From large language model adoption to operational performance: a process–decision conversion perspective

Purpose This study examines how large language model (LLM) adoption is converted into operational performance through process intelligence and decision agility. Design/methodology/approach Drawing on dynamic capability theory and information processing theory, the study tests a sequential mediation model using three-wave matched survey data from 462 managers in Chinese manufacturing firms. Findings LLM adoption is positively associated with process intelligence, process intelligence with decision agility, and decision agility with operational performance. LLM adoption also has a significant total effect on operational performance; however, its residual direct effect becomes non-significant after the mediators are introduced. This pattern is consistent with an indirect-only capability-conversion account. Practical implications Managers should treat LLM implementation as a capability-building process that connects language-based AI resources to process diagnosis, workflow adjustment and operational decision routines. Originality/value The study contributes to business process management research by explaining how LLM adoption becomes embedded in process intelligence and decision agility before producing operational value.