Flexible IT Architecture Boosts Digital Transformation, But Decentralized Governance Can Undermine It

A flexible IT architecture strongly builds the capabilities firms need for digital transformation, but pairing that flexibility with decentralized IT governance can quietly erode some of the benefit, according to a new survey-based study of 188 U.S. firms. The research, published in Information Systems Frontiers, finds that loosely coupled, standardized, and scalable IT systems powerfully support all three pillars of digital transformation capability, while pushing IT decision rights out to business units helps only two of the three and, when combined with a highly modular architecture, actually weakens the firm’s ability to sense and seize digital opportunities.

The study, conducted by Andreas Olsen, Henri Hussinki, and Patrick Mikalef, draws on modular systems theory and the dynamic capabilities perspective. Modular systems theory holds that systems whose components can be separated and recombined adapt more easily, and the authors apply that logic to both sides of the IT organization: the technical side, captured by IT architecture flexibility, and the organizational side, captured by IT governance decentralization, meaning how IT decision-making authority is distributed between central IT and business units. Dynamic capabilities theory, in turn, describes how firms adapt to turbulent environments through three distinct stages: sensing changes and opportunities, seizing them through strategic action and resource mobilization, and transforming structures, processes, and assets to support new ways of operating.

Prior research had largely examined architecture flexibility and governance decentralization in isolation, and where joint effects were studied, the outcome was typically a single aggregate construct such as agility or performance. The literature also disagreed on the direction of the joint effect: some work implied the two levers complement each other, while other studies suggested substitution once fragmentation effects dominate. The new study addresses that tension by modeling the three capability dimensions separately and explicitly testing an interaction between the two modularity levers.

To measure digital transformation capabilities, the authors developed a new 27-item instrument spanning nine micro-foundations: for digital sensing, digital scouting, digital scenario planning, and digital mindset crafting; for digital seizing, rapid prototyping, strategic agility, and balancing digital portfolios; and for digital transforming, navigating innovation ecosystems, redesigning internal structures, and improving digital maturity. The items were adapted from prior dynamic-capabilities frameworks, validated with 12 experts, and confirmed in a small-scale study with 37 respondents before the main data collection. IT architecture flexibility was measured with nine items covering four dimensions: loose coupling, standardization, digital reach, and scalability. IT governance decentralization was measured with three items. All constructs used seven-point Likert scales.

Data came from U.S.-based participants recruited through the Prolific platform, screened for full-time employment and visibility into IT-related decisions, with later rounds also screened on firm size and functional role. After excluding 12 invalid responses for insufficient response variation, 188 valid questionnaires remained. Respondents came from industries including ICT and technology, services, health and social care, financial and insurance, and wholesale and retail trade, and most companies were middle-sized, with 250 to 10,000 employees. The analysis used partial least squares structural equation modeling, chosen because the study combines a formative higher-order construct, newly developed measurement instruments, interaction effects, and an emphasis on predictive assessment.

The headline result is the strength and consistency of the architecture effect. IT architecture flexibility significantly and positively affected digital sensing, digital seizing, and digital transforming, with standardized path coefficients of 0.622, 0.715, and 0.663 respectively, all statistically significant at p < 0.001. The corresponding effect sizes were all large by conventional criteria. In practical terms, the authors note, a one-standard-deviation increase in IT architecture flexibility is associated with roughly two-thirds to one standard deviation in each capability dimension, holding governance and controls constant. The model explained 52.8 percent of the variance in digital sensing, 62.9 percent in digital seizing, and 57.9 percent in digital transforming.

IT governance decentralization told a more conditional story. It positively affected digital sensing and digital seizing, with smaller coefficients of 0.132 and 0.183, but had no significant effect on digital transforming. The most consequential finding, however, was the interaction: decentralization negatively moderated the relationship between architecture flexibility and both sensing and seizing. At one standard deviation above the mean of governance decentralization, the effective slope of architecture flexibility on sensing fell from 0.622 to roughly 0.48, and on seizing from 0.715 to roughly 0.60. The moderation on transforming was not significant. In short, technical modularity does the heavy lifting in capability formation, while governance modularity is a secondary, capability-specific lever whose value depends on how it is combined with the architecture.

The authors identify three mechanisms through which the joint use of the two modularity levers can backfire. First, coordination costs rise with the number of autonomous decision loci: when governance is decentralized, each business unit must independently negotiate priorities, interface contracts, data semantics, and roadmap timing with its peers, generating overhead that competes for the managerial attention that sensing and seizing require. Second, architectural drift and misalignment emerge, because the same ease of plugging in new components that makes a flexible architecture attractive lets units pick locally optimal but globally inconsistent platforms, data models, and integration patterns, eroding the shared architectural language that enables enterprise-wide sensing and rapid mobilization. Third, effort is duplicated and capabilities fragment, as decentralized units launch parallel digital initiatives that solve similar problems with different tooling, wasting investment and fragmenting the firm-level routines that should otherwise accumulate from local experimentation.

For managers, the study suggests that building digital transformation capabilities requires not only investing in modular, interoperable IT infrastructures but also aligning governance structures to guide how that flexibility is used. The authors recommend differentiated decision rights, centralizing decisions over IT principles, enterprise architecture, and infrastructure while devolving decisions over business application needs and investment prioritization; establishing a thin layer of non-negotiable interface standards with lightweight architectural review; instrumenting sensing and seizing with shared firm-wide routines such as a common digital-opportunity backlog and cross-unit horizon scans; and monitoring the architecture-governance balance with simple diagnostic indicators. They also suggest sequencing: invest in shared architectural standards before devolving IT decision rights.

The authors are careful about the study’s limits. The sample is relatively small and confined to the United States, respondents were single key informants whose answers reflect their perceptions rather than independently verified organizational arrangements, and 19 respondents, about 10 percent of the sample, held neither an IT nor a business-executive title, though re-estimating the model without them preserved every result. The cross-sectional design constrains causal inference, and reverse causation, omitted-variable bias, and self-selection remain residual concerns, though Gaussian-copula tests found no significant endogeneity and common-method bias diagnostics fell below recommended thresholds. The new measurement instrument is regarded as preliminary and in need of further validation. Even so, the central message stands: modularity has a coordination cost, and the question for firms is not how much modularity to adopt, but how architectural and governance modularity should be combined.

Subject of Research: How IT architecture flexibility and IT governance decentralization jointly shape digital transformation capabilities in U.S. firms

Article Title: How IT Architecture Flexibility and Governance Decentralization Shape Digital Transformation Capabilities

Article References: Olsen, A., Hussinki, H., & Mikalef, P. (2026). How IT Architecture Flexibility and Governance Decentralization Shape Digital Transformation Capabilities. Information Systems Frontiers. https://doi.org/10.1007/s10796-026-10832-1

Image Credits: AI Generated

DOI: 10.1007/s10796-026-10832-1

Keywords: IT architecture flexibility, IT governance decentralization, digital transformation capabilities, dynamic capabilities, modular systems theory, PLS-SEM, digital sensing, digital seizing, digital transforming, survey research, organizational design, Information Systems Frontiers