Email routing in customer service

For each service case, an administrator can configure keywords used for text analysis. The keywords identify the topic in an inbound email and associate the email with a suggested reply or a service case. For example, a disputed credit charge is associated with the Dispute Transaction case; a change of address is associated with the Account Address Change case. This association is made by comparing the email content with keywords that you configure for each case type.

You configure keywords in the email channel.

For this training, in the Pega Customer Service application, we have created an email channel called MySupport. The email bot associated with this channel is configured with service cases. The email bot's text analyzer is seeded with keywords that can help the bot identify the email topic and associate it with a service case.

For topic identification, you can start with keywords, but it is a best practice to use a topic model because the model can scale with email volume.

Pega Email Bot uses NLP to analyze and learn from patterns of conversation between customer service representatives (CSRs) and customers. Pega Email Bot learns from repeated behaviors: the more emails it processes, the more accurate it becomes in helping a CSR respond to customers.

For each topic that you add to the channel, you must train Pega Email Bot to interpret emails. By providing examples to Pega Email Bot and manually classifying them with a topic, you train the NLP model to understand these topics and classify similar emails.

A data scientist can set up and train Pega Email Bot by working directly with NLP models in Prediction Studio. 

In Pega Customer Service™, both the CSR and the customer service administrator (CS admin) contribute to training the email bot.

Routing rules use the result of text analysis to route mail to a queue or operator. You define the set of conditions that must be met for Pega Email Bot to invoke the routing action. Conditions can include the identified topic (for example, address change), whether an entity exists (for example, the email content contains the new address), and the sentiment of the email (positive, negative, or neutral).

A routing rule can have multiple conditions that must be true before it takes a routing action. If no conditions are met, then the otherwise action is taken, such as routing all emails to the Inbound Correspondence queue.

In the routing rule, you can also set a confidence threshold for topic detection. For example, given the text analysis of the email content, there is an 80% probability that the email is an address change request. The text analytics model becomes more accurate as it processes additional emails. A data scientist configures the model in Prediction Studio.

​