Customer stories

Intelligence classification engine

Challenges

Our client, a leading French bank, runs a number of customer service centers to support retail and corporate banking customers on various topics. They receive, for each customer segment, over a million emails every year. The response time is compromised due to the large number of inbound emails. The customer service centers require smart solutions that increase the productivity of customer service agents and improve the customer satisfaction through timely and accurate responses.

Solution

The bank has chosen Otto to classify customer requests and automate the processing workflows. We have undertaken the following steps iteratively:

  • Data collection and analysis: we collected and analyzed the historical customer emails, based on which we defined the meaningful categories with the client.
  • Data labeling: we labeled the sample emails per category. Each label corresponds to a customer intention, actionable by the customer service agents. We largely reduced the amount of manually labeled data required thanks to our active learning technology.
  • Model training: Otto was trained on labeled datasets to recognize customer intentions in each email, at an acceptable level of confidence.
  • Business rules integration: our client defined business rules regarding each and the combination of customer intentions. Rules determined the folders emails should be classified in, the replies that should be suggested or automated, and other relevant automation actions to be taken. Otto integrated the business rules as defined.
  • Testing: we tested Otto’s accuracy in classifying emails and automating intended actions.

Results

In less than 3 months, we delivered highly satisfactory results. Otto was able to classify 90% of requests at 99.5% accuracy. In addition, we were able to completely automate replies and actions for 40% of customer requests.

Agents spend 28% less time on email classification and processing, and the average response time is reduced by 43%.