Tata Power: Digital Innovation to ensure customer satisfaction during COVID-19 lockdown

Tata Power Company is India’s largest integrated power company and is a pioneer credited with steering the energy sector on technology, process and platform. The Covid-19 lockdown posed a unique challenge for them with respect to customer satisfaction.

During the lockdown customer billing was being done on estimate basis and when site activities were restored, billing resumed on actual meter readings with differential consumption which was really high for lockdown period and this resulted in an unprecedented customer complaint influx on email. The volume of complaints grew 8X and made it nearly impossible to manage with the existing systems.

Tata Power responded to this crisis through a Digital First approach where they came up with an NLP based Machine learning model that aimed to automate processing of email complaints in agent inboxes and create transactions in the SAP CRM backend.

The new model was designed to read the email, identify the transaction type and understand the customer’s emotion. The objective was to reduce the amount of activity for the customer care supervisor. This technology make uses of data accumulated after 5 years of using SAP CRM as a customer tool for email interaction. The NLP based model with machine learning Algorithm in conjunction with library tensor flow has effectively brought down the response time of replying to a customer complaint email from 15 mins to 7 mins.

Prior to implementing the AI/ML model for customer care, a typical supervisor would be required read each mail and identify the sentiment and knowledge area of each email complaint and forward it to suitable agents for responding and categorizing the email complaints in 42 odd type of category in SAP CRM

While the other many other succumbed to the unprecedented number of email complaints and stopped replying altogether to customer email queries, Tata Power continued to maintain a high level of customer satisfaction during the lockdown.

This new model ensures prediction with 85% of accuracy for complaint category identificaton and it predicts the customer sentiment with more than 98% accuracy.

“With the adoption of NLP model we are now able to predict the sentiment much greater accuracy. This enables the email supervisor swiftly respond to irate consumers which create experience and helps to retain customers better than before. This AL / ML model has really helped in identifying and bucketing different queries basis on their type and arranging a response basis on that with enclosing consumers' bill copy with that . This model was really helpful teams in managing this exceptional volume effectively to the E-care teams” - Roopesh Srivastava Group Head - Retail Customer Management, Tata Power Company