Geopolitical fragmentation forces firms to reassign AI decision rights from pure efficiency to political legitimacy, producing four distinct governance architectures; the paper proposes GADAT to explain how firms redesign AI–human authority to protect legitimacy and resilience.

Main Finding

The paper develops Geo-Algorithmic Decision Authority Theory (GADAT), a conceptual framework arguing that geopolitical fragmentation (sanctions, data-localization, friend‑shoring, contested digital infrastructure) systematically shifts AI delegation in global supply chains from an efficiency‑centric logic toward a legitimacy‑centric logic. Firms respond by redesigning the locus, granularity, and accountability of decision authority—reclaiming, partitioning, or insulating algorithmic decision rights depending on political exposure—enabled by newly emphasized capabilities (e.g., algorithmic buffering, delegation adjustment). The theory generates a set of constructs, mechanisms, and propositions predicting four organizational decision‑authority configurations and their implications for resilience, performance, and multinational adaptability.

Key Points

• Theoretical synthesis: integrates Resource Dependence Theory, Institutional Theory, Information Processing Theory, and Dynamic Capabilities to explain how external political forces interact with algorithmic delegation.
• Core causal mechanisms:
  • Coercive legitimacy pressures (regulation, sanctions, political narratives) that make algorithmic decisions contestable.
  • Political-triggered information constraints (data restrictions, opaque signals) that exceed algorithmic/information processing capacities.
  • Institutional heterogeneity across jurisdictions that requires differentiated authority designs.
• Constructs introduced/used (paper mentions multiple formulations across abstract and main text):
  • From abstract: Political–Algorithmic Misalignment (PAM); Geo‑Institutional Opacity (GIO); Legitimacy‑Centric Delegation (LCD); Algorithmic Legitimacy Buffering (ALB); Delegation Adjustment Capability (DAC).
  • From main development: Geopolitical Digital Exposure (GDE); Algorithmic Legitimacy Vulnerability (ALV); Decision Authority Locus Shift (DALS); Algorithmic Buffering Capabilities (ABC); Legitimacy‑Centric Delegation (LCD); Geo‑Algorithmic Ambidexterity (GAA).
• Causal chain (simplified):
• Higher GDE → 2. Elevated ALV → 3. DALS (reallocation of authority) → 4. Activation of ABC/ALB → 5. Adoption of LCD → 6. Emergence of GAA (dual architectures).
• Predicted structural outcomes (four): recentralized authority, federated hybrid authority, trust‑partitioned authority, and insulated algorithmic authority.
• Propositions: eight propositions link the above configurations and capabilities to supply‑chain resilience, operational performance tradeoffs, and adaptability of MNCs under geopolitical fragmentation.
• Managerial prescriptions: invest in modular/dual AI architectures, build buffering and delegation‑adjustment capabilities, partition decision rights regionally, and prioritize legitimacy criteria (compliance, transparency, sovereignty) when delegating algorithmic authority.

Data & Methods

• Nature of study: conceptual/theoretical paper—not empirical. The paper synthesizes prior literatures, develops new constructs, lays out causal mechanisms, and proposes propositions/models for future testing.
• Methodological approach: literature integration across multiple theories; construct clarification and causal mapping; development of a typology of organizational outcomes; articulation of testable propositions and a research agenda.
• No primary quantitative or qualitative data, no datasets, and no empirical estimation reported. The contribution is theoretical and prescriptive, designed to guide empirical work.

Implications for AI Economics

• Efficiency vs legitimacy tradeoffs: AI delegation decisions now internalize political/legitimacy constraints in addition to classical cost and performance metrics. This raises the effective cost of algorithmic delegation in politically contested environments (e.g., slower decisions, constrained data pools, added compliance overhead), altering the marginal benefit calculation for automation.
• Transaction costs and governance: geopolitical fragmentation increases transaction and monitoring costs associated with cross‑border data flows and algorithmic coordination. Firms may reorganize governance (recentralize, regionalize, or hybridize) to minimize political transaction risks, which affects vertical/horizontal boundaries and outsourcing decisions.
• Investment composition and barriers: firms face incentives to invest in algorithmic buffering (modularity, dual pipelines, redundancy) and delegation‑adjustment capabilities—these are fixed/capacity investments that raise entry barriers and may lead to concentration among firms that can afford geopolitical‑resilient AI infrastructures.
• Market segmentation and fragmentation: differential authority architectures and localized AI stacks can produce segmented markets (jurisdiction‑specific algorithms, datasets, and supply‑chain practices), reducing economies of scale for global AI models and potentially slowing diffusion of some algorithmic innovations.
• Welfare and policy interactions: regulatory regimes that prioritize sovereignty or impose localization may protect perceived legitimacy but can reduce global efficiency and raise consumer/producer costs. Conversely, predictable international frameworks for algorithmic governance could lower legitimacy vulnerability and restore some efficiency gains from delegation.
• Empirical research directions relevant to AI economics:
  • Measure GDE and ALV with proxies (share of data/compute routed through contested jurisdictions, frequency of regulatory interventions, incidence of compliance audits) and relate them to firm‑level AI adoption, automation levels, and supply‑chain performance.
  • Natural experiments/event studies: use imposition/lifting of sanctions, new data‑localization laws, or export controls as shocks to study shifts in delegation, investment in buffering, and subsequent productivity/resilience outcomes.
  • Structural models: build economic models of delegation under political risk that trade off efficiency (processing speed, error reduction) against legitimacy costs (fines, exclusion, reputation loss) to derive optimal allocation of decision rights and investment in buffering.
  • Cross‑country/firm panel analyses: test propositions linking authority configurations (recentralized, federated, trust‑partitioned, insulated) to outcomes like lead times, stockouts, cost volatility, and profit margins.
• Implications for market design and competition policy: policymakers and regulators should be aware that fragmented geopolitical rules not only alter firm incentives but also shape market structure through effects on AI architectures and fixed costs; regulatory harmonization or interoperable legitimacy standards could restore some efficiency gains from AI delegation while managing political concerns.