Invention alone no longer guarantees transformative change; system redesign, organizational leadership and aligned incentives turn technological breakthroughs — including AI — into scalable societal impact.

Main Finding

Despite exponential increases in scientific output and R&D spending, the real-world, transformative impact of innovation is declining (the “Innovation Paradox”). Technological breakthroughs—AI included—become game-changing only when embedded in redesigned business models, organizational capabilities, ecosystem coordination, and institutional arrangements. Absent those systemic changes, invention remains invention and rarely produces broad societal or market transformation.

Key Points

• Innovation paradox: More papers/patents and larger research teams are producing less field-disruptive work and diminishing marginal returns to R&D (Park et al. 2023; Bloom et al. 2020).
• Technology ≠ disruption: Technical novelty alone rarely reshapes markets; adoption requires alignment of pricing, usability, timing, business model, and institutional support (Rogers; Christensen; Teece).
• Organizational dynamics: Firms favor exploitation over exploration, biasing toward incrementalism (March). Successful reinvention requires absorptive capacity and dynamic consistency.
• Case studies:
  • Cinema: Survived waves of tech change by integrating innovations that strengthen storytelling, not by treating tech as an end (Elizabeth Daley).
  • Sony: Shifted identity from hardware to creator-centered platform, recombining capabilities across domains (Hiroaki Kitano).
  • Amazon/AWS: Turned infrastructure into a platform/business-model innovation (elastic pay-as-you-go cloud) that enabled whole new industries (Kim Majerus).
  • Energy (Digital Grid example): Renewable integration demands system-level coordination, market redesign, and real-time platform technologies rather than isolated device improvements (Yusuke Toyoda).
• AI-specific notes:
  • Generative AI can amplify creativity but also produce convergent outputs and reduce collective diversity (Doshi & Hauser 2024).
  • The future of work will center on task re-composition and human–AI complementarity; readiness of workers and institutions is decisive (Autor).
• Normative emphasis: Values, governance, and purpose matter—technologies must be embedded in societal and institutional frameworks to be transformative.

Data & Methods

• Quantitative analyses referenced:
  • Park, Leahey, and Funk (2023): Analysis of ~45 million scientific papers and ~3.9 million patents across six decades using a Convention Disruption (CD) index to measure disruptiveness; finds declining disruptive potential over time.
  • Bloom et al. (2020): Empirical study of R&D productivity showing exponentially larger teams/investments are required for incremental progress (e.g., maintaining Moore’s Law).
  • Doshi & Hauser (2024): Experimental/empirical work showing AI-assisted writing can increase assessed creativity but cause convergence in outputs.
• Conceptual and theoretical frameworks used: Diffusion of Innovations (Rogers), disruptive innovation (Christensen), exploration vs. exploitation (March), absorptive capacity (Cohen & Levinthal), business model innovation (Teece; Casadesus-Masanell & Zhu), digital transformation frameworks (Hanelt et al.; Westerman et al.).
• Empirical basis of chapter: synthesis of large-scale bibliometric/patent studies, economic analyses of R&D productivity, experimental studies on AI creativity, and qualitative evidence from a 2025 STS forum: expert panels and industry case studies (Sony, Amazon, Digital Grid, cinema stakeholders).
• Methodological character: mixed—combines citation/patent bibliometrics, macroeconomic R&D analysis, experiments, and case-based qualitative inference.

Implications for AI Economics

• Measure impact beyond counts: Economic studies of AI should move beyond input metrics (papers, models, compute) and incorporate disruptiveness/impact measures (e.g., CD-style indices), business-model change, and ecosystem effects.
• Complementarity and tasks: AI’s economic effects will largely depend on complementarities with human skills and how tasks are recomposed. Models of labor-market impact must incorporate task reallocation, skill complementarities, and firm-level adoption constraints.
• Business model and platform effects: AI diffusion can enable new platform models (like AWS) that change market structure, fixed vs. variable cost economics, and entry dynamics. Economic analyses should model platform-mediated spillovers, multi-sided markets, and winner-take-most dynamics.
• R&D productivity and returns: Large-scale AI investments risk falling into the same decreasing-return pattern unless paired with organizational change and absorptive capacity. Policy and firm strategy should consider how to increase marginal returns through deployment, complementary investments, and vintage human capital.
• Systemic vs. component innovation: For sectors with high systemic complexity (energy, healthcare, transport), AI’s value depends on institutional redesign (markets, regulation, governance). Economists should evaluate AI interventions in context of market design and coordination frictions.
• Diversity and cultural externalities: Research should track whether AI-induced convergence in creative outputs reduces cultural diversity and how that affects downstream markets and welfare.
• Policy & inequality: Platformization and winner-take-most dynamics can intensify concentration—policy must address competition, infrastructure access, skills, and redistribution to ensure inclusive benefits.
• Research agenda suggestions:
  • Integrate firm-level case studies with macro bibliometric indicators to trace when invention becomes commercial disruption.
  • Model adoption lags from institutional, governance, and business-model frictions.
  • Empirically estimate complementarities between AI capital and human skills across occupations and sectors.
  • Evaluate metrics for societal impact (resilience, inclusion, public goods) alongside productivity gains.

Summary takeaway: AI’s economic promise will be realized only when technical capabilities are integrated with new business models, organizational redesign, market institutions, and social governance. AI economics should therefore study not just the technology, but the ecosystems and institutions that convert invention into broad-based economic transformation.