Sofia Kyriakopoulou, a Fintech Strategy AI Champion and Group Chief Data & Analytics Officer at SCOR, spoke at InsurTech Insights revealing how GenAI innovation at one of the world’s largest reinsurers is transcending the realm of proof of concepts to become fully productive…
SCOR is as a tier one reinsurer – adaptable and business oriented. We are nimble, deeply technical and very focused on where and when we can play. We have an ambition to grow with our clients and see AI as the differentiator to allow us to innovate, offer new services, and to increase insurability. At SCOR, AI is not the future ambition. It’s here and now in InsurTech.
Delivering business value at scale with AI
In my role as SCOR’s Group Chief Data Analytics Officer I don’t just oversee data and AI initiatives, I aim to ensure they deliver business value at scale. Doing an AI proof of concept is relatively easy. Getting this into the hands of the user, that’s hard. And that’s what we strive to do at SCOR today. I want to give to you a glimpse of our approach and to share with you how we’re building the insurer of the future.
Since GenAI came into our lives in May 2022 we have all been following the frenzy of its unprecedented rise. It has the potential to change the way we work. And there are very few places where this potential is more relevant than insurance. We have ample data that has never been touched by digitalisation… The submissions, the contracts, the statements of accounts and so on. All of us who have been in data science who have tried the traditional models, we have seen the pain. The annotations, the laborious testing validation. And finally, if we could get it to work, it would’ve been so hard to scale it throughout the lines of business from the markets. Then came GenAI. And by now, many of us have figured out what it can do and what we would like to achieve with it.
GenAI can summarise what treaties look like across their addendums in agreements. And it can do more… Take out pieces of information and fill in my template, fill up my IT system, fill up my database. In the end, it’s not about what the technology does, it’s about what we do with it. And at the core we are focusing the opportunities in two places. Number one, the processes where it could create massive efficiencies. Number two, the new data points that could augment our analytics. And once we combine that with deep industry expertise, we believe that’s when we can go beyond pure automation.
Data in the DNA
At SCOR data is in our DNA. We’re a very technical company full of high calibre individuals and we’re growing. We see AI as a differentiator to be able to do more with the most precious piece of capital that we have – our people. So, we are embedding APIs where we believe it matters most. First of all, with our workforce. AI-powered tools are a commodity, but they’re essential. We are equipping our personnel with secure access to third party tools – essential to increase their effectiveness and efficiency. That’s where the value starts to arise – the processes. Identifying those document intensive processes where adding AI could significantly expedite them. Getting the data points for analytics that could make our decisions faster and more efficient. And that’s where it becomes really interesting, as we contemplate the jewels of the crown that we could be building.
AI-Powered InsurTech Underwriting
We believe differentiation comes from AI-powered underwriting and claims solutions. Through SCOR’s digital solutions, we’re coupling our internal knowhow with AI models to create an advantage for us as we use them internally and for our clients. We want to make the AI-powered underwriting process real when an application or submission comes in and triggers the engine to go straight to process. A certain percentage of cases work like that. Currently, if an underwriting repairer is needed the human looks at that. If we don’t have sufficient evidence that will need to be attained for the process to be reviewed again. Then we can decide what’s wrong with it. There are issues we can overcome… Number one – always back and forth with delays. Number two is human judgment – humans are not very consistent. And number three – we’re missing all the insights that we could have brought in from the past evidence. So, could we do better? Yes, we could add AI on all those human steps and augment them. We could do that. But could we look at it differently?
Could we think of this not sequentially, but at once, synthesising all the necessary data points when they’re needed. Looking at it again, we take evidence, we apply AI, we are structuring essential elements out, we’re triggering the underwriting rule engine, and then we’re adding any further information we have available that could support the decision making. Finally, we’re recommending to the underwriter what they should do. And to signify that this is supporting the underwriter, it’s an underwriting system and the small automation that could happen upstream. This is exactly what we’re currently using internally to augment life and health to support our underwriters.
SCOR’s AI Assistant
Our AI-powered underwriting capability is something we can provide to our clients through SCOR’s digital solution; we call it the AI Assistant. And here’s what it does in practice. When applications come in, we select the chain of thought that we’ll apply. For example, we ask it to think like a medical underwriter. It then extracts the essential pieces from the medical reports, joint records and vital family history. And then it creates a digital twin – the standardised pieces of information that the underwriters believe are the essential data points.
We store them and then we go deeper. We are putting the human in the loop so that humans can validate the actual sources of information. And then we complete the decision making. For example, the AI Assistant could detect an impairment and suggest the next course of action. This signifies the direction of sale. That’s the gold standard that we want to strive for.
Scaling AI with InsurTech
We don’t stop at experimentation. Data scientists like me, we love the tip of the iceberg. That’s where it’s exciting, and you can push to get the proof of concept to work. But in fact, under the water lies the very hard work one has to do… The building up of a data foundation; putting all the essential data assets together at the level of data quality that we can trust; establishing the necessary governance and then developing the IT platform in an equally robust way so it can scale. The proof of concept is not just an experiment.
We must plug in the actual IT landscape, the InsurTech tools where the AI is going to be consumed. And then you can go deeper and link the processes with the humans… In order to positively disrupt the process and keep the human in the loop they must be part of the journey from day one. We must educate our teams, demystify what AI is and isn’t. We must listen to their reactions because they’re the ones we will rely on to elevate the model performance.
Meeting the gold standard with InsurTech
Effective change management is for me, the essential element to allow us to go end-to-end. With insurance, and reinsurance, I believe we have come a long way. From the underwriting manuals to the rule engines, to the first AI models, probably now to the first cracking of the notorious submissions… It has been such a journey transforming both technology and the way we work. The shift, however, is beyond technology. It’s about how we operate, how we innovate, and how we create value for us and for our clients. Today, thanks to our ability to be nimble and technical at SCOR, we are in a position to connect all of the new capabilities of this value chain into what is an end-to-end comprehensive risk view. And for me, that’s the gold standard for InsurTech and what we are striving for with this AI revolution.
- InsurTech