2020 threw the world a curveball like no other. Yet, the digital transition that the world witnessed was nothing less than impressive. The manner in which businesses transitioned from a physical to digital setup is one for the books. Particularly, the service industry had to develop mechanisms to maintain touchpoints with customers and maintain the trust between the existing clientele. With customer relationships functioning as a key aspect to businesses in the service sector, companies cannot afford to leave any stone unturned in reaching out to them, serving them, resolving their issues, listening to them, and taking their feedback.

“The moment you can reduce the distance between a company and the customer to zero, that is digital. So digital is not about technology from our point of view, digital is about the customer, digital is about reducing the distance of communication and optimizing the ease of engagement. We are achieving it with the help of AI and bot capabilities with big data analytics,” says Manik Nangia, Chief Digital Officer & Director Marketing, Max Life Insurance, in an interview with ETCIO.

Artificial intelligence is making its mark in the service industry for improving customer relations through chatbots, speech recognition, and more. Major players in the industry are looking to tap artificial intelligence tools to discover customer needs at various life stages and offer appropriate products at each step.

Max Life Insurance Company Limited is a subsidiary of the publicly listed Max Financial Services Limited and is a joint venture between Max India Ltd and Axis Bank. The company was established in the year 2000 and is currently the 4th largest private life insurer and the largest non-bank life insurer in India. The company has a revenue (2021) of Rs. 16,184 crores and a market valuation of 20,000 Cr. The company has a wide portfolio of life insurance products - Protection insurance, health, pension, annuity, child, retirement, savings, and growth plans. The company employs around 10000+ people and is ranked the 18th Great Places to Work in India and the only Life insurance company in the top 50. 

The company witnesses a tremendous volume of customer interactions with over 3.5 Lakh customers voice calls happening daily across different touchpoints – Sales calls, policy renewal reminders, customer verification calls, tele-medicals, customer service calls, and more. The company has an online portal, where customers buy policies through the Max life website or partner websites, contributing to 30% of incoming volumes.

According to company sources, currently, out of every 100 customers, only 8 customers end up purchasing the policy online (8%), while others drop off. Thus, the key challenge is to clearly understand the exact pain points for not buying our policy (Pricing, product, competition, agent behaviour, etc.

The other area where AI-aided technology is critical is detecting fraudulent transactions and maintaining online security.

Improving Customer Experience with AI

In an attempt to improve customer engagement and efficient redressal of customer queries, Max Life Insurance Company Limited is now using VOX Speech AI Program. This initiative is harnessing the potential of AI to unlock insights from a huge heap of voice data to generate insight across customer needs, pain points, expectations for the ability to provide a customized experience and drive business from these customers.

The focus is to have an automated analysis of customer interactions over voice calls leveraging Speech AI (End to End speech analysis engine) using language-based speech analysis models along with NLP to derive insights from customer interactions across all types of the customer voice.

According to Company sources, “At Max Life, we have a tremendous volume of customer interactions with over 3.5 Lakh customer voice calls happening daily across different touchpoints – Sales calls, policy renewal reminders, customer verification calls, tele medicals, customer service calls. These personalized interactions with the customers have valuable insights beyond what can be captured in customer surveys and feedback but this data is not used currently due inability to manually analyze such huge volume of calls by listening each call.” 

The VOX program aims at building an end-to-end speech AI platform to analyze customer interactions happening through voice calls for every call interaction with the use of contextualized models across each type of customer interaction. In this phase, sales calls for the customer buying patterns and pain points to improve sales conversions shall be analyzed. In addition, customer verification calls for the identification of cases of misselling and obligatory sales to reduce complaints and improve the quality of sales will also provide actionable results.

Reminders such as renewal reminders to understand key reasons for not paying renewal premiums on time to improve renewal income and persistency calls for customizing interactions and provide upfront resolution can be done by VOX.

At the Core

The entire solution is spread across three layers: Data Layer - This layer is responsible for generating the audio call data from dialer equipment. This layer also processes the audio file to extract call metadata like callid, timestamp, agent id, etc. 

The next is the AI Layer - This is the layer where automated analysis of the customer interactions happens by leveraging AI to process calls. This layer is further divided into Validation Program, which validates the call type, labels and channels associated; The Max Life Versbose detects word phrases, short forms, phonetics, pronunciations, etc; The speech to Text Model Uses AI for Noise removal, Spectrogram, RNN-BLSTM lexicon, decoding search, word sequencing, etc; The transcription phase identifies accuracy, word-level confidence, and timestamps, along with speaker labels and call transcripts. Finally, the NLP program processes transcription outputs, calculates time and accuracy, processes text, performs stemming/ lemmatization and sentiment classification. 

The last layer is the Visualization Layer, which provides business insights in a graphical representation of data like call durations, monologue, conversations, overall quality rating, customer purchase levers, customer call sentiments, etc. 

The company is experimenting in various dimensions of Cognitive computing in delivering a conversational experience to the customer via bots along with the use of the machine and deep learning for underwriting purposes. Additionally, they are also automating some of their processes using Robotics.

There are challenges associated with the solution. These challenges are primarily around the handling of voice call data and ensuring customer privacy. Encryption of customer information like policy details and mobile number, masking customer personal information during transcript can be challenging and requires high security and scrutiny. Another challenge that lies in front is managing the volume data. With an average of ~5 mins call at a rate of ~15000 calls, the storage can reach up to ~300 GB per day. Yet another complexity stays in the voice call dataset. Splitting audio calls to extract customer and agent transcript, voice rate compression, background noise removal, and labeling of data for model training can be overly complex and sophisticated. 

Vox is based on a language model approach to maximize insights from the calls. It is customized as per the context of Max Life to enable transcription accuracy for more targeted actionable business insights. With an ability to handle Hindi, English, and Hinglish calls, Vox can handle various types of conversations. A high-quality dashboard for valuable information representation makes VOX an End to End platform for voice data processing. 

With changing technologies and growing awareness, consumers’ expectations are rising too, and to maintain a competitive edge in the market, the insurance sector has to build capabilities using customer insights. Its time to learn from the traditional principles of business and strengthen the approach with the power of the latest technologies and come up with a hybrid approach.