AI: Unpacking the power behind two letters
AI in Life & Health: Article 1
Article information and share options
In this new series of articles focused on the use of AI in Life & Health Underwriting (UW) and Claims, we highlight key considerations for utilising AI in your business and help break down concepts for those that may be new to AI. We discuss the notion of Responsible AI and also share practical lessons from Swiss Re’s experiences using AI and how it can help you move forward in Life & Health insurance. In this first article we introduce a few high-level concepts for those who may be new to this space.
What is AI?
AI is a broad term that can mean different things to different people. Establishing a common AI definition within your organisation helps meaningful collaboration, informed decision-making and proper understanding of regulatory responsibilities. The OECD, as one example, recently defined AI as “a machine-based system that […] infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments…”. This definition includes Machine Learning and its subsets Deep Learning and Generative AI (Figure 1), but excludes expert-rules systems as they don’t use models to infer patterns from data.
Generative AI
| Traditional AI | Generative AI | |
| What is it | Custom-built analytics models that perform specific tasks. They typically use structured data to identify complex patterns and make predictions. | Models designed to learn patterns, usually from less structured data, such as text, and generate content out-of-the-box. |
| Strengths | Quick, consistent and explainable analytics to make classifications and predictions and detect anomalies | Analysing and summarising unstructured data, e.g. scanned documents. The models can often be generalised to new scenarios. |
Opportunities to leverage AI in Life & Health insurance
AI is most often explored to realise time and cost savings and is most impactful when functional experts have a well-defined job to be done (a use case). In Life & Health insurance, AI is typically used to address pain points such as:
- Extensive manual review of unstructured medical documents
- Data analyses across multiple systems to enable data-informed decisions
- Automation and transformation of customer experience(s)
Traditional AI has been used to optimise a wide range of activities across the value chain (see Figure 2). These “narrow” AI models have helped from predicting and optimising underwriting, to triaging claims cases with the highest chances of rehabilitation or fraud red flags.
Generative AI, mostly using large language models, presents new opportunities. Areas of exploration in UW and Claims include using natural language to search documents, summarise cases and to create digital assistants such as Swiss Re’s Life Guide Scout. Most importantly, one AI model doesn’t necessarily replace the other.
The growing future value lies in the powerful combination of multiple models to deliver customer value seamlessly.
Therefore, it is essential to understand how models differ from each other and what value each one brings to the table.
AI is used across the value chain
With new opportunities, come new challenges and risks
AI regulatory considerations are increasingly prominent. Early explorations of AI highlighted challenges that could lead to issues of various levels of severity, from significant embarrassment to severe reputational issues and lawsuits when failed to be identified timely, e.g. Amazon scrapping exploration of a recruitment model that proved discriminatory.
Countries are adopting new laws and regulations to balance the benefits of AI with consumer protection. Fairness, transparency, explainability, oversight/ governance and accountability are key principles. Insurers need to implement appropriate strategies to ensure responsible use of AI and monitor the rapidly evolving regulatory landscape.
Additionally, Generative AI’s promise comes loaded with a new set of challenges and risks. While it enables us to work with new types of information and data, we need to consider carefully what data we can use and how it is acquired and curated. In the absence of a robust framework, Generative AI models can “hallucinate” (produce nonsensical outputs), causing trust and acceptance issues. Ethical concerns relating to indistinguishable synthetic content, data privacy violations as well as intellectual property risks can add complexity to adopting Generative AI technologies.
The human factor in AI
The use of AI will be a key component of our work going forward. To achieve the most impactful results, close collaboration between underwriters, claims and data scientists to develop strategies and tools is crucial. Managers must consider upskilling opportunities and hire roles with expanded and/or different skillsets and other change management tools to support this transformation.
Wrap up and watch out for the next article in this series
AI is a complex field requiring a multidimensional and multidisciplinary approach to both strategy and execution. We will explore some of these concepts in greater detail in upcoming articles and share insights from our experience based on projects across the globe, so we can help you move forward in this space. Amongst the many priorities we juggle it is worth keeping up to date on this rapidly evolving AI landscape. Watch out for the next articles in this series for more insights and learnings. Up next is Responsible AI and key considerations. For more information, or to discuss services, tools and/or Solutions, please contact your local Swiss Re representative.
Disclaimer: We recommend you discuss your own use of AI & rules engines with your legal / compliance team considering local AI definition(s).