Analytical Insight · Strategic Management

From Resource Rigidity to Adaptive Capacity:
Dynamic Capabilities in the Age of AI

An analysis of how the generative AI paradigm shift constitutes a discontinuous capability shock for incumbent firms — rendering resource-based advantages insufficient — and how the dynamic capabilities framework illuminates the sensing–seizing–reconfiguring pathways through which incumbents sustain competitive advantage under conditions of technological discontinuity.

Abstract

The diffusion of generative artificial intelligence constitutes a discontinuous technological shock that systematically erodes the rent-generating capacity of incumbent firms' resource stocks — particularly those intangible assets, routinized knowledge processes, and organizational competencies whose value was premised on scarcity conditions that AI capability directly undermines. Drawing on Barney's (1991) resource-based view and Teece, Pisano, and Shuen's (1997) dynamic capabilities framework, this paper argues that the AI transition does not merely alter the relative value of specific resources but reorganizes the structural conditions under which resource-based advantages are generated and sustained. Firms whose competitive position rested on proprietary knowledge accumulation, human capital specificity, and process optimization face the simultaneous obsolescence of resource stocks and the capability rigidities that generated them. The central theoretical proposition is that durable competitive advantage in an AI-intensive environment requires not the protection of existing resource configurations but the development of higher-order dynamic capabilities — sensing, seizing, and reconfiguring — that enable continuous realignment of resource portfolios with rapidly evolving technological frontiers. Empirical evidence on AI adoption rates, productivity bifurcation between early and late adopters, and organizational restructuring patterns supports the theoretical claim that the capability trap identified in the dynamic capabilities literature is the primary mechanism generating firm-level performance divergence under technological discontinuity.

I · The AI Paradigm Shift as Discontinuous Capability Shock

Technological Discontinuity and the Erosion of Incumbent Advantage

The emergence of large language models and generative AI systems from 2022 onward represents a qualitative departure from the prior trajectory of automation technology. Earlier waves of digital transformation — enterprise resource planning, cloud computing, robotic process automation — primarily altered the efficiency with which firms executed existing knowledge-based processes. Generative AI, by contrast, encroaches directly on the cognitive tasks — writing, analysis, synthesis, code generation, pattern recognition across unstructured data — that constituted the core of organizational knowledge work and whose tacitness had historically insulated them from substitution pressure. This distinction is theoretically consequential. The tacit knowledge stocks that Barney (1991) identified as the foundation of sustained competitive advantage were valuable precisely because their non-codifiability rendered them imperfectly imitable. AI systems capable of approximating tacit knowledge outputs — legal reasoning, strategic analysis, customer relationship management, research synthesis — erode this imitation barrier in a structurally novel way, not through competitor learning or organizational reverse-engineering, but through the commoditization of cognitive output itself.

The pace of diffusion compounds the theoretical challenge. Historical technological transitions — the adoption of electricity in manufacturing, the spread of enterprise computing — unfolded over decades, providing incumbent firms with adjustment horizons sufficient to enable incremental capability realignment. The generative AI transition, by contrast, compressed the interval between technological emergence and broad organizational applicability to approximately twenty-four months. GPT-4's release in March 2023 was followed within eighteen months by organizational AI deployment frameworks, sector-specific fine-tuned models, and AI-integrated workflow tools across knowledge-intensive industries. This compression eliminates the adjustment buffer that prior theoretical treatments of technological disruption implicitly assumed, and renders the dynamic capabilities framework's emphasis on temporal sensitivity — the notion that capability development trajectories matter as much as capability endowments — analytically urgent in a way that prior technological transitions did not demand.

A second distinctive feature of the AI transition concerns its pervasiveness across industry boundaries. Prior general-purpose technologies — the steam engine, electricity, the internet — exhibited pervasive effects but were mediated by substantial sector-specific adoption barriers in physical capital, infrastructure, and regulatory context. Generative AI's primary delivery mechanism, software-as-a-service and API-based integration, eliminates many of these mediation barriers. The implication is that the AI capability shock is not confined to a technology sector facing incumbent disruption; it simultaneously restructures the competitive environments of professional services, financial intermediation, healthcare administration, education, logistics, and manufacturing knowledge work. For strategic management theory, this pervasiveness implies that no sector provides a stable institutionalized refuge from the capability obsolescence pressure that the AI transition generates — a condition that renders dynamic capability development not a differentiating strategic choice but a baseline survival requirement for incumbent firms across industries.

