Abstract: A customer communication system is configured to automatically communicate with human customers in a conversation including an ordered sequence of messages. Messages from the system to the customers are selected based on a knowledge graph. The knowledge graph including multiple levels of clusters of customer messages, each of the clusters being associated with a responsive supporting message. The customer communication system is optionally configured to identify inadequacies in the knowledge graph and obtain corrections from a human expert. The system may be adapted to communicate with parties other than customers.

Abstract: In a crowd sourcing approach, responses to customer service inquiries are provided by routing a subset of the inquiries to an independent group of experts. The customer service inquiries are optionally routed to specific experts based on matches between identified subject matter of the inquiries and expertise of the experts. Embodiments include methods of classifying customer service inquiries, training a machine learning system, and/or processing customer service inquiries. Customer service inquiries and answers from a first enterprise and/or industry are optionally used to train an inquiry classifier for a second enterprise and/or industry. The classifier being configured to predict if a new customer service inquiry will require access to confidential information for a human to generate an answer that resolves the inquiry.