I interpret the question so, that the OP wants to know in which direction a literature search has to go. Because the topic he is asking for is very well researched by academics before. At first it is important to know, that the term CRM (Customer Relationship Management) is used in the business literature for describing marketing strategies and also for software which can be used in the intranet. This is a dead end, because It is very rare, that current management literature is already aware of artificial intelligence. The better starting point for a literature survey is the term "customer sentiment nlp", which is discussed in more mathematical motivated papers. The addition NLP (natural language processing) gives a rich amount of papers back, which are explaining how to clustering and annotate opinions of customers with the aim to build a model of it. Such models were used in later steps together with neural networks in so called opinion based databases.
Another potential direction for a literature search starts with the term "customer simulation". Here is the idea to simulate the behavior of customers in a huge computer game. The customers are agent, can make buying decisions at random and the results are feed into a model. Such a simulation can be run against an existing CRM system with the aim to fill it with pseudo-random-data.