AI Revolutionising Insurance: Insights on Data Unification, Customer Experience, and Ethical Innovation

Step into the world of innovation and digital transformation with Prasad Pimple, Executive Vice President & Head of Digital Business Unit at Kotak Life, as he shares his visionary approach to leveraging AI and data unification. In this interview, Prasad reveals how AI-powered solutions—such as chatbots, real-time analytics, and generative AI in contact centres—are revolutionising customer acquisition, retention, and operational efficiency. He highlights the importance of ethical AI adoption, phased deployments, and setting achievable goals, while offering a forward-looking perspective on groundbreaking advancements in AI-driven medical underwriting and insurance practices.

Key Highlights from the Interview

  • Data Unification and CRM as a Single Source of Truth: By centralising customer data through a CRM system integrated with analytics tools, the organisation achieves a 360-degree view of customers. This enables seamless tracking, personalised communication, and improved decision-making across customer acquisition and retention campaigns.
  • AI-Driven Campaign Optimisation and Engagement: AI tools, such as Meta’s Conversion API and Google Analytics, optimise campaigns by focusing on mid and bottom-funnel actions, resulting in precise targeting and improved conversion rates. Chatbots enhance customer interactions, assist in policy renewals, and deliver personalised updates, boosting satisfaction and loyalty.
  • Contact Centre Transformation through Gen AI: Gen AI solutions address variability in agent performance by providing consistent service, real-time feedback, and quality assurance. AI audits 100% of calls, offering actionable insights to agents and enabling real-time corrections for superior customer experience.
  • Ethical AI Implementation and Data Transparency: The organisation embeds ethical AI practices by ensuring transparency, obtaining informed customer consent, and adhering to regulatory standards. Processes like liveness checks and photo ID verification build trust and maintain data integrity.
  • Future Trends in Insurance with AI: AI is expected to revolutionise medical underwriting by replacing invasive physical tests with non-invasive, AI-driven assessments, simplifying the insurance process.

“In the long term, while some roles may be reduced due to AI implementation, many employees will move into more strategic and analytical positions, focusing on improving processes and planning.”

Can you share your experience with data unification and creating a single source of truth for achieving a 360-degree view of the customer? From a technology perspective, how has this evolved over the years, and what AI use cases have contributed to achieving this?

Prasad: From a data unification perspective, most of the industry is using multiple platforms rather than a single Customer Data Platform (CDP). While we don’t rely on a standalone CDP, the CRM acts as the central hub for unifying customer data and driving actionable insights.

  1. CRM as the Core: CRM serves as a mini-CDP, recording all customer interactions and touchpoints. It forms the foundation of our unification strategy by consolidating data from various sources.
  2. Tracking Digital Footprints: We use analytics tools like GA4 or GA360 to capture customers’ digital activities, from viewing and clicking on campaigns to expressing interest in our products. This data is integrated into the CRM, enhancing its value.
  3. Existing Customer Ecosystem: Information about current policies and customers within our ecosystem is also integrated into the CRM. This allows for seamless marketing campaigns targeting both new and existing customers.

By centralising these inputs, our CRM becomes the single source of truth for customer information, integrating tracking tools, policy data, and campaign management systems into one cohesive framework.

Regarding AI-driven enhancements, AI is leveraged to optimise campaigns more effectively. For instance, Meta’s Conversion API enables deeper funnel optimisation. Traditionally, campaign optimisation focuses on top-of-funnel actions like form submissions. However, we now optimise for mid to bottom-funnel actions, such as customers reaching payment or final conversion stages. This real-time optimisation, supported by both Meta and Google ecosystems, refines target segments based on actionable insights.

By passing conversion signals from CRM to these platforms, AI systems can identify and target audiences with higher conversion potential, leading to more precise targeting and improved campaign performance.

From a customer experience and retention standpoint, how has this data unification initiative impacted your metrics? Specifically, after implementing AI, have you observed a significant improvement in customer experience and retention outcomes?

