AI investment (In)congruence and innovation resilience in knowledge-intensive firms: unpacking the boundary role of knowledge structure

Purpose Drawing on the knowledge-based view (KBV), this study aims to deconstruct firm-level artificial intelligence (AI) investment into a supply-demand structure to examine how the (in)congruence between supply-side AI investment (SAI) and demand-side AI investment (DAI) affects innovation resilience in knowledge-intensive firms, along with the moderating role of knowledge structure. Design/methodology/approach Based on a sample of Chinese A-share listed knowledge-intensive firms from 2016–2023, this study uses fine-tuned BERT-based text classification to measure SAI and DAI, and tests hypotheses via polynomial regression and response surface analysis. Findings AI investment congruence enhances innovation resilience more effectively than incongruence, with structural configuration being more critical than investment scale. For incongruence, high SAI–low DAI is more effective than low SAI–high DAI. When knowledge depth is high, high SAI–low DAI increases innovation resilience more than low SAI–high DAI, whereas when knowledge breadth is high, low SAI–high DAI becomes more effective. Originality/value This study deconstructs AI investment into supply-demand dual dimensions, offers a structural perspective on how AI investment (in)congruence affects innovation resilience, clarifies the heterogeneous moderating effects of knowledge depth and breadth and extends the application boundary of the KBV in the AI era.