In order to drive customer experience, consumer brands need to know whether agent-consumer interactions were successful. In human-human interactions we know how to deduce resolution and satisfaction from post-conversation surveys, or, preferably, from automatic methods such as tracking of the dynamic sentiment of the conversation. I will show that these approaches are not successful in human-bot interactions, and will share how we understand the problem of automatically understanding the outcome of a bot-consumer interaction in LivePerson.
Moving Towards Data-Driven Assessment of Bot Success
Irit Sella is a Data Scientist at LivePerson developing algorithms to maximize the quality and efficiency of consumer-brand engagements, both human-to-human and human-to-bot. She currently leads LivePerson’s research into agent performance and utilization. Irit has a PhD in Neuroscience from the Weizmann Institute of Science, an MSc in Neuroscience from the Technion, and a double major BSc in Software Engineering and Cognitive Science from the Hebrew University.