Understand Call Center Conversations Without Reading Transcripts.
A major requirement for securing the NIKE contract was the ability to take call center data transcripts and distill it down to what the high level message, sentiment, and result of the call was without reading individual utterances. We enabled this by creating a robust transcript analysis feature which included front end UI to filter, analyze, and find your individual call center data points and back end updates to the Natural Language Processing engine to parse who said what.
Now You Know What They Both Said.
We created a feature in our underlying Machine Learning Models which enabled users to segment a transcript into employee utterances and customer utterances. This was surfaced on the front end by enabling users to see high level metrics on the overall call as well as the ability to filter the metrics based on if the customers statements and the employees states. All of this could then be rolled into high level charts and visualizations across thousands of calls.
Did Someone Order Instant Insights?
In addition to this extensive transcript analysis feature we also created a AI Summary feature which took the overall transcript and used AI to generate a summary of the interactions so that users could quickly see the main take aways from a one hour customer service call.
Devil in the Details.
Given how massive this undertaking was I invested time to ensure proper documentation was created to communicate with developers and upper management the function and purpose of every component of the project page. Not only did this accelerate development time but it also solved confusion about functionality and intention between design and engineering.
Learnings moving forward.
Currently we are in the process of enhancing the Smart Summaries feature by extending it to additional data source types