II · Resource-Based View Under Technological Disruption

When VRIN Conditions Become Liabilities

72%
Organizations reporting AI adoption (2024)
▲ from 55% in 2023 (McKinsey)
40%
Productivity gain in AI-augmented knowledge work
▲ task-level studies (MIT, Stanford)
$13T
Projected AI contribution to global GDP by 2030
McKinsey Global Institute estimate

Barney's (1991) resource-based view established that sustained competitive advantage derives from firm resources that are simultaneously Valuable, Rare, Inimitable, and Non-substitutable — the VRIN framework. The framework's predictive power rests on the assumption that the environmental conditions determining which resource attributes satisfy these criteria are relatively stable. Technological discontinuity of the magnitude represented by generative AI does not merely alter the value of specific resources; it reorganizes the structural conditions under which the VRIN criteria are satisfied. Resources that satisfied all four conditions under the prior technological regime may simultaneously lose value, become replicable through AI-mediated processes, and face substitution — rendering the entire resource portfolio strategically inadequate within a compressed timeframe.

Consider the category of accumulated knowledge assets — proprietary databases, codified analytical frameworks, domain-specific expertise embedded in human capital — that constituted the primary competitive advantage of knowledge-intensive incumbents across professional services, consulting, and research-intensive industries. Under the pre-AI regime, these assets satisfied the VRIN conditions through a combination of historical accumulation barriers (value), organizational embeddedness (rarity), tacit knowledge components (inimitability), and absence of functional substitutes (non-substitutability). The arrival of large language models, trained on corpora that encompass substantial portions of publicly available domain knowledge and capable of generating domain-specific analytical output at low marginal cost, directly attacks all four conditions simultaneously: it reduces value (output is substitutable at lower cost), eliminates rarity (similar capabilities are accessible to all firms through API subscription), undermines inimitability (the tacit knowledge component that generated imitation barriers is partially replicated by model training), and introduces a functional substitute of broad applicability. The theoretical implication is that resource stocks that satisfied the VRIN conditions under the prior technological regime can transition to value-destroying liabilities under the AI regime — not because they generate negative cash flows, but because the organizational commitments, governance structures, and investment patterns built around their protection prevent adaptation toward the capability configurations that the new environment rewards.

This dynamic is precisely what Levinthal and March (1993) identified as the capability trap: the systematic tendency of successful organizations to over-exploit existing capabilities relative to the exploration of new ones. In the AI transition context, the capability trap operates through multiple reinforcing mechanisms. Firms with deep investments in human capital configured around pre-AI knowledge work face dual pressures — the sunk costs embedded in existing talent structures and the internal political resistance of constituencies whose organizational status depends on the continued value of those structures. Firms with established process architectures built around human-mediated knowledge flows face coordination costs in restructuring those architectures around AI-mediated alternatives. And firms with strategic identities built around expertise-based differentiation face organizational identity challenges in repositioning around AI-augmented capability claims. These mechanisms are mutually reinforcing: each increases the perceived switching cost of capability reconfiguration, extending the adjustment horizon precisely when the technological diffusion rate demands compression of it.

AI Adoption Rate by Industry Sector (2022–2024)
Share of organizations reporting AI deployment in core workflows (%)

※ Illustrative trend based on McKinsey Global Survey on AI (2022–2024) and Stanford AI Index Report (2024).

Theoretical Note

The RBV's explanatory leverage is strongest under conditions of environmental stability. Penrose (1959) noted that firm growth is constrained by the productive services that can be extracted from existing resource bundles. Generative AI constitutes a Penrosean shock in reverse: it expands the productive services extractable from knowledge assets while simultaneously reducing the marginal value of the human capital that previously mediated their extraction. This asymmetric effect — capability expansion alongside human capital depreciation — generates a distributional challenge within incumbent organizations that neither the RBV nor classical dynamic capabilities theory has fully addressed.

