Product lines that combine automation and augmentation outperform peers; single-mode AI strategies do not — smart features must match task needs. The study finds positive market effects only for a flexible dual AI strategy, highlighting 'functional subsidiarity' across problem identification, solution development and implementation stages.

Main Finding

Only a dual AI strategy (combining automation and augmentation) positively affects market performance of product lines. Crucially, smart capabilities must be matched to the functional needs of each task within the solution delivery process — a principle the authors call functional subsidiarity. Automation alone or augmentation alone do not produce significant market-performance gains.

Key Points

• Sample: 667 distinct product lines observed in 2023, comparing AI-enhanced “smart” product lines to non-AI counterparts.
• Strategies examined: automation, augmentation, and dual (automation + augmentation).
• Core principle: functional subsidiarity — AI capacity yields benefits only when aligned to the specific functional requirements of tasks across the solution delivery process.
• Performance result: neither pure automation nor pure augmentation has a significant effect on market performance; only the dual strategy yields positive effects.
• Process focus: effectiveness is evaluated at the level of strategic processes (problem identification, solution development, solution implementation) rather than isolated tasks.
• Methodological rigor: Kernel-based propensity score matching used to create comparable treatment and control groups, reducing confounding from selection into AI adoption.

Data & Methods

• Data: Unique cross-sectional sample of 667 product lines from 2023 (product-line level analysis).
• Outcome: Market performance of product lines (study focuses on relative performance of AI-enhanced vs non-AI lines).
• Treatment classification: Product lines categorized by AI strategy implemented in solution delivery — automation, augmentation, or dual.
• Identification strategy: Kernel-based propensity score matching to balance observables between AI-enhanced and non-AI product lines, improving causal comparability.
• Process decomposition: Analysis disaggregates the solution delivery process into problem identification, solution development, and solution implementation stages to test where AI strategies matter.

Implications for AI Economics

• Strategy and complementarities: Firms capture value from AI when they combine automation and augmentation, highlighting complementarity rather than one-size-fits-all adoption. Investment decisions should favor flexible, hybrid AI deployments that can switch between automating routine tasks and augmenting human decision-making.
• Task–technology fit: Functional subsidiarity stresses the economic importance of matching AI capabilities to task requirements; misaligned AI investments may yield no market-payoff despite high technical capacity.
• Organizational design and capability investment: Optimal returns require process-level alignment (capabilities across identification, development, implementation), suggesting returns to coordination, governance, and managerial attention—factors that affect diffusion and productivity of AI at the firm level.
• Labor and skill implications: Dual strategies imply persistent complementarities between AI and human labor (augmentation) coexisting with task displacement (automation), affecting wage structures, training investments, and labor reallocation within product lines.
• Market structure and competition: Firms that successfully deploy dual strategies and manage functional subsidiarity may gain competitive advantages in product-market performance, potentially increasing concentration in markets where process complexity favors hybrid AI deployments.
• Policy relevance: Policies that support workforce retraining, modular adoption of AI, and incentives for firms to develop hybrid AI systems could improve aggregate value capture and mitigate risks of misaligned automation.

(Brief caveat: summary reflects the study’s reported sample, methods, and high-level findings; the original paper should be consulted for metric definitions, robustness checks, and auxiliary analyses.)