Algorithms are reshaping strategic decision-making: firms fall into human-dominant, sequential hybrid, or aggregated human–AI structures, and the core task is designing governance that preserves accountability and contestability when algorithms shape outcomes.

Main Finding

Algorithmic systems are increasingly active participants in strategic decision-making (SDM). Whether SDM remains meaningfully human or becomes effectively algorithmic depends on how decision rights, oversight, and contestability are organized. The paper develops a human–AI decision-structure framework (human-dominant, sequential hybrid, aggregated human–AI governance) and shows that each form reshapes not only decision accuracy but also power, learning, agency, scalability, and accountability. The central policy and managerial challenge is designing governance that preserves meaningful human control and accountability when algorithms materially shape outcomes.

Key Points

• Purpose and contribution

  • Moves beyond "AI as decision support" to examine when algorithms function as active decision participants and what that implies for firm strategy and governance.
  • Synthesizes 62 contributions across strategy, information systems, behavioral research, and governance.

• Comparative dimensions (how human vs algorithmic decision-making differ)

• Interpretive authority — who defines what counts as relevant evidence, objectives, and success.
• Search structure — how alternatives are generated, filtered, and prioritized (heuristic, exploratory, brute-force, optimization).
• Time orientation — emphasis on short-term responsiveness vs long-term planning and path dependence.
• Accountability — how responsibility, traceability, and contestability of decisions are assigned.

• Scalability — ability to replicate decisions across contexts and volumes at low marginal cost.

• Human–AI decision-structure typology

• Human-dominant: humans retain primary authority; algorithms play assistive roles (e.g., analytics, scenario generation).
• Sequential hybrid: decision flow is staged — either AI-to-human (algorithm narrows options, human finalizes) or human-to-AI (human sets goals/constraints, AI executes).

• Aggregated human–AI governance: decisions emerge from coordinated interactions/aggregation among many human and algorithmic actors (e.g., marketplaces, platform governance, ensemble systems).

• Consequences beyond accuracy

  • Power: who wields influence shifts (e.g., control of models or data confers leverage).
  • Learning: feedback loops differ; algorithms can scale learning but may lock-in suboptimal routines.
  • Agency: human sense of agency and capacity to contest decisions can erode when algorithms automate choice framing.
  • Accountability: tracking, auditing, and legal responsibility become more complex when multiple actors and opaque models are involved.

• Normative takeaway

  • The goal is not to resist algorithmic participation but to design governance that preserves decision rights, oversight mechanisms, and contestability when algorithms are active strategic agents.

Data & Methods

• Method: Theory-building integrative literature review.
• Scope: 62 scholarly contributions spanning strategy, information systems, behavioral science, and governance literatures.
• Approach:
  • Comparative analysis of human vs algorithmic decision-making across five conceptual dimensions (interpretive authority, search structure, time orientation, accountability, scalability).
  • Synthesis used to construct a conceptual framework of human–AI decision structures (three archetypes).
• Limitations (implicit/acknowledged):
  • Conceptual/theory-building work — empirical validity and boundary conditions of the typology require testing.
  • Review-based selection may reflect disciplinary emphases of the sampled literature.

Implications for AI Economics

• Firm organization and boundaries

  • Algorithmic scalability and replication can shift comparative advantages toward firms that control data, models, and deployment pipelines—favoring scale and potentially increasing concentration.
  • Hybrid decision structures alter the locus of strategic control; ownership of model design and governance becomes a core asset.

• Allocation of decision rights and incentives

  • Economics of delegation: the trade-off between automation (efficiency, scale) and maintaining human oversight (robustness, accountability) affects optimal delegation policies.
  • Incentive design must account for altered learning dynamics and principal–agent problems when algorithms learn from decisions they influence.

• Market competition and dynamics

  • Faster, algorithm-driven decision cycles compress reaction times and can increase winner-take-most dynamics in markets where speed and personalization matter.
  • Path dependence and lock-in risk rise as algorithmic routines entrench strategies, potentially reducing effective competition unless contestability is preserved.

• Labor, skills, and complementarity

  • Demand shifts toward roles that set objectives, audit models, and manage contestability; routine decision-making tasks can be displaced.
  • Complementarities between managerial judgment and algorithmic analytics determine productivity gains; mismatches can cause strategic friction.

• Regulatory and accountability considerations

  • Economic policy should aim to preserve contestability and clear assignment of liability—e.g., requirements for explainability, audit trails, human sign-off regimes depending on SDM archetype.
  • Antitrust and market regulation may need to account for power concentrated through data and algorithmic governance, not just traditional capital or IP.

• Empirical research directions for AI economics

  • Measure the degree and form of algorithmic mediation (human-dominant vs sequential hybrid vs aggregated governance) and link to firm outcomes (profitability, innovation, risk-taking).
  • Study how changes in interpretive authority and search structure affect investment, experimentation, and long-run adaptability.
  • Quantify externalities from reduced contestability (consumer harm, reduced product diversity) and benefits from scaled accuracy (lower costs, personalization).

• Policy and managerial prescriptions

  • Design governance that combines operational efficiency with oversight: clear decision rights, auditability, escalation paths, and mechanisms for contesting algorithmic outputs.
  • Invest in institutions (internal audit teams, external regulators, standards) that can certify algorithmic processes and enforce accountability to mitigate market risks from opaque, scalable decision automation.

Overall, the paper reframes SDM as a family of human–AI configurations with distinct economic effects. For AI economics, the key task is to model and measure how these configurations change firm incentives, market structure, and social welfare — and to design governance that balances gains from algorithmic scale with the preservation of contestability and accountability.