Republish from the special edition of Nextdeal (Issue 560)

Sofia Kyriakopoulou, Group Chief Data & Analytics Officer, SCOR

ChatGPT entered our lives just before the end of 2022, bringing Artificial Intelligence into our daily routines more directly than ever before, through what became known as Generative AI. For the months that followed, boardrooms across the globe were preoccupied with the same question: what is this new technological capability, and what does it mean for us?

Machine Learning had already been around for many years, full of promises, and disappointments. Many companies invested time and resources in developing algorithms through simpler or more complex statistical methods. Huge efforts were made to collect data, build models, and integrate them into enterprise systems, only for companies to later realize the results fell short of expectations. The benefits, though real, were limited.

Yet, this new evolution of AI felt different from the outset.
And what if it wasn’t just an evolution, but a true revolution? 

Sofia Kyriakopoulou

Two and a half years later, among professionals working with AI - especially in the financial services sector - the answer is becoming increasingly clear. We are witnessing the dawn of a revolution. This new generation of models is visibly different from anything we have seen before. Generative AI models come pre-trained, understand written language with remarkable proficiency, and most importantly: they excel in processing unstructured data – the very kind that is abundant in the insurance industry.

Looking at the current capabilities of AI that we have seen in practice, we know it can classify documents (e.g., what type of file is this?), summarize information (e.g., what changed in this policy over the years?), extract details (e.g., specific terms from a contract), and more recently, even perform reasoning (e.g., resolving a problem through logic).

However, the real discussion is not about AI’s capabilities in isolation, but whether they can be practically leveraged and how an enterprise-wide strategy can be implemented to deliver tangible outcomes and value.

Two Strategic Hypotheses with Real-World Impact

1. Faster and more accurate decisions.
In the insurance industry, every decision is based on data, starting from client submissions, to pricing, underwriting, claims, reserving, and finance. The hypothesis is to examine whether AI can enable teams to make faster, more informed decisions, with deeper understanding of each parameter involved.

2. Automation of repetitive processes.
The insurance lifecycle involves many steps that require manual intervention to interpret or transform data. Today, this takes up countless person-hours without adding real value. The second hypothesis, then, is whether we can fully automate these time-consuming processes, freeing up valuable time for truly strategic, human-driven tasks.

From Theory to Practice
We now know that both hypotheses can be realized.
The Data and AI teams of the world’s leading organizations, including the SCOR team I have the privilege of leading, have moved beyond experimentation and into real implementation.

We have seen unparalleled effectiveness when placing the human at the center of decision-making, supported by new AI-powered tools, especially in underwriting and claims handling. The true power lies in the ability to synthesize information from dozens of sources, documents, emails, systems, to present a comprehensive picture, enabling the opportune decision at the opportune moment. 

At the same time, we have seen remarkable acceleration in process digitization, including file classification, summarization, and data extraction. However, before we can speak of full automation that would free up human involvement, AI models must reach near-perfect accuracy and reliability. At present, the prevailing approach is to keep the human element as safety net that monitors AI-driven data flows and intervenes when necessary (Human-In-The-Loop). These interventions are precious, allowing models to improve over time.

A Practical Example: Augmented Underwriting
Take a typical insurance application. During the medical records review, it often becomes evident that additional documentation is required. Until recently, this triggered a back-and-forth process: identifying missing documents, communicating with the applicant, collecting, and verifying new data, comparing it to similar cases, and finally pricing and deciding on the risk.

In today’s new reality, we can execute all these steps at once:
By using AI models, the case is analyzed, all key information extracted, any missing evidence automatically identified, comparable historical cases retrieved, including their outcomes in terms of risk, and all findings synthesized. Only then is a summary presented for final decision-making. The result: faster, better-informed decisions, with improved consistency.

The Future Has Already Begun
Artificial Intelligence is no longer a theoretical possibility. It is a strategic differentiator, to drive informed decisions, to operate efficiently and to deliver innovative solutions, in a dynamic global risk landscape. The question is no longer if it will be adopted, but how quickly each organization will adapt to the new reality.

The future is here. And it is up to us to seize its full potential, with courage, imagination, and a clear strategy.

Read below the article from the special edition of Nextdeal (Issue 560) in electronic format, page 39 (click bottom right to enlarge)