When Hiding Hurts: How Stealth Use of Generative AI Resources Impairs Task Performance

AbstractThe rapid proliferation of Generative Artificial Intelligence (Gen AI) has created an intriguing paradox in organizations: while employees increasingly recognize Gen AI’s potential to enhance productivity and innovation, many use these tools covertly. Drawing on dual-process theory, we introduce and validate the concept of Stealth Use of Generative AI Resources (SUGAR)—employees’ hidden use of Gen AI tools within professional settings. Through rigorous scale development across multiple samples in the UK and China, we establish SUGAR as distinct from related constructs including AI usage and attitudes. Analysis of multi-wave data from 207 employees reveals that SUGAR undermines task performance by disrupting employees’ natural learning processes and triggering feelings of professional inadequacy. Specifically, while general AI usage positively relates to performance, concealing such usage reduces reflective learning opportunities and increases impostor feelings, ultimately reducing performance. Our findings advance organizational theory by revealing how AI concealment creates unique cognitive demands that cannot be overcome by motivational factors, while also demonstrating how employees’ well-intentioned efforts to leverage AI capabilities through covert use may ultimately impair their performance. This research initiates important theoretical conversations about the unintended consequences of undisclosed AI usage in organizations and provides practical insights for developing more nuanced approaches to AI governance that balance productivity with transparency. Keywords: Generative AI, stealth use of generative AI resources, dual-process theory, reflective learning, imposter feeling, self-improvement motivation, task performance