Prasad: Today, the insurance industry’s current initiatives with Google and Meta are primarily focused on customer acquisition. By integrating conversion APIs with their ecosystems, we share data with their advanced models to identify and target high-potential customers. This enables us to optimise our marketing campaigns, resulting in improved conversion rates and more efficient customer acquisition efforts.

Regarding customer retention and experience, we have implemented AI and machine learning-based solutions to enhance customer interactions. While these are not generative AI (GenAI) initiatives, our use of chatbots has significantly improved the way we engage with customers at various stages of their journey.

During the acquisition phase, chatbots assist prospective customers by addressing queries and providing detailed information about our offerings. For existing customers, chatbots play a pivotal role in policy renewals, answering questions, and resolving issues promptly.

Additionally, we use chatbots as proactive engagement tools, sending personalised updates and communications through platforms like WhatsApp. This ensures that customers stay informed and connected, contributing to a more satisfying and seamless experience.

These AI-driven efforts not only enhance the efficiency of our operations but also improve customer satisfaction and loyalty by providing timely, relevant, and personalised interactions.

Are there any specific areas in your day-to-day activities where you would like to highlight recent AI use cases you’ve developed?

Prasad: There are several initiatives we are currently working on, many of which align with industry trends. Here are two key use cases we’ve focused on in the contact centre space:

1. AI-Powered Contact Centres for Enhanced Customer Experience

A significant challenge in contact centres is the variability in customer experience due to human factors. Despite uniform training, individual skills, fatigue, emotional state, and time of day can impact the quality of service. For instance:

  • Skill Variations: Some agents naturally excel at providing exceptional service, while others may struggle.
  • Fatigue: Performance tends to decline toward the end of a shift, affecting the consistency of interactions.
  • Emotional State: Personal circumstances can influence an agent’s ability to deliver a positive customer experience.

To address these challenges and bring consistency, we have piloted a Generative AI-based contact centre. This system can:

  • Answer customer queries accurately.
  • Assist customers throughout the purchase journey.
  • Maintain consistent performance without being affected by fatigue or emotions.
  • Reduce reliance on a large workforce by enabling one system to handle the workload of multiple agents.

The initial pilot has shown promise, with improved efficiency, consistent service delivery, and cost savings. However, we are cautious about scaling the solution, ensuring it adheres to all guardrails and maintains the desired customer experience.

2. AI-Assisted Callers for Real-Time Feedback and Quality Assurance

Another area of focus is improving agent efficiency and contact centre performance through AI-driven assistance. Traditionally, quality assurance involves reviewing a sample of calls and providing feedback to agents, which:

  • Limits insights to a subset of interactions.
  • Relies on agents to remember and implement feedback during live calls.

To overcome these limitations, we’ve implemented a solution that:

  • Audits 100% of calls: Instead of sampling, all calls are reviewed, and detailed summaries are provided to agents.
  • Feedback Summaries: Daily summaries highlight areas of improvement, showing what has been addressed and what needs attention.
  • Live Assistance: We are trying to enable real-time assistance during calls. If an agent veers off track or fails to address a customer query effectively, the system can provide instant feedback, guiding the agent to correct the course during the interaction.

This initiative has been promising, enhancing real-time support, reducing the reliance on post-call feedback and enabling agents to provide a consistently better experience.

These AI-driven solutions are transforming our contact centre operations, improving both agent performance and customer satisfaction.

There’s ongoing discussion about AI replacing jobs. How has AI impacted staffing in your contact centre? Have roles been reduced, or have employees transitioned to other responsibilities as AI takes over certain tasks?

Prasad: Currently, AI has not fully replaced roles in contact centres, but its potential to handle transactional activities more efficiently is clear. While some jobs may eventually be replaced, the transition is likely to redefine roles rather than eliminate them entirely.

