Solution 

First, we reviewed the prior DTMF solution to see who was using it and how. We found that it was mainly power users who were using it to do simple tasks, like access their checking account balance. But the average user was not using it, likely because they didn’t want to memorize a handful of simple DTMF interactions to get to their task. 

So, we designed an interactive voice response (IVR) application with a natural language interface. This allowed members to use their voice to access their member accounts. No more listening to menu prompts and pressing a series of buttons. Now, they could simply say “checking balance” and be done with it. And to make it even better, the new IVR experience was more secure than the existing solution. 

We also recommended moving away from the client’s existing on-premise model to a secure, flexible, cloud-based solution using the latest Amazon Web Services Lex technology.  

But we didn’t stop there. We ran an optimization exercise where we had a speech science expert listen to calls and review interactions to understand caller behavior and suggest modifications to fine-tune the new IVR experience and improve containment.  

This included adding words or phrases that callers were using, but which had not been originally programmed, and designing changes based on caller behavior, such as error handling around collecting authentication tokens that verify a user’s identity within the application.