This article proposes an intelligent customer service robot system based on machine learning to address issues such as large user inquiries, insufficient customer service personnel, low efficiency caused by repetitive work, lack of service during non working hours, and poor user experience. The system uses natural language processing technology to analyze user problems and applies long short-term memory (LSTM) networks to identify user intentions and extract key information. The robot system proposed in this article also integrates a knowledge graph to more accurately retrieve relevant information and provide customized answers. Experimental data shows that compared to traditional systems, intelligent systems exhibit higher accuracy in all situations, with an average accuracy of over 85% in most cases, and even approaching or exceeding 90% in a few instances; in terms of response time, the fastest response time of intelligent systems is 39 milliseconds, while the fastest response time of traditional systems is 229 milliseconds, further emphasizing the speed advantage of intelligent systems.