Leaders use it First: How Leadership Role Modeling Shapes the Sources and Grounding of AI Self-efficacy in Knowledge-intensive Roles in Morocco

Abstract

Artificial intelligence is reshaping knowledge-intensive work, yet organizations repeatedly confront a gap between technical access to AI tools and workers’ confidence in using them. This exploratory qualitative study asks how leadership role modeling shapes the development of AI task self-efficacy among knowledge workers, and how that self-efficacy relates to their job performance. Drawing on ten semi-directive interviews across five professional domains in Morocco (consulting, finance, technology, medicine and academia), we bring the Technology Acceptance Model (Davis, 1989) into dialogue with Bandura’s (1997) Social Cognitive Theory, treating organizational support as a moderating condition. Thematic analysis surfaces four interlocking patterns. Formal training appeared largely inconsequential for self-efficacy, as participants experienced mandatory programmes as compliance-driven box-ticking. Vicarious learning, watching leaders use AI in the ordinary course of work, emerged in participants’ accounts as the primary organizational pathway to self-efficacy, supplying a symbolic signal that legitimated adoption; where such modeling was absent, notably in medicine and academia, self-efficacy appeared to stagnate. Perceived usefulness seemed to generate self-efficacy only indirectly, by prompting the engagement that produces mastery, and only where credible role models were present. Finally, the self-efficacy-to-performance link appeared conditioned by attribution style, itself shaped by leadership context. We theorize a mechanism chain (modeling, legitimacy, engagement, mastery, self-efficacy, performance) moderated by attribution style and sectoral context, and offer it as propositions for later quantitative testing. The study is deliberately exploratory and non-generalizable; its contribution is to reposition leadership as a cognitive, not merely behavioral, enabler of AI adoption in an under-studied emerging-economy setting.