Deploying automated ticket router across the enterprise

With the recent advances in machine learning, use of natural language processing (NLP) technology to support various business processes has been increasing. This paper discusses NLP route more than one million live client tickets annually appropriate service personnel 67 missions across IBM. Each mission supports a product family with multiple teams, each requiring different skills for engineers. We discuss three important aspects such large-scale deployment: (i) The centralized team common learning infrastructure and practices entire enterprise. (ii) quality deployment from perspective mission, namely, IBM's z/OS family. (iii) Careful monitoring deployed models detect drifts routing behavior. Despite vast differences technical contents missions, it is possible define metrics enterprise, without dedicated mission. In addition, we provide examples policies demonstrate utility approach outcome.