Email classifier

Context

A financial services company receives over 1000 emails per day to their customer service desk

Problem

The customer service desk employees had no way of identifying the most urgent emails and which ones they should open first. The long response times frequently led to customer dissatisfaction and other escalations.

All emails arrived into a central inbox and an email could be forwarded up to seven times before it reached the right department or person. This wasted a lot of time and caused frustration among employees.

Solution

We built an AI email classifier which can recognise the topic of an incoming email. All known topics get an internal urgency score. With this score the emails can be ranked for urgency, allowing the service desk employee to focus on the most pressing matters first.

The topic is also linked to a specific department or person, so the email can automatically be forwarded to the right destination.

Results

  • The classifier managed to recognise the correct topic with an accuracy of 87%
  • The average response time on urgent emails dropped from 7.5 hours to 3 hours
  • The number of times an average email is forwarded dropped from 7 to 2