Traditional models of intellectual capital miss a fourth locus created by self-learning systems: 'Algorithmic Capital' — a unified asset composed of data, models, agents and learning infrastructure that generates and renews knowledge with relative autonomy. The paper argues this warrants reconstructing the three-part taxonomy into a four-dimensional framework, though the claim remains theoretical and untested.

Main Finding

Abuhaimed (2026). Algorithmic Capital: Toward Reconstructing Intellectual Capital Theory in the Age of Artificial Intelligence — ARADO Business Journal, 4(1):103–126. DOI: 10.64190/abj.2026.69

The classical tripartite intellectual-capital taxonomy (human, structural, relational) implicitly rests on a single classificatory criterion: the locus of knowledge embeddedness (where knowledge resides). The emergence of self‑learning AI systems constitutes a new, distinct locus of organizational knowledge — termed Algorithmic Capital — that warrants elevating intellectual capital to a four‑part model. Algorithmic Capital is composed of four interdependent components (data, models, agents, learning infrastructure), unified by their capacity to generate and renew knowledge with relative autonomy from direct human intervention, and it can satisfy RBV/VRIN conditions as a strategic asset distinct from static structural capital.

Key Points

• The tripartite taxonomy can be coherently read through the criterion “locus of knowledge embeddedness”: human (individuals), structural (codified organizational systems), relational (external networks).
• Learning AI systems instantiate a new locus: knowledge resides and evolves within algorithmic systems, not reducible to humans, static structures, or external ties.
• Definition of Algorithmic Capital: the organizational asset-locus whose components enable automated knowledge generation and continual learning with relative independence from direct human action.
• Four proposed components of Algorithmic Capital:
  • Data (input resources/pipelines)
  • Models (learned representations, algorithms)
  • Agents (deployed decision-making/operational actors)
  • Learning infrastructure (pipelines, experimentation, compute, MLOps)
• Distinction from structural capital: structural capital is codified but largely static; algorithmic capital is dynamic and self‑modifying through learning.
• Data alone is not a separate locus — raw data are inputs; their epistemic character arises within the algorithmic locus.
• Algorithmic capital is positioned as a resource (asset) that composes organizational capabilities (AI capability, digital capability, dynamic capabilities) rather than being a capability itself.
• The construct is theoretically evaluated against RBV/VRIN (Value, Rarity, Inimitability, Non‑substitutability) and argued to plausibly satisfy strategic-asset conditions.
• Practical exemplar: Saudi institutional architecture (SDAIA) is cited as an organizational manifestation where data, models, agents, and infrastructure are managed as an integrated, distinct ecosystem.
• Research approach: conceptual/theoretical reconstruction rather than empirical measurement; the paper offers testable propositions for future empirical work.

Data & Methods

• Methodology: Integrative Conceptual Review / theory‑reconstruction. The author conducts a critical analysis of foundational intellectual-capital literature (Stewart; Sveiby; Edvinsson & Malone) to extract the latent classificatory criterion and then tests its adequacy conceptually against contemporary AI artifacts.
• Theoretical grounding: synthesis with organizational knowledge literature (Nonaka; Spender; Kogut & Zander), information-systems literature on AI components (Iansiti & Lakhani; Mikalef & Gupta), and the Resource‑Based View (Barney; Wernerfelt) including VRIN criteria and dynamic-capabilities framing (Teece).
• Outputs: (a) articulation of the locus-of-knowledge criterion as the taxonomy’s governing principle; (b) conceptual definition and internal decomposition of Algorithmic Capital into four components; (c) theoretical propositions (three main propositions) proposing the quadripartite reconstruction and inviting empirical testing.
• Limitations: no empirical dataset or quantitative measurement is provided; conclusions are conceptual and require empirical validation and operationalization (measurement models, indicators, econometric tests).

Implications for AI Economics

• Valuation and firm-level accounting
  • Intellectual-capital measurement frameworks and valuation models should be updated to treat algorithmic assets as a separate locus; existing measures (e.g., VAIC) that assume a tripartite structure may undercount or misattribute AI-derived value.
  • Algorithmic Capital can generate sustained rents (VRIN), implying persistent intangible value that affects firm valuation, Tobin’s q, and investment returns in AI-intensive firms.
• Strategy and the Resource-Based View
  • Firms’ strategic advantage increasingly depends on integrated investments across the four components (data, models, agents, learning infrastructure) and on governance that preserves rarity and inimitability (e.g., proprietary data pipelines, model fine‑tuning practices, MLOps sophistication).
  • Dynamic capabilities remain crucial: the ability to reconfigure and scale algorithmic capital (experiment fast, adapt models) mediates the translation of algorithmic assets into competitive advantage.
• Market structure and concentration
  • Algorithmic capital’s high fixed costs, scale economies (data and compute), and path dependence in model improvement suggest tendencies toward concentration and winner‑take‑most dynamics in industries where algorithmic assets are central.
  • Platform and ecosystem strategies (control of data flows, model marketplaces, agent deployment) can amplify market power — connecting organizational-level algorithmic capital to macro-level platform capitalism debates.
• Labor, productivity, and complementarities
  • Algorithmic capital reshapes complementarities between human and machine capital: some tasks are substituted, others augmented. Economic models must capture how algorithmic capital reorganizes labor demand, productivity growth, and wage dispersion.
• Policy and regulation
  • Recognizing algorithmic capital as an organizational locus implies policy relevance: data governance, model accountability, competition policy (anticompetitive lock‑in via proprietary datasets), and investment subsidies/standards for shared learning infrastructure.
• Measurement and empirical agenda
  • Operationalizing Algorithmic Capital requires new indicators: data quality/coverage metrics, model performance/uniqueness measures, agent deployment breadth, and learning‑infrastructure KPIs (update frequency, compute capacity, MLOps maturity).
  • Empirical work could test propositions linking algorithmic capital endowments to firm performance, persistence of returns (VRIN), entry barriers, R&D investment patterns, and labor outcomes.
• Macroeconomic and growth implications
  • Aggregate accumulation of algorithmic capital may affect total factor productivity (TFP) dynamics; the distribution of this capital across firms and sectors will shape productivity divergence and the pace of AI‑driven growth.

Overall, the paper reframes a core taxonomy in intellectual-capital theory to include algorithmic assets as a distinct, strategic locus. For AI economics, this invites rethinking measurement, valuation, firm strategy, competition, labor complementarities, and policy — and sets an empirical research agenda to operationalize and test the proposed quadripartite model.