Partner or driver? How generative AI role framing shapes perceived cognitive agency and employee-driven digital innovation
S. Uppal Memon, Shang Yufan, Amjad Hussain Laghari, Wali Muhammad Khoso · Behaviour and Information Technology · 2026
Organisations are rapidly embedding Genenerative AI (GenAI) into day-to-day work. Still, it remains unclear how GenAI role cues, partner (collaborator) vs driver (directive), shape employees’ perceived cognitive agency (PCA) and, in turn, employee-driven digital innovation (EDDI). We advance a multifactorial weighting model grounded in sense of agency, in which role cues reweight prospective vs retrospective signals, identifying PCA as the mechanism and role – skill fit as the lever. Two between-subjects studies with a 2 × 2 experimental design were conducted with Chinese professionals, Study 1: (GenAI role × locus of control); Study 2: (GenAI role × skill match). The partner role reliably elevates PCA above driver, especially for individuals with an internal locus of control, and PCA mediates between GenAI roles and EDDI. Partner-in-skill maximises co-creation, where higher PCA leads to higher EDDI, whereas driver-out-of-skill yields greater EDDI by supplying structure when expertise is low. Theoretically, this paper extends PCA as the operative mechanism and specifies when/for whom GenAI roles enable innovation. Practically, this paper provides actionable guidance on generative AI role framing (partner vs driver) to preserve authorship, pair it with training that strengthens internal-control cues, and prioritise in-skill task assignment to unlock agency-driven innovation.