How does the Message Cleaning work?

Playvox Customer AI is designed to analyze your user messages to reflect their voice so we remove any agent or bot responses from the analysis automatically. However, when users respond to support conversations via email, the original agent message can sometimes get embedded in their reply. As a result, your topics might trigger unwanted matches and affect the accuracy of sentiment.

In the example below, you see a user message (in green) that has an email thread attached at the end (in red). This can sometimes happen due to the way different support tools process email replies. 

The attached thread contains text from the support agent which is not what we want since we are analyzing customer messages. 

Customer AI can automatically detect that there is an email thread attached and remove it, leaving just the user message for analysis. This helps dramatically improve the accuracy of sentiment and topic analysis.

How to enable Message Cleaning?

Message Cleaning is enabled on all workspaces by default. However, if it was disabled you can easily enable it by following these steps:

  • Navigate to Settings > Workspace Settings;
  • Under Analysis Features, set the Message Cleaning toggle to ON (green).

Once enabled, any new incoming messages will pass through our Message Cleaning algorithm. Please note that it will not automatically clean any messages that have been already imported.

How does it work?

  • When Message Cleaning is enabled, Customer AI will attempt to detect the part of the message that does not contain useful user content such as remnants of email threads, signatures or disclosures;
  • The unwanted message part is automatically removed and only the user part of the message will be kept.

How to add new cases?

We are continuously adding new cases to the Message Cleaning but due to the ever-evolving nature of free text conversations, inevitably new cases will occur.

If you spot a message that wasn't automatically cleaned, please send us the message example and we will add a new case to handle those types of messages.