Upping the game in Online Insurance

Background Our customer is part of a new breed of fast-growing insurance companies that promise to change the way people feel about insurance. The group currently covers 8 Asian markets and is fast approaching 10m clients.

They were one of the first on-line insurance companies to introduce a chatbot in Singapore. However, despite significant investment, the bot never really gained traction. By the time INTNT engaged the customer, their Customer Satisfaction ratings on their chatbot were stuck at ~1 (in a scale of 1-5).

We were engaged in late-2020 to train, turnaround and enhance their chatbot.

400%
Increase in new sales leads
50%
Increase in Users
60%
Reduction in Agent Escalations
INTNT uses patented algorithms to auto-detect false positives and suggest alternative answers. False Positives typically make up to 67% of bot errors. And yet they are often overlooked because it takes manual effort to detect them. INTNT provides an easy and simple process to train bots against False Positives.

What we did to transform their chatbot Using our proprietary and patent pending INTNT NLP/NLU engine, we turned around the knowledge base, adding 30% more intents, pruning 20% of conflicting training phrases and establishing a weekly bot training cadence with their chatbot.

Our INTNT training engine is fully automated, which allowed us to perform a 100% review of all customer utterances on a regular basis (unlike manual reviews which usually make do with a small statistical sample on a periodic basis). The process takes only a matter of hours to identify and review false positives and missed intents, as opposed to an estimated 27 man-days to review a medium sized chatbot manually. This allowed the bot to be trained regularly - resulting in increases in the accuracy of their natural language chatbot from an average of 40% (typical industry standard) to 90+% in all our use cases.

Impact

The impact on their chatbot was immediate, with the biggest impacts being on customer satisfaction and sales lead generation.

With our INTNT engine, their chatbot accuracy increased by 90% within the first 3 weeks of training.

Simply by being able to understand and address customer questions and concerns about product features (and other enquiries) accurately during the customer product search and in-purchase journeys, the chatbot ensured that customers remained in-journey in their website, and that they were able to complete their purchase journeys seamlessly.

The results

12 months into the project:

- the number of unique users were up by 50%

- customer satisfaction in the chatbot went from 1 to 4+

- agent escalations were down by 60%

- sales leads were up by 400%

- renewals were up by 900%.