Like most industries, Insurers have been leveraging AI technologies for some time. Fraud detection has capitalised on machine learning based systems to thwart bad actors and ensure authentic claimants are supported as fast as possible. In underwriting, tools like Swiss Re’s Magnum automate decisions and bring more flexibility to how premiums are priced. Also, various advanced analytics capabilities use a lot of different AI tools for risk analysis.
That said, the real transformation is yet to come. The latest advancements in AI, particularly Generative AI, present groundbreaking opportunities to reshape customer experiences and redefine how insurers deliver value.
However, despite the excitement AI has generated, insurers face significant barriers to fully realising its potential. Legacy mindsets and systems, regulatory complexity, and ethical concerns stand in the way of true disruption. The result, AI adoption is in somewhat of a pilot purgatory, where insurers, who are still largely in monolithic legacy systems and siloed technology architectures, try to avoid solving the unstructured and non-operational data problem. But, this isn’t working.
The Data Dilemma
Historically, when we think of having dialogue with a machine-based solution, most of us think of rubbish chatbot experiences. These have largely been “programmatic” and deterministic. Whilst the machine is identifying certain words, it is also categorising the question so that it can then refer to specific default answers that it’s allowed to give. A very narrow and prescribed structured data set.
It’s been a long journey, but we have now started to build more probabilistic models, where language and intelligence are fluid. However, the outcomes are therefore less predictable. So, insurers now need to train these models on their own data sets, and apply them in a way that means they can keep a human in the loop and a human in control. Or typically both of these things.
Big Data has the ability to change the way we see people, which impacts on things like insurance and risk pricing. It should also change how we operationalise and use data in insurance as well. But there’s a problem. Legacy technology in insurance is rife, and most of the newer technologies in play mimicked the way insurers worked in those legacy systems.
This has created a fracture with analytical data often kept separate from operational data. That’s a problem. While historical data can help with long-term analysis, delivering smart, real-time customer experiences requires immediate access to fresh, operational data.
To make real progress, insurers need to treat data like a perishable asset, constantly extracting insights and acting on them in real-time, either through human decisioning or enabling AI to do it. For technologies like Natural Language Processing (NLP) to work effectively, we need better-quality data, delivered through systems designed to make that data accessible and actionable in the moment.
On the flipside nearly 80 percent of enterprise data is unstructured, coming in the form of emails, text documents, research, legal reports, voice recordings, videos, social media posts and more. For insurers looking for answers, this unstructured data is a goldmine. However, compared to structured data, it’s much more difficult to analyse.
Fortunately, evolving technologies, such as NLP, can enable insurers to unlock the value. However, once this insight needs to turn into action the legacy trap will again bite hard. Making changes to products and services, for example, is painful and slow.
AI and Insurance Needs a New Business Model
AI needs some hard yards put in to become the powerhouse driver of positive change we all know it can be in insurance. It needs the right foundations, and in insurance this means new foundations, of that I have absolutely no doubt.
Data fluid: able to treat data as a perishable asset, constantly mining it for insight and acting on it as close to real-time as needed.
Built around a customer and not a policy: For example, operating an “API” first model and assuming this is the way you successfully integrate is not sufficient for orchestrating expanding ecosystems. You need to be able to rapidly adapt to make those partnerships work in the context of a process, efficiency, and customer/employee experience.
Multi-agent & multi model: Whatever AI you are using today it will likely be obsolete very soon. The ability to harness AI is also about inter-operability and control. Probabilistic models offer us a new way of interacting and shaping outcomes. Agentic AI will take this to another level. Insurers must make sure they have the right kit in place to move quickly and outpace their competition.
The reality is that as tempting as key AI use cases might be, wiring them into legacy in insurance could be a very costly mistake. Complex point solutions welding you to key GenAI models creates dependencies on specific AI models that aren’t easily changed. Equally harnessing these services to ensure you don’t fall foul of a hallucination, or ways to control the data can be highly complex.
Let’s not build AI Skyscrapers on bungalow foundations.
A vision for an AI-Driven Insurance future
At EIS, we believe insurance needs to be reimagined. Concepts like policies and rigid product structures that don’t allow insurers to think and act like eCommerce businesses will give way to new intelligent and customer-led businesses.
This vision paves the way for “embedded insurance”, where coverage is seamlessly integrated into the products, services, and experiences we use every day. Imagine car insurance that’s usage-based, automatically detecting incidents and initiating claims without any paperwork. Or protection products that evolve with us, adapting to life’s changes in real time.
Even better, insurance can move beyond just paying out after something goes wrong. It can actively help prevent issues, like using smart water valves and leak detection tools to stop water damage before it happens, a problem that currently accounts for half of all home insurance claims.
All of these examples need intelligent data orchestration manifesting in customer and employee experiences, and in parallel changing the nature of our relationship with insurers. Moving us away from merely price led choices to longer term and more trusted partnerships. New AI capability makes all of this theoretically far easier than ever before.
Where Next Then?
We really do need to think AI can do everything in insurance, but then rationalise this quickly. The feasibility of AI to add value to most use cases I track today is high, but the viability is relatively low.
One of the quickest ways through this is to look at low risk but high value use cases, and build these on top of properly thought through technology foundations that are evolving rapidly. Doing this in turn solves a lot of the knowledge gaps, systems transformation and organisational and regulatory barriers as well. Thus making use cases that were higher risk but equally high value a lot more viable in the future.
AI models can make customer experiences more intuitive for customers, they can help personalise experiences and they can even help us optimise these – or even optimise themselves and how they make all this possible.
AI can also be used in operational context shaping call centre scripts real-time, providing call handlers with context information like where the customer has come from and what they have experienced. Or easier stuff like guiding operational processes, and how to optimally complete tasks.
On the right foundations AI models can evolve rapidly and reshape insurance so we can start to finally grapple with risk-mitigation, embedding insurance into people’s lives, things and businesses and make insurance a constant adaptable force for good.