III · Dynamic Capabilities as the Adaptive Mechanism

Sensing, Seizing, and Reconfiguring in an AI-Disrupted Environment

Teece, Pisano, and Shuen's (1997) dynamic capabilities framework was developed to explain how firms sustain competitive advantage in rapidly changing technological environments — a context that precisely describes the AI transition. The framework's central insight is that sustainable competitive advantage in turbulent environments derives not from the possession of superior resource stocks but from the higher-order capacity to sense environmental changes before they fully manifest in market signals, seize new resource configurations before competitors establish position, and reconfigure organizational architectures at the pace the environment requires. Each of these three capability classes maps onto a specific adaptive challenge that the AI transition places on incumbent firms, and together they define the organizational capability profile that distinguishes firms capable of sustained competitive repositioning from those that will undergo capability-driven performance decline.

A critical distinction structures the analysis that follows. Operational capabilities — the organizational routines that execute existing activities efficiently — and dynamic capabilities differ not in degree but in kind. An incumbent firm may achieve world-class operational performance in AI-augmented workflow execution while simultaneously lacking the dynamic capabilities required to reconfigure its fundamental business model, knowledge architecture, or competitive positioning in response to AI-driven industry restructuring. The firms that achieve sustained competitive advantage through the AI transition are those that invest not merely in AI tools deployment (an operational capability enhancement) but in the organizational capacity to continuously realign their entire capability portfolio with the evolving technological frontier — a dynamic capability investment of fundamentally different character.

Sensing
Detecting AI-Driven Capability Shifts
Sensing in the AI context requires organizational processes capable of interpreting capability-level signals — not merely monitoring AI tool releases, but assessing their implications for the firm's existing competitive position, value chain configuration, and human capital architecture. Effective sensing requires cross-functional intelligence integration that connects technology assessment, competitive analysis, and strategic planning in continuous rather than periodic cycles. Incumbents that treated early AI developments as IT procurement decisions rather than strategic sensing inputs systematically failed to classify the transition as a structural capability shock, compressing their adjustment horizon and constraining subsequent adaptive options. The sensing challenge is heightened by AI's cross-sector pervasiveness: signals from outside a firm's primary competitive domain — AI-driven restructuring in adjacent industries — carry early-warning information about trajectory that sector-specific sensing routines fail to capture.
Seizing
Committing Resources to AI-Reconfigured Value Creation
Seizing denotes the organizational capacity to commit resources to new value-creating configurations once the sensing function has identified the opportunity or threat. In the AI transition context, seizing extends beyond AI tool procurement to encompass strategic decisions about which elements of the existing value chain to augment, which to automate, and which to fundamentally redesign around AI-native architectures. Teece (2007) emphasizes that seizing quality is constrained not by financial resources but by organizational decision-architecture: the capacity to overcome inertia in human capital governance, accept short-run efficiency costs from parallel capability development, and devolve commitment authority to organizational units closest to the AI integration frontier. Firms with highly centralized human capital governance and rigid workflow architectures exhibit systematically slower and lower-quality seizing responses, independent of their financial AI investment levels.
Reconfiguring
Transforming Organizational Identity for AI-Native Competition
Reconfiguring represents the deepest and most organizationally demanding dynamic capability class: the transformation of organizational structures, knowledge bases, incentive systems, and strategic identity to fit a qualitatively altered competitive environment. In the AI transition, reconfiguration exceeds workflow automation. It requires redesigning the organizational knowledge architecture around human–AI complementarity rather than human cognitive primacy; rebuilding performance evaluation and incentive systems to reward AI-augmented output rather than individual expertise demonstration; restructuring alliance and partnership networks to incorporate AI capability providers as strategic partners rather than vendors; and, most profoundly, reconstituting the firm's strategic identity around a capability claim that is credible and differentiated under AI-intensive competition. This last dimension — strategic identity reconfiguration — is the most undertheorized aspect of the dynamic capabilities literature as applied to technological discontinuity, and the most difficult to execute under conditions of organizational inertia and constituency resistance.

A central theoretical proposition emerging from this analysis is that the AI transition generates a capability sequencing problem that prior applications of the dynamic capabilities framework have not adequately addressed. In the supply chain fragmentation context analyzed in Brief No. 1, the sensing–seizing–reconfiguring sequence operated at broadly compatible temporal horizons: geopolitical signals, while difficult to interpret, unfolded over years, providing adjustment time consistent with organizational reconfiguration capacity. The AI transition compresses this sequence: the interval between the emergence of a materially disruptive AI capability and its broad adoption by competitive actors has contracted to quarters rather than years. This compression means that firms lacking pre-existing sensing and seizing routines — organizations for which dynamic capability development was not a prior strategic investment — face the AI transition with capability deficits that cannot be remedied within the available adjustment window. Path-dependence thus generates a structurally disadvantaged cohort of incumbents: those whose prior competitive success in stable environments provided no organizational impetus to develop the dynamic capabilities that the current environment demands.