At present, many contact centre tasks involve transactional processes, such as addressing customer queries and fulfilling requests. These are areas where AI excels due to its consistency and ability to process vast amounts of data. However, AI still cannot fully replicate the human touch needed to strategise, analyse, and enhance the customer experience.

As AI takes over transactional tasks, human roles are evolving toward higher-value activities, such as:

  • Strategising and Process Improvement: Employees can leverage their experience to analyse customer interactions, generate insights, and develop plans to enhance the overall experience.
  • Backend Support for AI: Humans will play a crucial role in managing and improving the AI systems themselves, ensuring bots are trained with accurate and relevant information. They will also monitor AI interactions, stepping in when necessary to handle complex scenarios or nuances that require human judgment.

For instance, during pilot tests of AI-driven calling bots, human agents observed and intervened when the bot encountered complex situations, ensuring a seamless customer experience. This hybrid model highlights the evolving collaboration between AI and human agents.

In the long term, while some roles may be reduced, many employees will move into more strategic and analytical positions, focusing on improving processes and planning. This shift will elevate their roles from execution-based tasks to insight-driven functions, emphasising the enhancement of customer experience rather than merely delivering it.

AI is not expected to fully replace contact centre roles but will reshape them, creating opportunities for employees to move up the value chain as these technologies mature.

What metrics or KPIs have you developed to measure the success and ROI of these initiatives, particularly in terms of their impact on customer experience and process standardisation?

Prasad: Our KPIs remain consistent with what we have tracked traditionally. For the acquisition process, we focus on metrics like lead conversion ratios and final conversion ratios. From a marketing campaign perspective, we track KPIs such as cost per click (CPC), cost per lead (CPL), and cost per final conversion (CPF).

What we primarily evaluate is the delta—the improvement brought by deploying these solutions. For instance, are we observing a positive shift in payment conversion ratios? Are we seeing a reduction in cost per acquisition (CPA)?

While the core business KPIs remain unchanged, the emphasis is on assessing how these initiatives enhance efficiency and drive improvements in these metrics.

Have the organisations you’ve worked with established an AI ethics committee or a team of leaders to ensure the responsible and ethical use of data and AI?

Prasad: In the life insurance industry, ensuring the ethical and responsible use of AI is not solely the responsibility of a dedicated committee. Instead, it is ingrained in the organisation’s approach, mainly because data privacy and security are of paramount importance in this highly regulated and compliance-oriented sector. Every solution, whether AI-based or not, undergoes rigorous checks to ensure compliance with regulatory requirements and data security standards before deployment.

A critical aspect of ethical AI involves transparency with customers. Whenever data is collected, it is essential to communicate why it is being gathered and how it will be used. For instance, in our life insurance purchase journeys, we provide tooltips for every data field. These tooltips explain the purpose of the data collection—such as asking for the name as per government documents for KYC verification—ensuring customers understand the rationale behind each request.

Additionally, certain processes, like fraud prevention measures, are designed to maintain integrity while being transparent to customers. For example:

  • Liveness Test Check: As part of the online purchase journey, where there is no face-to-face interaction, customers are required to take a selfie using their mobile phone. This selfie undergoes a liveness check to verify that the individual is alive, helping prevent identity fraud.
  • Photo ID Verification: The system cross-references the selfie with the photo ID proof provided by the customer. A matching score is generated, and this information is shared with the customer to ensure transparency.

These processes demonstrate the organisation’s commitment to ethical AI by ensuring customers are informed and confident about how their data is being used. By prioritising transparency, data privacy, and security, we build trust, empowering customers to provide informed consent.

Rather than relying solely on a committee, the principle of ethical data handling is embedded in our practices, ensuring customer data is always treated with the utmost respect and care.

What challenges have you faced in implementing AI technologies, and what guardrails would you recommend for those planning to implement AI solutions, particularly in the insurance industry?

