Whether AI 'dominates' work is a strategic choice, not a technical inevitability: codifiable manufacturing tasks show clearer displacement patterns, while marketing's judgment‑heavy tasks primarily see augmentation, and governance, task-allocation design, and reskilling determine longer‑run outcomes.

Main Finding

AI does not deterministically "dominate" work; its labor-market effects are strategically contingent. Task composition (codifiability vs. judgment/relational intensity), organizational deployment choices (full automation vs. calibrated human–AI collaboration), accountability weight (verification overhead and non-transferable liability), and governance quality together determine whether AI substitutes for or augments human labor. Empirically and conceptually, manufacturing—because of more codifiable tasks—shows stronger substitution signals, while marketing—because of judgment- and relationship-heavy tasks—shows stronger augmentation signals. Across domains, hybrid human–AI designs more often outperform full automation in productivity, resilience, and workforce capability.

Key Points

• Sectoral divergence is explained by task composition, not by an intrinsic difference in AI capability:
  • Manufacturing: codifiable, routine tasks (quality inspection, packaging, predictive maintenance) generate clear substitution pressures in many contexts.
  • Marketing: judgment-, creativity-, and relationship-centered tasks (personalization, creative strategy, emotional engagement) produce augmentation and reinstatement effects.
• Task-based automation–augmentation theory (Acemoglu & Restrepo tradition) provides the mechanism: automation displaces tasks while augmentation can raise labor demand and create new tasks; displacement typically acts faster than reinstatement.
• Fully automated deployment architectures often underperform calibrated human–AI collaboration:
  • Evidence shows gains from explainable, human-overseen systems and hybrid teams (e.g., human–AI teams raising iterative-design productivity; explainable systems improving performance vs. opaque automation).
  • Full automation can erode human skill acquisition, autonomy, and organizational resilience, making recovery from failures harder.
• Accountability weight matters: functions with regulatory/fiduciary accountability (triangulated via a Philippine accounting case) impose verification overhead and non-transferable liability that constrain automation depth independent of technical feasibility.
• Governance and leadership quality are critical, sector-agnostic mediators:
  • Calibrated leadership commitment and transparent algorithmic governance improve implementation success and employee well‑being.
  • Poorly designed algorithmic management (low transparency, high surveillance) harms job satisfaction; transparent, autonomy-preserving designs can have neutral or positive welfare effects.
• Practical recommendations (conceptual):
  • Task-level portfolio mapping to identify which tasks to automate, augment, or retain human-led.
  • Parallel reskilling investments emphasizing both technical and soft skills.
  • Calibrated leadership involvement (timing and intensity) during rollouts.
  • Explicit assessment of accountability weight when deciding deployment depth.
• Limitations noted by the authors: the paper is a structured conceptual literature review (non-empirical) and calls for empirical validation of the proposed framework.

Data & Methods

• Approach: structured conceptual literature review and thematic synthesis across five areas:
• AI and manufacturing automation
• AI in marketing and customer-facing functions
• Substitution vs. augmentation theory (task-based frameworks)
• Organizational redesign and human capital strategy
• Strategic governance of AI-integrated work systems
• Sources: interdisciplinary scholarship spanning labor economics, innovation management, organizational behavior, plus empirical studies and sectoral case evidence (including a triangulating case from Philippine accounting practice).
• Theoretical grounding: task-based automation–augmentation models (e.g., Acemoglu & Restrepo lineage) and organizational design literature on human–AI complementarities.
• Empirical references used illustratively in the review include: VAR analysis in South African manufacturing; firm and sector case studies (manufacturing, e‑commerce, marketing deployments such as Unilever); human–AI collaboration experiments; surveys on worker perceptions; and Philippine-specific organizational studies.
• Methodological caveat: the paper synthesizes existing empirical findings but does not present new primary empirical analysis; authors emphasize need for cross-sector empirical validation (panel studies, microdata, field experiments) of the conceptual framework.

Implications for AI Economics

• Model heterogeneity: economic forecasts and models of AI’s labor impact must incorporate task composition and organizational choice variables, not just aggregate technical capability measures. Predictive models that ignore firm-level deployment design and governance risk overestimating substitution.
• Timing and dynamics: displacement tends to materialize faster than job-creating reinstatement; macro forecasts should account for temporal lags and transitional frictions (skills mismatch, reskilling delays).
• Sectoral policy targeting:
  • Manufacturing-focused policies should anticipate displacement risk concentrated in codifiable tasks and prioritize social insurance, targeted retraining for maintenance/oversight roles, and policies that encourage augmentation pathways where possible.
  • Service/marketing-type sectors may see net complementarities; policies should focus on enabling smaller firms to access augmentation technologies and on data-privacy/consumer-trust regulation to preserve returns to personalization.
• Labor-market distributional risks: productivity gains may be uneven (by firm size and sector), potentially exacerbating inequality unless paired with inclusive reskilling and access initiatives.
• Governance as an economic variable: algorithmic governance, accountability weight, and leadership calibration meaningfully alter technological returns and welfare outcomes—these should be operationalized in empirical work (e.g., indices for governance quality, accountability cost measures) and included in policy evaluation.
• Measurement priorities for researchers and policymakers:
  • Develop microdata linking task-level automation intensity, deployment architecture (full automation vs. hybrid), governance metrics, and labor outcomes (employment, wages, skill composition).
  • Use longitudinal and quasi-experimental designs to estimate displacement vs. reinstatement timings and magnitudes.
  • Experimentally compare deployment architectures (e.g., opaque automation vs. explainable human-overseen systems) for productivity, skill retention, and well-being outcomes.
• Practical takeaway for firms and policymakers: avoid technodeterministic prescriptions. Strategic choices—task allocation, reskilling, governance—shape whether AI will substitute or augment labor and determine long-run organizational and economic outcomes.

Reference: Tolentino, G., Espelita, C. A. M., Quinto, L., Teodosio, G. M., & Atento, A. G. (2026). Artificial Intelligence and the Strategic Reconfiguration of Work: A Conceptual Analysis of Substitution, Augmentation, and Organizational Redesign in Manufacturing and Marketing. Journal of Enterprise Strategy and Management Innovation. https://doi.org/10.65166/bt13jr83

(If you want, I can extract and summarize the specific empirical findings cited in the paper—e.g., the VAR result, conversion/retention numbers, human–AI performance differentials—into a short table for use in modeling or policy memos.)