BEHAVIORAL OPERATIONS AND HUMAN–AI COLLABORATION: EFFECTS OF ALGORITHMIC DECISION SUPPORT ON MANAGERIAL JUDGMENT, EMPLOYEE TRUST, PRODUCTIVITY, AND ORGANIZATIONAL PERFORMANCE

Abstract

The rapid penetration of artificial intelligence (AI) and machine learning into business processes has transformed the way managers make decisions, the way supply chains execute, and organizations design themselves. Indeed, early technology utopias promoted a fully automated “no-touch” style of environment, but what is becoming apparent is that value creation for organizations lies in the quality of the human–AI interaction as a socio-technical system not a product or service that can be easily be switched out of a solution stack. This review integrates behavioral operations research to explore the impact of algorithmic decision support on managerial judgment, employee trust, productivity, and organizational performance. Based on trust in automation theory, dual‐process cognitive paradigms, and literature on algorithm aversion and appreciation, this paper provides a theoretical analysis on four collaboration modes (full automation, adjustable automation, human augmentation, and strategic delegation). Results demonstrate a strong delegation asymmetry: a high confidence accurate self-assessment allows AI systems to delegate efficiently, human managers, however, are metacognitive impaired, they hold on to tasks they are bad at while delegating tasks they are best at. Environmental boundary conditions add another layer of complexity: human overrides are found to improve accuracy in long-horizon, low-uncertainty contexts, but to impair accuracy in short-horizon, volatile contexts. Trust is identified as the dominant psychological process through which transparency, interpretability, and decision alignment influence planning effectiveness, with task complexity exerting a positive moderating influence on such relationship. The review also surfaces a blamemaintenance paradox in which managers offload ethically fraught decisions to algorithms while holding onto credit for positive outcomes, as well as productivity hazards from uncontrolled introduction. The article argues for context-sensitive override rules, explainability as the default, explicit delegation support, metaknowledge training, and institutional governance that maintain human moral authorship for sustainable performance.