The copilot paradox: configurational pathways to human-AI symbiosis in knowledge work

Purpose This study aims to examine the “Copilot Paradox” in safety-critical operations, offering insights into how the chronic tension between the algorithm’s increased efficiency and operational vulnerability is subdued through socio-cognitive configurations across time. Design/methodology/approach With the use of a two-wave longitudinal fuzzy-set Qualitative Comparative Analysis (fsQCA), the paper explores the changing dynamics of human–artificial intelligence (AI) symbiosis in the oil and gas industry. Findings Analysis of necessity shows that there is no single condition required in all cases. Sufficiency analysis reveals three distinct, equifinal pathways: buffering, synergy and substitution. At early adoption, institutional re-skilling coupled with psychological safety acts as a buffering mechanism against job insecurity induced by AI. As integration progresses, synergy (fuelled by empowering leadership) and substitution (where team resources fill the gaps in leadership) come to dominate. Practical implications Team psychological safety and capability building are structural safeguards for managers in high-reliability sectors, not optional cultural flourishes. These lessons point to pathways for realising AI copilots while preserving human agency and operational resiliency. Originality/value This paper contributes to paradox theory by developing a socio-cognitive infrastructure that highlights its dynamic, temporal configuration and causal asymmetries. Methodologically, it showcases the new promise of longitudinal fsQCA to model dynamics in algorithms’ infusion by capturing their embeddedness through time.