One of the world’s largest global financial institutions was looking for ways to improve its call center performance to both enhance customer experiences and reduce operational costs. It particularly wanted to gain a better understanding of all inbound correspondents, so that it could prioritize calls based upon related subject and also sentiment of the customer, automating calls where possible and routing calls to the most specific agent groups where automation was not possible.
We used an enormous volume of historical written and verbal inbound communications to train the Intelygenz AI to learn a classification priority order based on the bank’s key specifications. Combing AI solutions of voice to text, NLP and sentiment analysis meant all calls could be accurately diverted to either an automated response for simple inquiries (such as bank balance or branch open times) or to the correctly trained call center agent for more complicated issues.
The solution was successfully deployed into production having followed the Intelygenz AI deployment process. That is starting with a Discovery phase to understand the business challenge, establish if correlation in the historic data existed and experimenting on which model combinations could be trained to provide the most beneficial results.
After establishing positive results through the Discovery process, a POC solution was built to test the results in a live environment. This allowed us to determine the best way to integrate the technical results into the business processes of the customer to drive the business value.
Finally the POC was transitioned into production through an integration exercise using the learning from the POC to maximise results.
By basing the AI’s learning on specific requirements, the bank is now able to take full, automatic control over which issues demand the time and attention of its call assistants. This allows it to better serve those with more complicated queries and saves costs by automating a large volume of incoming calls. The customer experience has also improved, as queries are answered much faster, thanks to a reduction in hold times and the call being handled by agents with specific knowledge on the subject.