Moral dilemmas are the most difficult and prevalent type of ethical challenges in organizations. Whereas prior research indicates that moral insight is key to resolving moral dilemmas, there is little knowledge about how to acquire and foster moral insight. This research investigates the potential of using large language models (LLM) to pave the way for moral insight in resolving moral dilemmas. Specifically, we theorized how LLM can emerge with moral insight by combining their creativity and moral alignment, and benefit human users in decision-making for moral dilemmas. Through two real decision-making experiments conducted across different countries and involving LLM products, we found that: (1) LLM such as ChatGPT generate more moral insight solutions than humans in moral dilemmas; (2) individuals who use LLM to resolve moral dilemmas generate more moral insight solutions than those who do not in moral dilemmas; and (3) individuals who have previously used LLM to resolve moral dilemmas generate more moral insight solutions than those who have not, even when they do not use LLM. Overall, our research showed that by using LLM, individuals can increase the likelihood of discovering moral insight solutions by 44.21% to 80.21%. This research provides significant theoretical contributions to the understanding of the relationship between morality and creativity, and it offers a novel and useful approach for resolving moral dilemmas.