Get claims smart: Using AI to transform your Life & Health insurance claims management

The claims stage is when insurance truly delivers on its promise to support beneficiaries in their moment of need. It's a moment of profound emotional burden for those involved, when a high degree of sensitivity and an understanding of basic human psychology are paramount. In this vulnerable moment, customers seek empathy, clear guidance on the claims process, accessible support and transparency in the claim's evaluation. They also seek reassurance through positive claim experiences shared by their peers.

The use of AI has the potential to enhance insurance claims management by reducing manual work and offering decision support, resulting in faster and seamless service for claimants.

Consider the case of "John", an individual who has just lost a loved one. Amidst his grief, John submits a claim online with all the necessary documentation. He receives confirmation that the documents are in good order and that the policy covers the loss. As the claim moves into review, John is able to track progress and reach out to a claims expert for support and guidance. At the end of the process, he is reassured that the payout is on its way, offering a glimmer of relief in this testing time.

By the same token, the claims stage also fulfils the role of safeguarding the collective. Take the example of "Peter", an individual who submits a claim. In a first step, a responsibly designed AI system initiates the task of cross-referencing the documentation provided. The system detects inconsistencies across the information submitted, which indicates a potential misrepresentation. The human claims expert reviewing the case considers this as one of many inputs in the decision-making process. This illustrates how AI can augment the capabilities of human claims experts and expedite the claims processing.

Hurdles in Life and Health insurance claims management encompass the manual extraction of necessary data from lengthy and complex documents, the comparison of information across various sources such as medical and financial records, and the application of judgment to triage cases for subsequent steps. Common questions may include whether the event is covered by the policy, if all required documentation has been submitted, and if any indication of non-disclosure, misrepresentation or indication of fraud exists.

Technology plays an important role in facilitating a fast, smooth and transparent claims process for policyholders, while automating claims processing and reducing manual work for insurers. AI can assist in compiling the necessary data to support human decision-making and support the efficient triaging of more complex claims cases from straightforward ones (Figure 1). This article will explore how AI can aid in the notification of loss and claims review process.

The application of human judgment in complex life and health insurance cases remains essential. This approach is commonly known as 'human in the loop'. To create optimal and responsible Human-AI interactions, the concept of Responsible AI (RAI), as discussed in the previous issue of our AI-article series, is crucial.

While customers may still notify insurers via paper forms or phone calls, many insurers now offer digital claim portals, such as Swiss Re's Automated Claims Experience (ACE). This facilitates the reflexive identification of required information, document uploads and status updates.

The key challenge lies in managing unstructured data. Optical Character Recognition (OCR) technology combined with Natural Language Processing (NLP), Large Language Models (LLMs) and other AI forms can significantly improve the structuring of incoming information, a prerequisite for insurance claims automation and validation. In this process, documents are digitized, cleaned, classified and structured, with any missing or inconsistent data flagged for review.

Swiss Re has successfully piloted the use of AI to extract information from death certificates with high degree of accuracy. Additionally, AI has been employed to structure free-text inputs and classify them into predefined causes of claim, reducing the frequency of uncategorised entries from 20% to 5%. For both initiatives, it was essential to train models using varied document formats, update models with new information types, and classify and split documents for extraction purposes.

Data from the Notification of Loss step flows into a claims or policy administration system, which may incorporate business rules and AI models to assist the triaging of claims into differentiated workflows. In this environment, AI can be employed to detect anomalies or to conduct a preliminary analysis of claims based on various factors.

Swiss Re has leveraged AI to identify patterns that can assist claims experts in their assessment and management of claims. This can be a valuable aid to enhance resource allocation. For example, AI can be used to triage complex claims for operational efficiency or to identify opportunities to support disability claimants in returning to work or improving their health. To the latter, statistical models can help prioritise personalised support for those who need it most by leveraging bio-psycho-social (BPS) factors, such as motivation and coping skills. These factors have significantly higher predictive power than traditional metrics like age and cause of loss. The BPS Triage Tool has been validated with several of Swiss Re's clients and just recently won the UK Vocational Rehabilitation Association's project of the year 2024 award. It is poised to enhance Swiss Re's Claims Automation Solutions as of early 2025.

Attention to the following areas is critical to use AI to enhance claims management successfully:

1. Focus: Identify and understand the business issues you want to address and manage stakeholder expectations regarding the potential of your present AI capabilities and its limitations. AI is a means to help you achieve your business goals, so the basic questions are: Is my investment in this project worth it? Can it be done? Will people want it?
2. Follow a Responsible AI framework: When designing an AI system for life and health insurance, it is essential to incorporate the principles of the Responsible AI framework from the outset. It is crucial to rigorously test for unfair discrimination, promote transparency by making AI model outputs understandable, and ensure robust data and AI security to prevent data breaches or cyber incidents. Additionally, privacy must be safeguarded through data pseudonymization and minimization. Human oversight is vital to challenge AI outputs and prevent over-reliance on the technology.
3. Secure high-quality data: Poor data undermines the accuracy of AI. The historical data used to train a model needs to be accurate, representative, well-labelled and capturing relevant information in a variety of formats. 
4. Keep it simple and consider partnering: Given the high costs associated with training and maintaining AI systems, it's crucial to approach this space with an open mind. Sometimes, the best solution may not involve AI at all. If AI is necessary, purchasing an established AI capability might be a quicker and more cost-effective option than developing one in-house. Key considerations should include whether the AI system addresses a fundamental aspect of your business and whether an existing vendor can provide a level of quality that your company cannot achieve internally.
5. Designing Human-AI interactions aligned with business processes: AI should be seamlessly integrated into systems, with Human-AI interactions thoughtfully designed and aligned with claims processes to maximize impact. In the disability claimant's example above, a holistic claims management strategy for support and intervention is essential for enhancing customer outcomes.

AI has the potential to greatly improve customer interactions for routine inquiries and submissions in Life & Health insurance, allowing claims experts to focus on more complex claims and human interactions. To achieve this, it is essential to effectively integrate AI into systems that leverage its speed and precision alongside human creativity, curiosity, empathy, and critical oversight. The goal is to augment the claims expert by combining the strengths of both human and technology to make the world more resilient, one claim at a time.

In the next article, we will explore practical applications of traditional AI in Life and Health insurance underwriting. 

For further information on our lessons learned and to partner with us on the journey, please contact us. Read also issues 1 and 2 in our AI-articles series, "AI: Unpacking the power behind two letters" and "Responsible AI: what it is and why it matters for insurers", respectively.