This study examines how Generative AI (GenAI) adoption influences exploratory and exploitative innovation and, in turn, organizational performance, through an AI-maturity lens that evolves from isolated pilots to fully integrated, operations-embedded capabilities. AI maturity is conceptualized as a dynamic, data- and model-centric capability encompassing governance, data infrastructure, human capital and MLOps, distinct from broader digital-transformation readiness.
Survey data from 302 advanced manufacturing firms in China were analyzed using Smart-PLS. Reliability and validity followed established guidelines, and multiple procedural and statistical remedies mitigated common method bias. The model integrates the TOE framework with an AI-maturity model to test stage-contingent relationships among GenAI adoption, innovation ambidexterity and performance. AI maturity was measured across four capability domains and classified into five descriptive stages using transparent, replicable rules.
As firms progress through maturity stages, GenAI adoption becomes more strategic and is linked to greater innovation ambidexterity, balancing efficiency gains with radical AI-enabled breakthroughs. Several classic adoption factors (e.g. compatibility, competitive pressure) weaken at higher maturity.
The cross-sectional, single-informant design limits causal inference; future longitudinal, multi-informant studies are encouraged.
A benchmarking checklist and predictive-maintenance case guide sequencing of GenAI investments and monitoring of key digital performance metrics.
The study advances existing theory by redefining AI maturity as a dynamic, operations-integrated capability linked to measurable digital performance outcomes. It extends the TOE/DOI perspective through a capability-progression lens consistent with dynamic-capability and ambidexterity theories and translates this conceptualization into actionable tools for managerial practice.