The theoretical complement to the capability trap is what might be termed the complementary asset recombination opportunity — a concept that Teece (1986) anticipated in his analysis of how incumbent firms can leverage co-specialized assets to appropriate returns from technological innovation even when they do not originate it. Incumbent firms possess organizational assets that AI-native entrants lack: established customer relationships, regulatory approvals and compliance infrastructure, domain-specific operational knowledge, and institutional trust accumulated through sustained market presence. These complementary assets retain substantial value under AI-intensive competition and, critically, can be recombined with AI capabilities to generate competitive positions unavailable to new entrants. The dynamic capability required to execute this recombination — sensing the AI capabilities that most powerfully augment existing complementary assets, seizing the organizational configurations that activate the recombination potential, and reconfiguring governance and knowledge structures to sustain the combination's competitive advantage — represents the theoretically precise account of how incumbents achieve sustained advantage through the AI transition rather than despite it.

Firm Performance Divergence: AI Leaders vs. Laggards (Illustrative)
Indexed productivity growth (2020=100), based on McKinsey & Stanford AI Index estimates

※ Illustrative representation. Derived from McKinsey Global Institute (2023), Stanford AI Index (2024), and IMF Working Paper on AI and productivity.

IV · Strategic Implications for Incumbent Firms

Capability Investment Priorities for Sustained Competitive Repositioning

For incumbent firms in knowledge-intensive industries, the primary strategic imperative is what the dynamic capabilities framework would identify as complementary asset recombination: the deliberate organizational investment in identifying which elements of the existing resource portfolio retain VRIN conditions under AI-intensive competition, and developing the sensing and seizing capacity required to combine those assets with AI capabilities before competitive actors establish dominant positions. The analytical error to be avoided is the conflation of AI tool deployment — an operational capability enhancement — with dynamic capability development. Firms that achieve high rates of AI tool adoption without investing in the organizational capacity to continuously realign their resource portfolio with evolving AI frontiers will generate one-time productivity improvements that competitors rapidly replicate, without achieving the dynamic capability endowment required for sustained competitive repositioning. The distinction between operational AI adoption and dynamic capability development is the central strategic diagnostic challenge of the current transition period.

A second strategic implication concerns human capital architecture. The complementary asset that most clearly retains value under AI-intensive competition — and that AI-native entrants cannot rapidly replicate — is accumulated domain-specific judgment: the capacity to evaluate AI-generated outputs, identify their failure modes in specific institutional and organizational contexts, and integrate them with tacit relational knowledge that AI systems cannot generate. Incumbent firms that invest in reconfiguring their human capital architecture toward this complementary function — developing what might be termed AI-augmented judgment capacity rather than AI-substituted labor — position themselves to extract sustained competitive advantage from the human–AI complementarity that the AI transition enables. This reconfiguration requires not merely reskilling investments but the redesign of organizational roles, performance evaluation systems, and knowledge management structures to reward and develop AI-augmented judgment as the primary human capital asset.

For firms in emerging markets and developing economies, the AI transition creates a structural opportunity analogous to — and potentially larger than — the GVC integration opportunity of the prior two decades. AI capabilities accessible through API-based interfaces can compress the knowledge and analytical capacity gaps that historically constrained knowledge work quality in resource-constrained organizational environments. However, the conversion of AI access into sustained competitive advantage requires the same dynamic capabilities — sensing, seizing, reconfiguring — that the framework identifies as the determinants of sustained advantage under technological discontinuity more broadly. The risk of AI-enabled capability catch-up being reversed by more rapid AI frontier advancement among leading-economy incumbents is real and requires a strategic response centered on the development of organizational dynamic capabilities rather than mere technology adoption. The theoretical implication is that dynamic capability development is not a luxury investment for resource-constrained organizations in developing economy contexts; it is the primary mechanism through which AI access translates into sustained organizational improvement rather than temporary efficiency enhancement.

