Conversational AI

This Manning title walks developers through the full life cycle of a customer-facing AI assistant, from the first design decisions to the upkeep that follows launch. Readers learn how to choose an assistant type and the channel it will run on, how to give its dialogue a deliberate tone and level of detail, and how to build a classifier from scratch, starting with where its training data comes from. The book also covers building question-and-direct-response assistants, mapping and tuning process flows for web and voice, measuring how accurate an assistant is, and planning improvements over time.

The chapters move in five parts. The first introduces conversational AI and guides readers through building a basic assistant. The second focuses on designing processes, dialogue, and a complete assistant. The third covers training, accuracy measurement, and testing dialogue flows. The fourth addresses deployment, management, and ongoing improvement. The last part offers optional material on building a custom classifier and on extra training for voice assistants. Throughout, the author combines design theory with hands-on technique and pairs the material with a practical, lighthearted style. Examples are written with Watson Assistant and Python.

The book is intended for software developers who want to automate common customer questions, sort and route issues, and reduce support costs. The author, Andrew R. Freed, is a Master Inventor and Senior Technical Staff Member at IBM, and has worked on AI solutions since 2012.