In LUIS, regexes are features used by the machine learning (ML) algorithm. Think of a feature as a "hint" to the ML algorithm that helps it see the difference between words that match the regex and words that don't. Regex features are used by the ML learner for both intents and entities.
A regex can help recognize an entity -- for example, if the regex matches the terms in the entity, like the regex "kb\d+" for identifying knowledge base article IDs like kb8732827 or kb23737. In addition, you would also need to enter some utterances with the entity labeled – e.g. "look for article kb22716" with "kb22716" tagged as a "KnowledgeBaseArticle" entity.