For academic and research institutions, the AI transition generates a distinctive reconfiguration challenge. Universities and research organizations built competitive positions on the scarcity of expert knowledge access — a scarcity that AI systems are systematically eroding. The dynamic capability response required is not resistance to AI adoption but a fundamental reconfiguration of institutional value claims: from knowledge access provision toward the cultivation of AI-augmented analytical judgment, critical evaluation capacity, and the institutional trust that sustains the credibility of knowledge claims in AI-saturated information environments. The RBV assets that research institutions retain — methodological rigor, reputational credibility, peer validation infrastructure, and the relational networks that sustain research collaboration — represent precisely the complementary assets that AI-native information providers cannot replicate. Institutions capable of sensing this strategic landscape, seizing AI capabilities that amplify rather than substitute for these complementary assets, and reconfiguring their organizational identity around the resulting capability combination are positioned for sustained relevance in the AI-intensive academic environment.

Graduate-Level Discussion Questions
Q1
Barney (1991) argues that resources satisfying the VRIN conditions generate sustained competitive advantage by definition. Generative AI appears to erode the inimitability condition for knowledge-based resources by approximating tacit knowledge outputs through model training. Develop a theoretically grounded account of whether this constitutes a fundamental challenge to the RBV's predictive framework — requiring revision of the VRIN criteria — or whether it is better interpreted as a shift in which specific resources satisfy the existing criteria, leaving the framework's theoretical structure intact. What empirical evidence would allow a researcher to distinguish between these two theoretical positions?
Q2
Teece (2007) identifies reconfiguring as the dynamic capability class most resistant to imitation because it requires organizational identity transformation rather than merely resource redeployment. In the AI transition context, the dominant observable firm response has been AI tool adoption — a seizing-level response — rather than the organizational identity reconfiguration that the framework identifies as the basis of sustained advantage. Using the sensing–seizing–reconfiguring framework, construct a set of empirically observable indicators that would allow a researcher to distinguish firms that are executing genuine organizational reconfiguration from those that are engaging in seizing-level AI adoption while framing it as transformation. What organizational variables would you prioritize in a firm-level longitudinal study?
Q3
The complementary asset recombination opportunity — the proposition that incumbents can leverage co-specialized assets unavailable to AI-native entrants — depends on the assumption that incumbents can develop dynamic capabilities rapidly enough to execute the recombination before entrants develop substitute complementary assets. Evaluate the theoretical and empirical conditions under which this window of opportunity closes for incumbent firms: specifically, identify the organizational, industry-structural, and AI capability development variables that determine the duration of the incumbent recombination window, and derive strategic prescriptions for incumbents facing different combinations of these variables.
Data Sources & References

Barney, J. (1991). Firm resources and sustained competitive advantage. Journal of Management, 17(1), 99–120. · Teece, D.J., Pisano, G., & Shuen, A. (1997). Dynamic capabilities and strategic management. Strategic Management Journal, 18(7), 509–533. · Teece, D.J. (2007). Explicating dynamic capabilities: The nature and microfoundations of (sustainable) enterprise performance. Strategic Management Journal, 28(13), 1319–1350. · Teece, D.J. (1986). Profiting from technological innovation: Implications for integration, collaboration, licensing and public policy. Research Policy, 15(6), 285–305. · Levinthal, D.A., & March, J.G. (1993). The myopia of learning. Strategic Management Journal, 14(S2), 95–112. · Cohen, W.M., & Levinthal, D.A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35(1), 128–152. · Penrose, E.T. (1959). The Theory of the Growth of the Firm. Oxford: Basil Blackwell. · McKinsey Global Institute (2023). The Economic Potential of Generative AI: The Next Productivity Frontier. McKinsey & Company. · McKinsey & Company (2024). The State of AI in Early 2024: Gen AI Adoption Spikes and Starts to Generate Value. · Stanford University (2024). AI Index Report 2024. Stanford Institute for Human-Centered Artificial Intelligence (HAI). · IMF (2024). Gen-AI: Artificial Intelligence and the Future of Work. IMF Staff Discussion Note SDN/2024/001. · Brynjolfsson, E., Li, D., & Raymond, L.R. (2023). Generative AI at Work. NBER Working Paper No. 31161.