AI reshapes firms' ability to learn rather than simply replacing people: absorptive capacity now depends as much on data, models, and model governance as on human knowledge and routines, producing new complementarities and wider heterogeneity in innovation outcomes.

Main Finding

The paper argues that the classic absorptive capacity construct—built on the assumption that humans are the sole agents of knowledge absorption—must be reworked for the AI era. By combining conceptual mapping with qualitative interviews in an abductive, theory-building approach, the author proposes new AI-specific dimensions of absorptive capacity and shows how these interact with traditional human-centered dimensions. The result is a richer, hybrid view in which AI both complements and reconfigures antecedents, mechanisms, and outcomes of absorptive capacity rather than simply replacing human roles.

Key Points

• Critique of the human-only assumption: Traditional absorptive capacity theory (acquisition, assimilation, transformation, exploitation) implicitly treats humans as the only learners; AI emergence challenges that premise.
• AI introduces new functional and dimensional elements to absorptive capacity:
  • Data and model acquisition (automated sensing and intake of external knowledge/data),
  • Algorithmic assimilation (machine-driven pattern extraction and internalization),
  • Model-based transformation (AI-enabled reframing/recombination of knowledge),
  • Continuous machine learning and updating (persistent adaptation of internal representations),
  • Human-AI orchestration (coordination, interpretability, and governance of AI outputs).
• Interactions and complementarities:
  • AI dimensions can substitute for some human tasks (e.g., large-scale scanning) while amplifying others (e.g., human strategic judgment, contextualization).
  • Feedback loops arise: human inputs shape models; models change which human skills are valuable; organizational routines evolve.
  • Heterogeneity across firms and sectors: gains depend on data access, infrastructure, governance, and human capital.
• Reaffirmation, not replacement: Absorptive capacity remains fundamental, but its antecedents (e.g., prior knowledge, routines, diversity) and mechanisms (e.g., learning processes) must be reconceptualized to include AI-driven processes.
• Practical concerns highlighted: interpretability, trust, incentives, data quality, and governance mediate whether AI-driven absorptive processes lead to productive exploitation.

Data & Methods

• Approach: Abductive theory-building that iterates between conceptual mapping (literature synthesis) and primary qualitative data to generate and refine theoretical propositions.
• Conceptual mapping: Comparative analysis of existing absorptive capacity literature and emerging AI/ML literature to identify gaps and potential new dimensions.
• Empirical component: Qualitative interviews with practitioners and experts (AI developers, managers, organizational scholars) to surface real-world mechanisms, boundary conditions, and examples of human-AI interaction in knowledge processes.
• Analysis: Thematic coding of interview data integrated with conceptual insights to build propositions about new dimensions and interaction patterns. Emphasis on theory construction rather than hypothesis testing.
• Note on limits: The study is exploratory and hypothesis-generating; it does not provide causal estimates or large-N empirical validation.

Implications for AI Economics

• Measurement and modeling:
  • Empirical work on absorptive capacity should add AI-specific indicators (e.g., model deployment intensity, data assets, algorithmic update frequency, human-AI interaction metrics).
  • Economic models should treat AI as an active learning agent that can alter knowledge production, diffusion, and spillovers—not merely a productivity multiplier.
• Firm behavior and strategy:
  • Investments in data infrastructure, model governance, and human-AI interfaces become central determinants of returns to innovation and R&D.
  • Complementarity shifts: labor-technology complementarities will vary by task and sector; policy and firm decisions should account for reallocation effects.
• Productivity and growth:
  • AI-enabled absorptive capacity could accelerate knowledge diffusion and firm-level productivity if governance and interpretability enable effective exploitation.
  • Conversely, unequal data access and governance failures could increase heterogeneity and market concentration.
• Policy and market design:
  • Policies to broaden data access, support interoperable infrastructure, and ensure trustworthy AI can increase social returns from absorptive capacity.
  • Antitrust, data portability, and training programs matter for equitable diffusion of AI-enabled absorptive capabilities.
• Research agenda for economists:
  • Quantify causal impacts of AI on knowledge absorption and downstream outcomes (use IVs, diff-in-diff, natural experiments).
  • Develop micro-level measures and surveys tracking AI-enabled absorptive processes across firms/sectors.
  • Study distributional consequences (winners/losers across firms, workers, regions) and welfare implications of AI-induced reconfiguration of absorptive capacity.