In the reality of the highly uncertain external environment and the widespread dependence of Chinese enterprises on a few large customers, building a diversified customer network is an important way to create a flexible, controllable, and highly resilient industrial and supply chains. This paper regards the construction of AI Innovation Application Pilot Zone as a quasi-natural experiment of deep integration between AI and the real economy. Based on the multi-time point PSM-DID, it examines the impact and mechanism of AI applications on the diversification of enterprise customer configuration. The results show that AI applications effectively promote the diversification of enterprise customer configuration. Mechanism testing reveals that enhancing the internal product quality advantage of enterprises and weakening the external information constraints of both parties in transactions are the key factors for AI applications to promote the diversification of enterprise customer configuration. Heterogeneity analysis shows that this optimization effect is more prominent in enterprises in industries with fierce competition, small- and medium-sized enterprises, and enterprises with lower production efficiency, as well as regions with a higher level of information fragmentation. It is also found that the main manifestation of the optimization effect is to explore new customers or increase sales to existing small- and medium-sized customers. This paper deepens the theoretical understanding of the logical relationship between AI applications and micro subject customer configuration, providing important experience and policy references for promoting the deep integration of AI and the real economy, enhancing the coordination ability of enterprise customers, and improving the security and resilience of industrial and supply chains.