Voicebox is already celebrated for its speech technology products and solutions for cars, homes, and mobile devices. The company’s collaborative, customer-oriented development style has created game-changing voice control interfaces and systems for a variety of partners, including Samsung, Toyota, and TomTom. Voicebox experts have also worked with mobile leaders to create conversational assistants for use in customer support, shopping, and navigation.
One example is the recent collaboration between Voicebox and a major U.S. retailer, when the two companies jointly developed a mobile conversational assistant. In this case, the client sought to reduce costs while increasing customer satisfaction. The assistant provides customers with a dynamic chat interface that automatically answers routine questions and routes customers to the proper departments for more in-depth assistance. By deflecting calls that would otherwise go to live support agents, the retailer reduced costs. By providing rapid, accurate answers and reducing call times through automated routing, the Voicebox conversational assistant also increased customer satisfaction.
Voicebox created a service to route calls, aid in finding stores, and guide customers to the appropriate departments. The conversational assistant was added to the retailer’s existing mobile application, and may be added to the retailer’s website in the future.
Customers use the retailer’s mobile app to access the conversational assistant, which uses advanced machine learning technology to determine what the customer needs (their “intent”). If the intent is for a simple answer to a routine question, the assistant provides the answer. For example, if the customer requests store hours, they are given that information. The Voicebox team, in concert with the retailer, developed a body of questions and appropriate responses that the assistant can handle directly.
More complex intents trigger the call routing service, which hands the customer off to a live support agent in the correct department; thus, bypassing the need for menu-trees or two stages of live assistance.
The Voicebox method for developing the conversational assistant uses advanced statistical methods to learn how to answer customer questions, by mining the retailer’s customer support logs and crowdsourcing additional training data. As most of Voicebox’s competitors cannot yet deliver similar solutions, Voicebox’s large-scale use of statistical methods is part of why the Voicebox conversational assistant stands out.
The initial project was completed in only six weeks just prior to Black Friday 2016, thanks to Voicebox’s agile development team. The ability of those within the team to bootstrap an initial system from the retailer’s support log data — and augment it with our proprietary crowdsourcing system for data generation — allowed the team to create a robust solution. Their close collaboration with the retailer and openness to revise plans when necessary ensured success.
Since its launch, the conversational assistant has successfully answered a high percentage of customer queries. Further, the assistant continues to improve, thanks to the team using customer interactions that end unsuccessfully to discover new intents the assistant should address.
The conversational assistant’s future is bright: Work continues on an expanded range of intents and responses, as well as improvements in the underlying system’s natural language understanding technology, response generation, and ability to incorporate customer-specific information (e.g., purchase history or order shipping status) into its understanding of what customers need. As businesses target the dual goals of decreasing costs and increasing revenue, the demand for conversational assistants and the business intelligence they can provide will only grow.