The partner is one of the leading insurance providers in the US. Their repertoire includes everything from health and life insurance to travel and vehicle insurance policies for both corporates and individuals. Selling over 30 million policies every year across 35 states, they are scaling fast with a double digital YoY growth.
The mission was to build a scalable solution that could help them streamline their support processes and deliver a cohesive customer experience across channels. Relying on emails and IVR-based phone network was no longer effective to address their ever-growing queue of customer queries. Delayed response times, ill-prioritized inbox and lack of customized responses / insurance quotes were leading to lost business for the partner and a huge dissatisfaction among their customers.
The first thought was to adopt a SaaS solution that could help them overcome this challenge. However, plans of SaaS companies in the space are often structured based on per agent pricing. This would have been an extremely expensive deal for the partner, considering they currently have over 100+ employees in their customer support team, with a fast-growing headcount as they scale.
They needed a technology partner that could build them a customized multi-channel chatbot to suit their business needs.
How Velotio Helped?
A team of six was deployed on the project. Velotio built a cost-effective, highly-personalized AI-powered chatbot solution which was deployed on Alexa, Facebook, and the partner’s website.The bot was launched in 11 weeks and helped the partner offer a streamlined customer experience.
This involved building a Machine Learning (ML) and Natural Language Processing (NLP) system that could learn from customers’ historic data and read customers’ queries / tickets to offer extremely customized quotes and replies. This reduced query response time, improved sales and customer satisfaction for the partner.
RASA NLU was used to train the bot to handle over 110 intents along with hundreds of entities. This enabled the bot to interact with customers in a more humane manner.
Botkit provided the provisioning to develop the bot once and deploy it on the two initially chosen platforms (Facebook and the partner’s website).
The historic data from past cases was required to help the bot learn different variables that are taken into consideration while offering a personalized insurance quote for a customer. The team needed to retrieve eight years of data from email, partner’s CRM and core IVR system, which spiralled into millions of records.
As the data was retrieved from three different platforms, there was a lot of overlap and variations in variable names, etc. The bot, however cannot learn from unstructured data. Velotio’s team needed a solution to convert this unstructured data into a structured one.
The dictionary needed for this business problem was very niche. The team needed to identify all the names of different insurance policies from the transaction free text. Probability of finding such a reliable dictionary like this was very low. Hence, the team needed to create a dictionary for scoring the entire dataset.
The team used text mining techniques with R and SAS together to build an unstructured data model. R was used to create dictionary on smaller datasets and SAS was used to score this dictionary on the entire dataset.
The bot created by Velotio team provided a comprehensive solution to ensure great customer experience across channels by:
1) Automating mundane queries that didn’t require human intervention, like registering a claim, checking policy status, renewing insurance, and FAQs. Such queries made almost 30% of the queries partner gets on a daily basis. Now their support executives spend time on queries that actually require human intelligence.
2) Utilizing Natural Language Processing (NLP) to help the bot read customers’ queries / tickets to tailor replies in their language. The technology also came in handy in building features like skill-based routing, automatic prioritization scoring, canned responses and many others, which reduced overall query resolution time by 40% and improved customer satisfaction.
3) Offering extremely accurate insurance quotes based on the information provided by the customers after answering a couple of quick questions on the chat. Since customers could now get personalised insurance quotes right away without any manual intervention, it speed up the sales process by 2x.