The Client
The client is an online insurance marketplace in India where users compare price and features and purchase insurance policies based on their requirements. The client currently accounts for 50% of the online insurance market space in the country.
The Problem
About an 80% of users in the India purchase insurance after speaking to a representative on the phone. This makes a sales pitch an important factor in deciding whether a customer will purchase an insurance policy. The client wanted to build a system that could:
- Correlate the rating provided by the customer with the words, tone and speed of the conversation
- Automatically assess a sales agent’s performance over the phone
- Enhance a proprietary chatbot that would mimic the behaviour of an effective sales agent in recommending insurance policies
Key Challenges
The calls were in different languages that needed to be translated to English.
The client receives ~75,000 calls per day. The number spikes up during the travel season when users are looking for travel insurance and during taxation season when users buy health and term insurance to claim tax rebates.
There were multiple stakeholders – customer support, quality assurance and sales.
How ElementalAI Helped
The team designed and built a system that would transcribe all recorded phone calls into text and translate them into English using open source APIs. This data was enhanced to include insurance terms and jargons.
The system would assign ratings to the phone conversation by extracting the relevant keywords and phrases from the conversation. Other relevant information about the call such as speed, pitch were also extracted for analysis. This was a self-learning application and it improves itself with each new phone conversation.
Over a period of 12 months, the chat bot was handling 1,200 incoming leads per day and thereby reducing the operating costs of the sales operations by ~$1,000/day.
Technologies Used
Several technologies were also critical to the project. These included:
- Databases – SQL Server, MongoDB, Elastic Search
- Web Application – Python, Django, Javascript, D3.js
- AWS Services – Glacier, S3, Polly
- Data Analysis – Tableau, Sisense
- Other tools – Jira, BitBucket