., 2003).

Finally, the interdependencies revealed by the correlation analysis explain why incremental fixes often fail. Culture’s negative relationship with ease of use and regulation’s positive link with readiness and integration show that inhibitors are systemic, not isolated. Progress in one domain cannot compensate for stagnation in another. These dynamics are intensified in the Greek context, where post-crisis restructuring, ECB supervision, and profitability pressures bias spending toward compliance and continuity, leading to phygital equilibria that sustain service delivery but delay deep renewal. In that sense, the Greek systemic banks succeeded in doing what the crisis demanded of them; the question this study raises is whether the institutional conditions that enabled survival are the same ones that now need to be recalibrated to enable renewal. Comparable dynamics are likely in other post-crisis European banking systems (e.g. Italy, Portugal), where institutional legacies constrain experimentation and scale-up.

The findings highlight implications that could extend well beyond Greece. As a European Union member state, Greece operates within a stable supranational architecture governed by ECB supervision, shared regulatory directives, and common financial standards, yet lived through one of the most severe economic crises in recent European history. That tension between institutional belonging and domestic disruption is not uniquely Greek, and comparable dynamics may well be present in other post-crisis or heavily supervised financial systems.

Three priority areas emerge. First, core modernization must be treated as a precondition, not a discretionary project: legacy systems constrain scalability and innovation. Leaders should approach core renewal as a multi-year build: phased domain replacement, API-first decoupling, and progressive cloud migration (Kane et al., 2015; Manta et al., 2024). Crucially, modernization budgets must be explicitly ring-fenced to prevent reabsorption into routine “keeping the lights on” maintenance (Bharadwaj, 2000).

Cultural change must be anchored in customer-outcome metrics. The negative correlation between culture and ease of use shows that cultural inertia directly worsens customer experience. Transformation efforts should tie incentives and routines to tangible metrics such as onboarding time, task-completion rates, and digital journey satisfaction by Net Promoter Score (NPS). Quick wins that couple back-end decoupling with visible front-end improvements can build momentum and demonstrate value (Acosta-Prado et al., 2024). Dedicated governance structures—such as Chief Digital Officer mandates and transformation offices—are necessary to maintain strategic alignment across organizational silos (Singh and Hess, 2020).

Regulation can be used as a modernization catalyst rather than viewed purely as a constraint. Regulatory clarity can accelerate modernization, while ambiguity maintains inertia. Banks can adopt a proactive stance through early supervisory dialogue and “regulatory sprints” that align legal, risk, and engineering teams before scaling initiatives. Policymakers, in turn, can amplify positive effects by designing clear, enforceable mandates (e.g. PSD2, DORA) that lock in integration milestones rather than diffuse effort (Buttigieg and Zimmermann, 2024). Shared regulatory utilities—digital identity or Know Your Customer (KYC) platforms—can lower duplicate costs and raise sector-wide resilience. From a managerial perspective, the key issue is not simply the scale of investment in digital transformation but the organisational and institutional conditions that determine whether those investments translate into meaningful results. The Greek systemic banking case demonstrates that these conditions can be identified, prioritised, and addressed through measures such as regulatory clarity, governance restructuring, and cultural realignment.

The study’s theoretical significance lies in reframing digital transformation as a balance between change and stability. Instead of being mere obstacles, inhibitors such as legacy systems and regulatory compliance act as stabilizers that absorb innovation resources to preserve organizational legitimacy. This shifts the focus from listing barriers to explaining why transformation inertia persists even when the necessary enablers are present. The two-stage design—Delphi consensus followed by quantitative ranking—addresses a common limitation of inhibitor studies by adding comparative weightings to expert-validated lists, producing a usable hierarchy rather than a catalogue (Okoli and Pawlowski, 2004). The design is transferable to other regulatorily dense domains, preserving validity while improving decision relevance (Aromal and Ma, 2023; Kerpedzhiev et al., 2021).

The empirical results also refine the TOE–V framework in three specific ways. The strong support for P7 is perhaps the most theoretically significant: usability, personalization, and stakeholder relevance emerge as maturity benchmarks that adoption-focused TOE models do not capture, confirming that the Value dimension shifts the framework from measuring inputs to measuring outcomes — consistent with service-dominant theory (Acosta-Prado et al., 2024; Filotto et al., 2021; Religia et al., 2025). The partial confirmation of P5 adds a further nuance: regulatory clarity can act as a modernization catalyst rather than a constraint, extending Institutional Theory by showing that coercive pressures enable transformation when they are specific and enforceable, not merely when they are present (Buttigieg and Zimmermann, 2024; DiMaggio and Powell, 1983; Gegenhuber et al., 2022; Taeihagh et al., 2021). Finally, the joint support for P1–P4 shows that technological and organizational constraints do not operate independently; together, they moderate how investment translates into value, refining TOE–V as a model of conditional value realization rather than linear capability accumulation.