Prasad: When implementing AI solutions, it’s crucial to prioritise a thoughtful and measured approach rather than rushing into deployment. A hasty rollout can lead to unforeseen issues, especially when dealing with potential data biases and the variability in how AI interprets the data provided. These variations can significantly impact customer experience, making thorough testing an essential part of the process.

A phased deployment strategy is highly recommended. Begin by developing the solution and conducting rigorous user acceptance testing (UAT). Before launching broadly, keep the solution in a Closed User Group (CUG) environment for an extended period. This allows you to test the AI against multiple scenarios, particularly extreme scenarios that are harder to handle than typical use cases. These extreme cases help identify weaknesses or areas where the AI may provide inappropriate responses. Feedback from this testing should be incorporated into the system to refine its capabilities further.

Another key consideration is setting clear and realistic objectives. Avoid overly ambitious expectations, such as assuming the AI can immediately replace an entire contact centre or replicate complex human functions overnight. Instead, focus on specific use cases and KPIs. For instance, in a contact centre context, aim for incremental improvements, such as providing better assistance to agents or enhancing the customer experience, rather than attempting to fully automate all functions at once.

It’s important to manage expectations realistically and take a step-by-step approach. Begin with smaller, manageable implementations, evaluate their success, and expand gradually. This iterative process ensures the solution is robust, aligns with business goals, and delivers tangible value.

Finally, resist the urge to rush deployment in the name of innovation or being the first to market. Success lies not in speed but in the quality and effectiveness of execution. A well-tested and carefully deployed AI solution will ultimately provide far more sustainable advantages than a rushed implementation that fails to deliver on its promises.

As someone with extensive experience in this field, what advancements or trends do you foresee in the insurance industry over the next 3-5 years with the implementation of AI across various areas?

Prasad: In the life insurance industry, one of the most critical aspects is underwriting customers for the long term. Life insurance contracts often span decades—some as long as 40 to 50 years. For instance, a term insurance plan for a 25-year-old with coverage until age 75 involves a 50-year commitment. Similarly, savings and investment plans often extend over 40 to 50 years.

Underwriting for such long-term contracts involves assessing customers both financially and medically. Currently, the industry relies on two primary mechanisms for medical underwriting:

  1. Physical Medical Tests: Customers visit diagnostic centres for prescribed tests, including blood and urine samples, followed by a medical examination conducted by qualified doctors. These results help determine the customer’s eligibility for insurance products.
  2. Video Medical Assessments: For certain products, customers participate in video medical examinations. During these sessions, they answer health-related questions, and, if necessary, a follow-up physical test is conducted before issuing the policy.

These processes, while effective, require significant effort from customers, such as visiting medical centres or arranging home visits for sample collection.

With advancements in AI, the future of medical underwriting is poised for significant transformation. In the next 4-5 years, we may see a shift toward more automated, non-invasive medical assessments. Emerging technologies already allow for clinical correlations through video or photographic analysis, providing vital information such as blood pressure, blood glucose levels, and even stress indicators. These innovations can eliminate the need for intrusive procedures like blood or urine sample collection.

Once these technologies are clinically validated and widely adopted, they will simplify the underwriting process, enabling insurance companies to offer policies entirely online. This will streamline the customer journey, removing the need for physical tests and making the process more accessible and convenient. As a result, insurers can expect to reach more customers, enhance their digital capabilities, and drive greater adoption of life insurance products.

From your leadership perspective, what insights or benchmarks would you find most valuable in a report exploring AI adoption strategies in the insurance industry?

Prasad: It’s primarily about understanding what others have done, including their successes and, more importantly, their failures. Often, only success stories are highlighted publicly, showcasing the positive impacts or ROI achieved through AI initiatives. However, behind every success, there are likely multiple failures or challenges that brands face during the process.

Having access to a repository of these failures or pitfalls in AI and generative AI implementation would be invaluable. It could serve as a guide for others to avoid similar mistakes, making their experimentation and adoption more informed and efficient. Such insights would significantly contribute to better execution and understanding of AI-based strategies.