The Non-Human Innovator: Agentic AI, Physical AI, and the Transformation of R&D Management

Éditeurs invités

• Joon Mo Ahn, Korea University
• Alberto Di Minin, Scuola Superiore Sant'Anna
• Sungjoo Lee, Seoul National University
• Ahreum Hong, Kyung Hee University

Synthèse

This special issue addresses how artificial intelligence systems function as autonomous innovators rather than tools within R&D management. As AI systems including large language models and physical AI become active participants in innovation processes, existing frameworks assuming human actors must be reconceptualized. The core challenge extends beyond how AI assists humans to how much cognitive work should be delegated to AI and what consequences this has for organizations, ecosystems, and intellectual property.

The special issue examines four interconnected themes: determining optimal levels of AI delegation; understanding how agentic and physical AI reshape innovation ecosystem architectures as autonomous actors; reconceptualizing absorptive capacity for evaluating AI-generated knowledge; and addressing IP ownership and inventorship when non-human systems generate patentable outputs. Collectively, these themes interrogate how the distinction between AI-as-tool and AI-as-innovator requires new governance mechanisms and strategic frameworks for innovation management.

Thèmes proposés

• Is there an inverted-U relationship between the degree of AI delegation in R&D and innovation performance, mirroring the over-search paradox?
• How do firms determine the optimal breadth and depth of AI involvement across different stages of the R&D process? Specifically, under what conditions does algorithmic management enhance or hinder innovation performance?
• What is the AI-era analogue of the NIH syndrome: how does uncritical acceptance of AI-generated knowledge erode internal expertise?
• What governance architectures enable effective human–AI teaming while preserving accountability, creativity, and strategic judgment?
• How do agentic AI systems alter the architecture of knowledge flows in open innovation ecosystems, and what new governance mechanisms are required?
• How does the participation of autonomous AI agents change the logic of platform-based open innovation, including roles, incentives, and boundary conditions?
• How do digital twins and AI-enabled simulation reshape the scope and speed of distributed experimentation across organisational boundaries?
• What new forms of inter-organisational trust, contracting, and coordination are needed when AI agents act as innovation partners?
• How must absorptive capacity (or dynamic capabilities) be reconceptualised when the primary external knowledge source is an AI system rather than a human partner?
• What organisational routines and managerial processes enable firms to transform AI-generated knowledge into innovation value?
• What individual competencies — technical, cognitive, and relational — distinguish high-performing innovators in human–AI contexts?
• What learning mechanisms allow firms to continuously upgrade AI-related innovation capabilities over time, particularly in relation to exploration and exploitation?
• How should inventorship and IP ownership be attributed when agentic AI systems autonomously generate patentable outputs, and what theoretical frameworks from open innovation research best capture this challenge?
• What strategic logic governs firms' decisions to release model weights, training data, or fine-tuned AI capabilities into open-source commons?
• How do trained model weights, fine-tuning data, and emergent AI capabilities constitute a new category of strategic asset, and how do firms govern access to and monetisation of these assets?
• Under what regulatory and institutional conditions does the open-source AI movement accelerate versus impede innovation diffusion?