|M.Sc Student||Ratner Tatiana|
|Subject||Applications of Artificial Intelligence to the Management|
of Call Center Waiting Lines
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Reuven Karni|
|Full Thesis text - in Hebrew|
One of the main channels between a company and its clients is a call center; customers buy products and services, ask questions, and order field service. Effective management requires the organization to focus on three central operational factors: characteristics of the calling clients and incoming calls, assignation of skilled agents, and implementation of its policy regarding relationships with customers and agents. We base ourselves on the observation that many factors influencing call center activity are under the control of the organization and should be considered in order to empower the organization: decisions concerning characterization of clients and services given by the call center; decisions regarding client and call prioritization; determination of client segmentation to guarantee advantages given to preferred clients; differentiation of wait time thresholds for each segment; assignment of agents and their skills for the effective processing of calls and clients; correlations between agent capabilities of the agents and service levels; and decisions regarding the tradeoff between service quality and speed.
No mathematical model that considers all these factors exists at this time. Our research therefore implements a computer based model for allocating and routing incoming calls. The model comprises five algorithms: an "outer loop" for high-level management of the routing process; two "inner loops" for selection of an agent to process a specific call in "regular" and "backup" modes; an expert system for the dynamic prioritization of calls in the queue; and an expert system for computing the expected service level when matching agents to calls.
The significance of dissimilar clients, calls and agents was studied to determine the relationship between organizational decisions and call center performance. It was found that the recommended policy for maximizing performance center should be: training agents to be competent in at least two service types and languages; formulating business rules that prioritize calls from important clients using preference weighting and thresholds on maximum waiting times; and avoiding the granting of "favored client" status to too many customers.
The model contributes to the science of call center management and call routing: operational factors explicated and consolidated in business rules which allow easy assignment and modification of: the differentiation and preference grading of client and service types; thresholds that define a maximum waiting time for each client type; partitioning of skill levels to distinguish between "regular", "backup" and "unqualified" agents; and multiple characterizations of calls and agents.