Cognitive Ambidexterity: Balancing Algorithmic Efficiency and Contextual Judgement in AI-Enabled Decision Systems
AI-enabled decision systems make algorithmic recommendations appear objective and ready for action before contextual, ethical, and stakeholder concerns are fully considered. This study examines how leaders preserve judgment under these conditions. We develop cognitive ambidexterity as a leader-level meta-cognitive capability through which managers balance algorithmic exploitation with contextual exploration in AI-mediated decisions. Using an exploratory sequential mixed-methods design, we first examine five U.K. procurement and infrastructure organizations and identify four mechanisms of cognitive ambidexterity. We then examine these mechanisms through vignette experiments with managers. The findings show that leaders preserve contextual judgment by treating AI outputs as interpretive cues, calibrating trust under ambiguity, slowing decisions when reflection is needed, and building supports that make algorithmic recommendations contestable. The study extends ambidexterity theory by showing how exploration–exploitation tensions are cognitively regulated when exploitation is embedded in algorithmic decision processes. It also uses leadership cognition and Upper Echelons Theory to explain why leaders differ in their capacity to enact this regulation. Together, these insights position cognitive ambidexterity as a critical leader-level capability for using AI’s efficiency gains while preserving the contextual judgment.
Authors
- Tahir Abbas Syed (ORCID: https://orcid.org/0000-0002-0770-1753)
Institutions
- University of Manchester (GB)
Publication Details
- Journal
- Journal of Management
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1177/01492063261501111
- Primary Topic
- Artificial Intelligence Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00