Delegating for the Greater Green? How competence and autonomy shape acceptance of agentic AI for carbon-intensive choices

Would you be willing to delegate control to AI for environmental benefits? This study examines users’ willingness to entrust decisions to agentic AI systems designed to promote sustainable behaviours, focusing on how satisfaction of the basic psychological needs – autonomy, competence, and relatedness – mediates intentions to delegate. A 3 × 2 online experiment (N = 1,522) asked participants to evaluate three activities: two carbon-related (food shopping and digital usage) and one non-carbon task (email management), under either full-autonomy or human-in-the-loop conditions. Acceptance differed by context: delegation intentions were lowest for food shopping and highest for digital usage. A preference for human-in-the-loop over fully autonomous systems emerged only in the food domain. Psychological needs fully mediated the advantage of digital usage over email management, but only partially mediated both the food domain's lower intentions and its sensitivity to agency level, suggesting that identity-relevance and hedonic value exert an independent influence beyond the SDT framework. Competence was the strongest mediator throughout, indicating that users favour agentic AI primarily when it credibly enhances their capacity to achieve task goals. Effective sustainable AI design should therefore prioritise utilitarian domains, avoid full automation in identity-relevant contexts, and support competence and autonomy together, preserving user agency.