As companies must support an expanding base of products and customers, the need for automated helpdesk systems is increasing. Development of an automated helpdesk system requires automating a four step troubleshooting process involving context establishment, feature extraction, solution presentation, and verification. This thesis examines and evaluates current approaches utilized in modern automated helpdesk systems and presents the design of a new automated helpdesk application called the Conversational Case-Based Helpdesk System (CCBHS). CCBHS employs a conversational case-based reasoning approach to infer the most likely cause of a troubleshooting problem. The CCBHS is implemented in Java 2.0 and utilizes a web-based interface created dynamically using Java Server Page technology. In addition, multiple concurrent user sessions are supported. We improve on existing conversational case-based reasoning approaches by adding support for "I don't know" answers, multiple question ranking and case ranking strategies, and the ability to index cases from multiple case-base files (of any format) in a single index. We also introduce a method for generating inference rules automatically from a set of cases and show that this method effectively increases the questioning efficiency and usability of the system. We introduce four case ranking strategies and two new question sorting algorithms and perform and analyze experiments that measure the effectiveness of these heuristics relative to other heuristics that were proposed in the literature.