Leading AI Adoption in Organizations: Introducing a Behavioral Human-Centered Approach
Citation. Schweitzer, S., Narayanan, D., McGuire, J., & De Cremer, D. (2025). Leading AI adoption in organizations: Introducing a behavioral human-centered approach. International Journal of Human–Computer Interaction. https://doi.org/10.1080/10447318.2025.2531287
Summary. Schweitzer and colleagues argue that organizational AI adoption fails when leaders treat implementation as an engineering exercise and assume humans will simply fall in line. They integrate change management with human-centered AI to propose behavioral human-centered AI, a framework that treats design, adoption, and management as behavioral stages shaped by biases, perceptions, trust, resistance, and stakeholder well-being. The article is conceptual rather than empirical, but it gives the wiki a useful map of why AI adoption requires behavioral design and managerial discipline, not just technical capability.
Key claims
- AI adoption is a distinctive change-management problem because AI can perform sophisticated cognitive tasks, surveil and reward employee behavior, operate opaquely, and alter employee roles, values, autonomy, identity, and perceived skill value.
- Existing human-centered AI approaches often over-rely on rationalist assumptions about users, designers, and leaders; a behavioral approach should account for bounded rationality, algorithm aversion, trust loss after algorithmic mistakes, overconfidence, loss aversion, and other systematic biases.
- Behavioral human-centered AI should center relevant stakeholders across three stages: designers in the design stage, organizational members in the adoption stage, and leaders in the management stage.
- In the design stage, confirmation bias, inventor’s bias, and anchoring on technical specifications or KPIs can produce AI systems that appear technically sound but fail human usability, fairness, acceptance, or well-being tests.
- In the adoption stage, availability bias and loss aversion can make employees overweigh vivid AI failures, job-loss narratives, safety concerns, surveillance risks, or autonomy threats relative to potential benefits.
- In the management stage, leader overconfidence and escalation of commitment can sustain ineffective or harmful AI investments even when leaders lack sufficient AI expertise or evidence that the system benefits end users.
Evidence & limitations
- Conceptual synthesis drawing on change management, human-centered AI, behavioral decision research, and organizational AI adoption literature; it does not report a new empirical study.
- The article is useful for framing behavioral mechanisms and research questions, but its stage model and bias mapping should be treated as a theory-building contribution until tested in field implementations.
- Public provenance is verified through the publisher DOI and embedded Taylor & Francis DOI links in the PDF; the Drive PDF hash is recorded for version control.