AI infrastructures are redrawing firm boundaries: coordination and adaptive capacity increasingly flow through externally governed algorithms, decoupling legal ownership from effective control and creating new interorganizational power asymmetries.

Main Finding

The paper argues that the diffusion of externally governed AI infrastructures within platform ecosystems breaks the traditional alignment of ownership, coordination authority, and execution capability that underpins firm boundaries. It reconceptualizes firm boundaries as "algorithmic integration boundaries" determined by control over coordination and learning processes rather than by legal ownership or asset possession. Four cumulative mechanisms generate "algorithmic boundary permeability"—legal ownership can remain while effective coordination authority and adaptive capacity extend into external infrastructures.

Key Points

• Traditional firm-boundary theories assume alignment of ownership, decision authority, and execution capability; AI infrastructures decouple these elements.
• Firms embed algorithmic systems (pricing, forecasting, visibility management, workflow coordination) that are externally governed, enabling coordination without asset transfer or hierarchical integration.
• The paper identifies four mechanisms that reconfigure boundaries:
• Data dependency intensification — firms become increasingly reliant on external data flows and datasets controlled by platforms/infrastructures.
• Workflow embedding — algorithmic services are integrated into operational workflows, making coordination dependent on external APIs and platforms.
• Algorithmic visibility and control redistribution — monitoring, signals, and decision defaults are located outside the firm, shifting who sees and shapes behavior.
• Capability redistribution — learning and adaptive capabilities migrate to infrastructure providers, creating infrastructure-dependent capabilities for firms.
• These mechanisms cumulatively create algorithmic boundary permeability: firms retain legal ownership while effective coordination power and adaptive learning reside, in part or in whole, with external infrastructure.
• Theoretically, this refines boundary theory and extends resource-based and dynamic capability perspectives by introducing "infrastructure-dependent capabilities."
• The paper frames algorithmic mediation as a structural source of interorganizational power asymmetry (platforms/infrastructure providers gain coordination and learning power over dependent firms).

Data & Methods

• Conceptual/theoretical contribution: develops a mechanism-based framework rather than reporting new empirical data.
• Methodological approach appears to consist of:
  • Integrative literature synthesis across theory of the firm, platform ecosystems, resource-based view, and dynamic capabilities.
  • Mechanism identification and theorizing about causal processes (the four mechanisms) that link AI infrastructures to boundary reconfiguration.
  • Use of illustrative domains (e.g., pricing, forecasting, visibility management, workflow coordination) to ground the conceptual framework.
• No primary empirical testing is reported; the piece sets out propositions and a theoretical lens for future empirical work.

Implications for AI Economics

• Firm boundary modeling: Economic models should disentangle legal ownership from coordination authority and adaptive capacity; ownership is no longer a sufficient indicator of control over productive activities when external AI infrastructures mediate coordination.
• Market power and value capture: Platforms and infrastructure providers can accumulate coordination and learning rents even without formal ownership of complementary firms' assets, altering sources of market power and monopoly/oligopoly dynamics.
• Measurement of capabilities: Empirical work and productivity accounting must recognize "infrastructure-dependent capabilities"—firm performance may hinge on access to/embeddedness in external algorithmic infrastructures rather than solely on owned assets or internal skills.
• Contracting and governance: Transaction-cost and property-rights analyses should incorporate contracts, SLAs, API policies, and governance features of external infrastructures as mechanisms that shape incentives, hold-up risk, and ex-post bargaining power.
• Dynamic capabilities and strategy: Firms’ strategic choices include whether to embed, replicate, or internalize algorithmic functions; investments in data access, interoperability, and portability become strategic levers.
• Policy and regulation: Antitrust, data portability, interoperability, and platform governance debates need to account for coordination/control asymmetries that arise through algorithmic mediation; remedies focused only on ownership or asset transfer may miss how power is exercised.
• Labor and organizational design: Work design and managerial authority change when coordination logic is encoded in external algorithms; complementarities between human skills and infrastructure-mediated processes require new approaches to incentives and training.
• Research directions for AI economics:
  • Empirically measure algorithmic boundary permeability and its effects on firm performance, entry, and industry structure.
  • Quantify rents accruing to infrastructure providers via coordination and learning advantages.
  • Model contracting solutions and regulatory interventions that re-balance authority and data-access asymmetries.
  • Study dynamic investments in portability, API access, and competing infrastructures as strategic responses.

Overall, the paper urges economists and strategists to treat algorithmic infrastructure as a core element shaping firm boundaries, power relations, and the allocation of coordination rents in AI-enabled ecosystems.