Culture’s mixed metrics confirm its function as a latent moderator; future research should integrate survey data with ethnography and longitudinal cases to capture cultural dynamics without sacrificing validity (Powell et al., 2021). What the Greek case makes clear is that institutional legacies matter: the same inhibitors do not carry the same weight in every context, and the TOE-V framework could be most useful when applied with that sensitivity.

This study explains why modernization often reproduces stability: resource and leadership inputs are absorbed by stability-preserving processes, namely legacy systems, compliance governance, and risk-averse culture, that sustain an institutional equilibrium. In TOE–V terms, inhibitors act as forces moderating the pathway from inputs to Value outcomes. The Delphi–scaling priorities and correlations show that clarity in regulation and cultural flexibility are the pivotal levers. Consequently, DT is not linear capability accumulation but a managed balance; progress follows when that balance is deliberately shifted. From a theoretical perspective, the results suggest that existing frameworks such as dynamic capabilities, ambidexterity, and the standard TOE model are insufficient on their own because they assume that organisations can readily mobilise resources and leadership commitment to drive change. In heavily regulated, post-crisis environments, that assumption is often constrained by institutional pressures. The integration of TOE-V with Institutional Theory helps explain how those pressures shape the relationship between digital investment and transformation outcomes.

This study’s scope is limited to the Greek banking sector, whose post-crisis structure and concentrated market offer valuable context but constrain generalizability. Replicating the design in other settings would test the robustness of the inhibitor hierarchy and the TOE–V extension. The expert panel (N = 15) reflects depth consistent with Delphi but may over-represent senior perspectives; segmented panels could triangulate views on culture and customer experience. Future studies should continue to interrogate the transformation paradox by examining how equilibrium conditions shift when external pressures or regulatory frameworks change.

The single-country design limits statistical generalizability. However, the focus on a post-crisis, highly regulated environment enhances analytical generalization by identifying mechanisms likely to operate in comparable banking systems. Future research should test whether the inhibitor hierarchy and TOE-V dynamics hold across similar contexts.

Future work should pair expert assessments with observable outcomes, develop a comparable maturity index, and use longitudinal designs to track inhibitor salience. Culture’s weak reliability motivates mixed-method approaches treating it as a latent moderator. Together, these directions would refine TOE–V and advance a cumulative evidence base.

This study moves beyond descriptive barrier lists by identifying not only what inhibits digital transformation in banking but also the relative importance and systemic interaction of those inhibitors. It reveals transformation as a stable system that reinforces existing practices rather than a linear process of resource accumulation—one in which stability-preserving forces absorb innovation investment. Internal constraints—legacy IT, cultural inertia, and misaligned budgets—emerge as the primary factors of this equilibrium. External forces like regulation and competition act conditionally, accelerating renewal only when internal capabilities permit. This dynamic is particularly visible in the empirical setting studied. The study contributes a Delphi–scaling hybrid method that turns barrier lists into actionable priorities, and an extended TOE–V framework integrated with Institutional Theory to explain how adoption attempts translate into stakeholder value. For practice, breaking the paradox requires treating core modernization as foundational, anchoring culture to customer outcomes, and leveraging regulation as a catalyst. Ultimately, DT is less about adopting technology than about deliberately recalibrating the equilibrium between change and continuity.

What this means in practice follows the TOE–V dimensions directly. On the Technological and Organizational dimensions, executives should ring-fence core modernization budgets and tie cultural change programmes to measurable customer-outcome metrics—these are necessary conditions, not optional refinements. For the Environmental dimension: regulatory design is itself a transformation instrument, and clear, time-bound mandates (PSD2, DORA) create modernization pressure that individual banks cannot absorb or redirect. For the Value dimension: progress should be assessed by stakeholder-visible outcomes—usability, personalization, regulatory legitimacy—not by adoption counts. For the research community, the paradox warrants longitudinal investigation into how equilibrium conditions shift over time. TOE–V provides a replicable lens for that agenda. In doing so, it completes the reframing of digital transformation as a challenge of institutional design rather than merely of execution. Prior frameworks explain digital transformation through capability, alignment, or adoption lenses. This study shows that in systemic banking those lenses are insufficient because they assume that having resources and leadership commitment is enough to drive change. In regulated institutional environments, that assumption breaks down: structural constraints absorb investment before it reaches its intended outcomes. Institutional equilibrium, not capability gaps, is the primary explanatory variable. That is what this study adds, and that is what future research needs to address.

This article does not contain any experiments conducted by the authors on animal or human participants.

All participants in this non-interventional study were informed of the research purpose, assured of their anonymity, told how their data would be used.

ChatGPT (OpenAI) was used for language refinement and formatting support; the authors reviewed and approved all final text.

The supplementary material for this article can be found online.