French Telecom improved customer experience by converting Chatbot to Conversational AI

June 28, 2022 | Published by

Business Problem

Our client is a leading French Telecom company who had big plans to drive digital transformation. Its aim was to provide a seamless, personalized customer experience to increase customer satisfaction and loyalty. Our client had a chatbot on their Web Application to address end customer queries but had a low confidence score of 20-30% resulting in lower customer experience. The chatbot needed to identify the conversation and respond correctly to the customers with improved confidence score. To accomplish this, the telecom firm required a competent and experienced rollout partner to implement the system.

PureSoftware Solution

PureSoftware partnered with the client to analyze the root cause of the problems and decided to implement a machine learning based chat-bot to enhance customer experience to guide customers throughout the process. Our team with extensive experience performed the following:

  • Changed the machine learning model from Sklearn to deep learning model embedding TensorFlow.
  • Migrated the ML model from pre-trained to Deep learning embedding training in which model is trained based upon the corpus of own dataset.
  • Apply Keras policy of Deep Learning for response prediction.
  • Fine tune the TensorFlow embedding model according to the dataset so that it can give high confidence score.


  • After a successful rollout, our client experienced enhanced customer experience as the chatbot had an improved confidence score of 70 to 80%.
  • Chatbot was able to identify the conversation and respond correctly to the customers.
  • The Project went live within 6 Weeks and provided with improved UI for a better customer experience.