Serendipity with Generative AI: Repurposing knowledge components during polycrisis with a Viable Systems Model approach

Researchers Gordon Fletcher and Saomai Vu Khan have published a paper demonstrating how generative AI can serve as a 'serendipity engine' to help organizations navigate polycrisis uncertainty. Their approach focuses on discovering, classifying, and mobilizing reusable knowledge components—such as models, frameworks, and patterns—that are often overlooked within existing organizational documents. The researchers propose a theory of 'planned serendipity' where generative AI lowers the transduction costs between different subsystems of Beer's Viable System Model (VSM), enabling more efficient knowledge repurposing during complex crises.

The team empirically validated their approach by processing 206 academic papers through their AI pipeline, extracting 711 distinct reusable components (approximately 3.4 per paper) and organizing them into a structured repository aligned with VSM principles. Beyond the conceptual framework and empirical repository, the research provides practical managerial tools including adoption vignettes and process blueprints. The work suggests testable links between repository creation, reduced discovery-to-deployment timelines, and increased reuse rates, with implications for shifting organizational innovation portfolios away from breakthrough bias toward systematic repurposing that delivers both environmental and social benefits.