Project Naples
Cellular providers in the US and Canada
Conversational self-service for common caller intents
Demonstrate the effectiveness of open dialog and other conversational strategies to streamline interactions
Showcase seamless IVR-to-Digital handoff for use cases that benefit from GUI
Use case 1: confirming the balance and making a payment
Advanced conversational technology allows for flexible flows, where users can answer a question in a variety of ways.
For example, a payment amount can be specified:
as part of a multi-slot intent utterance,
as a direct answer to a follow-up question,
as a dollars and cents amount,
as an opt-into a proactive offer to pay the account balance,
as a proportion of the account balance.
Use case 2: upgrading to a new device
Handing over of the IVR session to the VA, and subsequently to an agent, in-session, allows for sharing of important context for a seamless experience:
User profile
Authentication tokens
Intent
Additional entities captured to refine the intent
Project Hansa
National wireline provider
Technical support line for a wide variety of technical issues with wireline services - home internet, TV and phone
Mature application with NLU only at the main menu; directed dialog and touchtone elsewhere
Dynamic troubleshooting flows, configurable in real time based on lab tests and call center data
The troubleshooting call flow for each user-reported issue is assembled dynamically. Based on the information reported by the caller, the diagnostics, and the outcome of previous steps, the next step is dynamically chosen among several dozen possible actions.
This can be modified in real time based on learnings from the hardware testing lab and call center data.
Each troubleshooting step is designed as a discrete module, with no predictable context as to what comes before or after.
Design challenge: making the exchange sound like a cohesive and intelligent conversation, when not working with a stable sequence of steps for each use case.
Examples of troubleshooting steps:
Polling the user’s devices and connection to the network, and remotely changing device settings.
Remotely sending a variety of signals to connected devices to power-cycle, refresh connections, etc.
Forcing select devices to emit sounds or light to help finding or identifying them.
Instructing the user to perform specific operations, or to try test task to evaluate if changes took effect.