Expert
Improve your virtual agents and the
customer experience with generative AI
Generative AI creates text and imagery based on user-provided prompts. These prompts are often provided by typing in a chat box, but they can also be given by voice or uploading an image.
While most people use it today to write content or compile artwork, it can be used for far more. Instead of creating media, Mosaicx Expert creates value for businesses and customers, using generative AI in four unique ways:
Answer Questions:
Not all intents can be predicted. Generative AI can pull information from an index, such as company websites, employee resources, internal spreadsheets, etc. to generate answers to unexpected questions. The search is restricted to approved sources, so answers won’t include any surprises.

Understand Customer Intent:
Customers contact a business for many reasons. People phrase each of these “intents” up to 100 different ways. So If a virtual agent responds to 50 intents, engineers must program up to 5,000 variant phrases. Generative AI creates and tests phrases automatically, reducing workload and error.

Identify Variant Entities:
“Entities” are data points like dates, times, monetary values, and locations. Virtual agents expect specific responses, but people still provide variations. For example, “Los Angeles,” “Los Angeles International,” “LA,” and “LAX” all describe the same airport. Generative AI lists variants that engineers may miss.

Automate Testing:
Engineers familiar with a system often struggle to create hard tests. Generative AI can initiate conversations with a virtual agent, pretending to be a customer who is angry, anxious, verbose, blunt, and more as prompted. These test conversations are easy to read and discover where virtual agents succeed and fail.

Carry Conversations:
IVAs (intelligent virtual agents) already carry out natural-sounding conversations with customers, but every variant intent and entity must be programmed before that’s possible. Through text, customers can comfortably communicate with generative AI without fixed dialogue and with less backend training.
