Industry: HealthcareRole: Conversation Designer
Skills: Conversation mapping, bot optimization, design workshopping, NLU development
A quick note: almost all of the projects I work on are under an NDA and are authenticated experiences, so I can’t share particulars here when it comes to names and detailed designs.
The Project
In an effort to move support volume away from the IVR system and help members navigate a large portal, the client wanted to create an FAQ bot to help members solve their most common issues.
Phase 1 : Basic FAQ Bot
In the first phase of this project, we focused on strategy and use case optimization. I included all of the IVR language data in my analysis and was able to help the client figure out which topics the bot could handle and help them structure the menus and flows.
When we launched the bot the first time, we did not launch with NLU as we did not have real-world messaging data yet. However, I was able to look at utterances and user behaviour in the bot shortly after the launch and present ideas for optimizations.
Phase 2 : Fully Optimized
In the second phase, I led a number of workshops with the client to review the data and discuss how we could make the bot more action-driven. We decided to:
- Add new flows that were more granular (i.e., moving away from general flows like “Claims” to more specific issues like “Claim denied”).
- Create a Knowledge Base to answer even more FAQ
- Use real user utterances to train the NLU model
- Allowed for the member to type their question or issue and get an accurate answer (as opposed to a guided flow with only buttons)
The combination of these optimizations increased the intent match rate by 9% and the containment rate by 3%. The client is now looking to offer IVR deflection to other rich channels and I’m currently working on the design for this next phase.
