It's Amazing – But Terrifying!: Unveiling the Combined Effect of Emotional and Cognitive Trust on Organizational Member' Behaviours, AI Performance, and Adoption

Abstract We conducted a qualitative, real‐life study where we tracked the introduction, implementation, and use of a new AI technology in a company. We identified four distinct trust configurations among organizational members: full trust (high cognitive/high emotional), full distrust (low cognitive/low emotional), uncomfortable trust (high cognitive/low emotional), and blind trust (low cognitive/high emotional). Furthermore, we found that organizational members exhibited distinct behaviours under the four trust configurations: Some responded by detailing their digital footprints, while others engaged in manipulating, confining, or withdrawing them. These behaviours triggered a ‘vicious cycle’, where biased (due to manipulation) and unbalanced and asymmetric (due to detailing, confining, or withdrawing) data inputs degraded AI performance, further eroding trust and stalling adoption. Our primary contribution is a model that explains how organizational members behave under different trust configurations and how these behaviours affect AI performance and, ultimately, AI adoption in organizations. We also provide valuable insights for managers.