Generative AI Use and Innovation among University Administrative Employees: The Roles of Dynamic Capabilities, Knowledge Sharing, and AI Sycophancy

Abstract

Generative artificial intelligence is increasingly used in university administration for document preparation, information retrieval, policy interpretation, data summarization, meeting support, and service improvement. However, faster content production does not necessarily translate into meaningful administrative innovation, particularly when AI feedback is overly agreeable and insufficiently critical. This study examines how generative AI–enabled dynamic capabilities and participation in AI-related knowledge sharing are associated with perceived innovation performance among university administrative employees. It further investigates perceived AI sycophancy as a negative boundary condition on both relationships. A survey was conducted among 428 administrative employees from higher education institutions in Liaoning, Jilin, and Heilongjiang provinces, located in Northeastern China, who had used generative AI in administrative work. Confirmatory factor analysis, structural equation modeling, hierarchical moderated regression, and simple-slope analysis were employed. The findings indicate that generative AI–enabled dynamic capabilities and AI-related knowledge-sharing participation are positively associated with perceived innovation performance, whereas perceived AI sycophancy weakens both relationships. The results extend dynamic capabilities theory and the knowledge-based view to the university administrative context and highlight the importance of preserving critical independence in human–AI interaction. Universities should combine AI capability development with verification procedures, cross-departmental knowledge-sharing mechanisms, and prompting practices that encourage alternative viewpoints and critical evaluation.