Companies that build comprehensive AI capabilities outperform peers, showing a roughly 10.7 percentage-point shareholder return premium, but only about 1% reach maturity. The study finds organizational redesign and workflow reconfiguration—not plug-in tools—are the critical bottlenecks as many firms abandon AI efforts due to data, skills, and integration challenges.

Main Finding

AI is reshaping business strategy by creating six distinct sources of competitive advantage (data differentiation; digital core strength; rate of learning; depth of capability reinvention; external partnerships; level of trust). Firms that built advantages across all six captured a sizable shareholder-return premium (≈10.7 percentage points). However, broad adoption has not translated into maturity: roughly three‑quarters of firms use AI in at least one area, but only ~1% report mature AI capability and ~42% abandon initiatives. The study concludes that substantive organizational redesign—especially workflow reconfiguration—is the strongest predictor of realized business value; superficial technology overlays are insufficient.

Key Points

• Six AI-driven sources of competitive advantage identified:
• Data differentiation (proprietary, high-quality data ecosystems)
• Digital core strength (robust engineering/platform capabilities)
• Rate of learning (ability to iterate and update models/knowledge fast)
• Depth of capability reinvention (reengineering core capabilities rather than point solutions)
• External partnerships (ecosystem and partner integration)
• Level of trust (internal and external trust, governance, ethics)
• Aggregate impact estimates: generative AI and related technologies imply potential economic value of US$2.6–4.4 trillion annually (cited from McKinsey & corroborated in the study).
• Market evidence: firms achieving strengths across all six capability areas realize a measurable market premium (~10.7 pp total shareholder return; median incremental market cap ~$300–500M for a Fortune 500 firm).
• Adoption vs. maturity gap: 55%→78% (2023–2024) adoption growth; only ~1% of firms are “mature” in AI; abandonment of initiatives rose to ~42% (2024/25).
• Principal barriers to scaling AI: data preparedness, skills shortages, systems/integration difficulties, and insufficient organizational transformation.
• Workflow redesign (end‑to‑end process reconfiguration) is the best single predictor of business impact from AI.
• Generative AI (post‑Nov 2022) accelerated prototyping, product development and broadened access to powerful AI tools beyond large tech incumbents.
• Theoretical synthesis: combines Resource-Based View (RBV), Dynamic Capabilities Framework (DCF), and Technology Acceptance Model (TAM), augmented with “situated AI” framing (grounding, bounding, recasting).

Data & Methods

• Timeframe: January 2022 – September 2025 (focus on the generative-AI era).
• Multi‑stream mixed methods (secondary data only):
  • Bibliometric and literature review: initial retrieval ~1,039 articles across Scopus/Web of Science/Google Scholar; refined sets reported (the paper cites systematic reviews of 180 articles and elsewhere a curated set of 314 articles used for thematic synthesis).
  • Systematic literature synthesis using Webster & Watson protocol; co-word and citation mapping (VOSviewer, R‑bibliometrix).
  • Financial panel analysis: data from Bloomberg, SEC EDGAR, company filings covering 427 Fortune 500 companies (sales growth, EBIT contribution, total return to shareholders, R&D, AI investments).
  • Case studies: 28 purposively sampled organizations across industries (documented AI implementations, outcomes).
  • Industry reports: synthesis of ~45 authoritative industry reports (McKinsey, Gartner, BCG, Deloitte, Forrester, PwC, IBM).
  • Patent analysis: >1,200 AI patents (2022–2024) to measure innovation outputs.
• Analytical approaches:
  • Bibliometric network & co-word analyses to map research trends.
  • Thematic qualitative synthesis of case studies and literature.
  • Quantitative panel comparisons linking AI capability metrics to financial performance (difference-in-means, panel regressions, descriptive return-premium calculations).
  • Hypothesis testing around RBV, DCF mediation/moderation, and TAM-related diffusion effects (H1–H6 in paper).

Implications for AI Economics

• Firm-level value creation and heterogeneity
  • AI creates heterogenous returns: firms that invest in complementary organizational capital (data assets, engineering platforms, workflow redesign) capture outsized returns — empirical support for RBV-style heterogeneity.
  • The 10.7 pp shareholder premium indicates measurable market recognition of integrated AI capability, implying that asset markets price organizational complements to AI, not just raw AI tools.
• Market structure and winner-take-all dynamics
  • Evidence of concentrated benefits and high abandonment rates supports potential winner-take-all or winner-take-most industry dynamics; scale and ecosystem advantages (data & partnerships) can be rent sources.
• Role of organizational capital and dynamic capabilities
  • Productivity gains from AI are conditional on investments in dynamic capabilities (sensing/seizing/transforming) and workflow reconfiguration. Economists should model returns to AI jointly with investments in organizational change, training, and process redesign.
• Innovation and aggregate productivity
  • Generative AI accelerates R&D and product design via faster candidate generation and evaluation; potential to reverse productivity declines in certain R&D-intensive sectors. Measuring spillovers from AI-enabled innovation (patent flows, cross-firm diffusion) is important.
• Labor markets and skills
  • High abandonment and skills shortage findings highlight frictions in labor reallocation and human capital formation. Policy and firm-level training investments determine how productivity gains translate into wages, employment composition, and inequality.
• Measurement and identification challenges
  • Secondary-data, observational design limits causal claims; future work needs causal identification (experiments, quasi‑experiments) to separate AI tool effects from complementary investments. Economists should incorporate measures of workflow redesign, data quality, and governance when estimating AI returns.
• Policy/regulatory considerations
  • Trust, governance, and ethics are strategic inputs that affect adoption and market rewards. Regulation affecting data sharing, privacy, and AI governance will influence the distribution of AI rents and competition across firms and sectors.

Limitations noted by the authors: reliance on secondary data and mixed-source synthesis (potential sampling and measurement inconsistencies), short post‑generative‑AI timeframe (2022–2025) limiting long-run inference, and heterogeneity across industries that complicates generalization. Recommended next steps: causal studies on scaling mechanisms, measurement of organizational-capital returns, and research on labor-market adjustments to AI-driven restructuring.