In an exclusive conversation with CXO Media and APAC Media, Santosh Bhat, Head – Data Science & Advanced Technologies, Policybazaar, explains how AI/ML is reshaping the insurance sector and enhancing its operational efficiency. GenAI is also being leveraged for a more personalised insurance product design.
How is AI/ML reshaping traditional insurance practices and enhancing operational efficiency and precision at Policybazaar?
At Policybazaar, technology is being used consistently to make the insurance journey for customers faster and simpler. By using Artificial Intelligence (AI) and Machine Learning (ML), we modernise traditional insurance processes to better serve the customers. One significant improvement is how we handle important documents like KYC forms, payment receipts, and policy copies. Using computer vision and machine learning, our systems can now read and process these documents instantly. This had led to fewer delays and less paperwork. These automated processes have improved our productivity by over 20–25%, thereby helping us to respond quickly and accurately to the needs of customers.
We believe that our use of AI is not just about technology — it is about making sure the experience for customers is hassle-free and offers reliable service every step of the way to earn their trust. Additionally, we have integrated AI-enabled vehicle inspections, which analyse over 40 attributes in customer-submitted videos, reducing inspection time by over 70% and improving customer satisfaction. Our hybrid chatbots also play a key role by combining natural language processing and human assistance to handle a wide range of customer queries efficiently. Over 50% of the customer chats are handled by AI bots without the need to connect to human advisors.
How is Policybazaar moving towards a personalised product design leveraging GenAI?
Policybazaar uses Generative AI (GenAI) to make insurance experiences for customers more personalised and efficient. A key innovation is our AI-powered call summarisation tool, which uses speech-to-text technology to capture and summarise conversations. This helps our contact centre advisors recall customer needs quickly and respond with tailored solutions. It also helps in saving time as there is no hassle of taking notes manually for the advisors. Policybazaar has
in-house speech recognition models, trained in Hindi, English, Tamil, Telugu and Malayalam with 90% accuracy. These models form the base for a fine-tuned AI system that understands the customers’ concerns, tone, and intent.
Built on advanced open-source models like Llama 3 and Mistral, this GenAI solution allows us to route their queries to the right team instantly and offer faster and more relevant support. This fine-tuning ensures our models not only summarise calls but also detect customer sentiment and intent, helping us deliver even more precise and personalised recommendations.
How does Policybazaar leverage the online platform to compare products and services and use digital touchpoints to handle claims and allied services?
Policybazaar has always believed in leveraging technology to improve customer experience and provide superior services. Our online platform offers insurance quotes from almost every insurer, helping customers compare policies to make an informed decision. Customers can easily log in to the website, enter details, and compare various policies in terms of features, benefits, inclusions, exclusions, price, etc. We have continually focused on innovations to drive seamless customer service and growth. For instance, we introduced telemedical services to make it easier for customers purchasing health insurance from home. So, instead of undergoing a physical medical test, customers can disclose their medical condition over a video call before buying health insurance.
Similarly, to address claims, we have dedicated Claim Managers who help with claim assistance to customers from start to settlement. For motor insurance, we have introduced AI-enabled video inspections, allowing customers to complete the process in 2–3 minutes with real-time feedback. Our hybrid chatbot system also ensures that routine inquiries are handled efficiently, while complex issues are escalated smoothly to human agents.
How does Policybazaar leverage analytics to generate meaningful insights from minimal data?
Policybazaar leverages advanced analytics and machine learning to derive meaningful insights, even from limited data. By using predictive modelling and behavioural analytics, the company identifies patterns in customer interactions across digital touchpoints. These models help understand user intent, preferences, and potential conversion behaviour. Our AI-driven systems continuously learn and improve over time. This helps ensure better targeting and higher engagement. By combining small data points with historical trends and aggregated behavioural data, Policybazaar makes informed decisions to enhance user experience and streamline product recommendations.
Additionally, transactional signals such as payment methods and device usage, along with behavioural data, are used to generate risk scores and personalised engagement, improving both marketing efficiency and risk management.
How does technology help to mitigate insurance fraud for Policybazaar?
Fraud management is critical in insurance, and at Policybazaar, we have transformed this area using a “Human-in-the-loop” AI framework, where AI and generative AI models assist quality and underwriting teams in making informed decisions. AI functions at the initial sourcing stage, flagging anomalies and outliers by analysing payment methods, device usage, and browsing behaviour. It leverages personal data and knowledge graphs to uncover suspicious patterns and hidden fraud networks. AI supports identity verification through face and voice recognition, along with liveliness tests to detect deepfakes and prevent identity fraud.
Additionally, machine learning models generate risk scores by analysing customer journeys, behavioural patterns, personal information, and call interactions. Our generative AI models also monitor customer-advisor calls to detect discrepancies between disclosures and proposal forms, prompting further investigation. These efforts have contributed to a 95% reduction in fraud and cut risk assessment turnaround time from two days to one, significantly strengthening trust and operational efficiency.
Rajneesh De, APAC Media
Also Read –
