Invisible Groupthink: AI Homogenization, Epistemic Diversity, and Organizational Decision Resilience

Organizations worldwide are investing rapidly in AI tools for knowledge work, yet evidence consistently shows that this investment is not translating into proportional decision quality improvements. This paper argues that one underexamined mechanism explains both the productivity gap and an overlooked organizational risk: when teams simultaneously consult homogeneous AI systems, they unknowingly undergo a process we term Invisible Groupthink. Invisible Groupthink is an AI-mediated conformity that operates without the social pressure or directive leadership that classical groupthink requires, and which is therefore structurally invisible to the organizational safeguards designed to prevent it. Drawing on AI sycophancy research, collective intelligence theory, epistemic diversity literature, and organizational learning theory, we develop a formal mechanism called the Informational Quiet Homogenization Effect (IQHE). We derive from it the Amplification Paradox, a counterintuitive prediction that organizations investing most in uniform AI deployment may develop the least resilient decision cultures, because AI investment reduces the epistemic diversity that enables teams to detect and correct errors. We propose Diversity-Preserving AI Governance (DPAG) as a targeted organizational remediation framework. This paper makes organizational-level contributions that extend individuallevel and model-level work in existing literature.