Purpose This paper examines why organizations can generate highly differentiated diagnostic evidence yet make undifferentiated resource commitments. It asks how strategically meaningful distinctions are lost as evidence moves from diagnosis and recommendation through allocation, commitment and delivery. Design/methodology/approach Theory building. The diagnosis-delivery gap is developed by integrating research on resource allocation, strategy implementation, evidence-based management, organization design and organizational justice. A descriptive audit of 200 administrative records from a capability development program illustrates the data conditions required to trace conversion. The extract is not used to test the propositions. Findings Four mechanisms compress resolution: delivery indivisibility, allocation legitimacy, split authority and horizon mismatch. The paper distinguishes diagnostic, recommendation, allocative and delivery resolution and separates variety, alignment and fidelity as components of each. In the administrative extract, ten diagnostic dimensions are complete for all 200 participants while allocation, commitment and delivery fields are unpopulated, so the record cannot establish whether diagnosis altered provision, or whether individualization was genuine or symbolic. Practical implications Senior leaders should assess delivery modularity, decision rights, allocation rules, override visibility and conversion traceability before investing in more sophisticated diagnostic or recommendation systems. Originality/value The paper introduces resolution as a missing dimension in evidence-led execution, showing that the strategic value of analytics depends on the organization’s capacity to preserve meaningful distinctions through allocation and delivery.
Purpose From a cybernetic perspective, leadership can be approached as informed practices of status monitoring and intervention. This paper aims to use this approach to describe leadership capital and expenditures in complex-adaptive systems, addressing the efforts to keep an organization in the state of boundary stability between order and chaos. Design/methodology/approach Drawing on recent works on complexity leadership, the authors capture two different monitoring and intervention activities. One is directed at the creation of order, the other at its disturbance. A combination of both is needed to avoid both stagnation and complete disintegration. In consequence, it also becomes necessary to spend effort on handling contradictions resulting from this combination. To illustrate the validity and relevance of their approach, the authors discuss its application to (1) identify leadership capital and expenditures (2) understand implications of different leadership styles. Findings The model distinguishes three types of issues that leadership must address to enable successful management, which can be further broken down to monitoring, intervention and processing contradiction. This leads to a radically different account of leadership capital and expenditures than elsewhere in organizational studies. Requisite variety in the organization emerges as a critical factor regarding the frequency of interventions and confrontation with contradictions and tensions. Originality/value The model shows how leadership in complex-adaptive systems can be formalized to identify intangible assets for leadership in organizations, identify necessary expenditures and discuss strengths and weaknesses of different leadership practices. It departs from earlier work in cybernetics focusing exclusively on floating equilibria, shifting the focus on boundary stability between order and chaos as the goal of monitoring and intervention.
Purpose Firms increasingly depend on a small number of external providers for artificial intelligence (AI) models, yet that dependence is not only imposed by supplier concentration; it can also be reproduced by a firm’s own technology-integration decisions. This paper aims to examine how provider-specific integration progressively erodes a firm’s ability to switch providers, and what company leadership can do about it. Design/methodology/approach This paper is conceptual. Using resource dependence theory as its base, it draws on transaction cost economics, modularity and real options reasoning, organisational routines theory, and the attention-based view to develop a mechanism and an organisational capability, formalised in four core propositions, a typology matrix and a remedy audit. Findings The paper develops integration drift, a reinforcing process in which locally rational integration decisions cumulatively raise the cost of switching providers. It also introduces architectural hedging capacity (AHC), an organisational capability enacted through three recurring routines: governance-enforced modularity, rehearsed recalibration and sustained managerial attention. And lastly, the argument also identifies a boundary condition in resource dependence theory: where dependence is continuously reproduced through organisational integration, maintaining autonomy requires continuously reproduced exit capacity rather than a one-time remedy. Originality/value The paper extends resource dependence theory to settings where dependence is reproduced inside the firm through its own architecture, not only imposed externally. A typology and a remedy audit checklist give boards and managers a diagnostic for assessing and managing AI-provider dependence.
Purpose This paper introduces the AI-ification paradox and explains why organizations that invest in increasingly autonomous AI often keep it in advisory rather than delegated roles. We examine the paradox in healthcare services, where the promise of AI is high but autonomy is tightly bound. Design/methodology/approach The article develops a conceptual model by integrating servitization and task-delegation literature with an illustrative synthesis of 11 interviews from healthcare settings, supplemented by two subsequent expert interviews used for model refinement. Findings AI is most readily delegated to reversible, auditable tasks such as documentation, information retrieval, drafting and second-opinion support. Delegation stalls when verification remains costly, tasks require socio-emotional alignment or professional judgment, and organizational risk, liability, and workflow integration are unresolved. The result is a gap between AI capability and delegated authority that can trap value and weaken return on AI investment. Resources and costs also shape the scope and value of delegation. Originality/value The paper extends research on digital servitization by identifying delegation, rather than technical capability alone, as the next bottleneck in AI-enabled service delivery. It offers a managerial framework for governed autonomy in services.
Purpose This paper introduces the AI Throwaway Economy, a framework explaining when a still-usable knowledge-based output becomes economically disposable because generating a replacement is expected to cost less than refining what already exists. It also considers what may be lost when replacement becomes the default. Design/methodology/approach The article develops a conceptual framework through an integrative synthesis of research on generative AI, automation and augmentation, knowledge management, and the knowledge-based view of the firm. Findings Generative AI can reverse the economics of preservation when the total cost of regeneration falls below the total cost of refinement at comparable quality and risk levels. Both options may use AI. Regeneration is most attractive for codified, standardized, context-light, and readily verifiable outputs. Preservation remains advantageous when outputs embody tacit knowledge, proprietary context, institutional memory, or high integration and error costs. Repeated replacement may also weaken organizational learning. Practical implications Leaders should compare full lifecycle costs, classify outputs by preservation value, retain provenance and decision context, and invest in verification. They should also ask: What will our people stop learning or practicing if we routinely replace rather than refine their work? Originality/value This framework adds a comparative-cost explanation of why organizations may replace a usable knowledge-based output even when neither the underlying knowledge nor the task has become obsolete.
