Bridging the expectation chasm: Insurance and the impact of AI

One of the best parts of my work at hyperexponential is collaborating first-hand with the insurance market. I get to witness how cutting-edge technology is being implemented across a broad spectrum of scale and pace and have insightful conversations with leaders and practitioners in the field.

This unique opportunity, combined with my background in software engineering, and front-line experience as an actuary, give me a distinct perspective to evaluate the intersection of AI and insurance.

While it’s clear that we are at the beginning of a significant AI-driven technology revolution, it’s not in hyperexponential’s DNA to parrot buzzwords without underlying substance. As the hype starts to subside (slowly!) and we in the insurance industry begin to see clarity in the role AI can play, I’m excited to share some of our thinking with you.

AI and insurance

Insurance is a critical part of the world’s socioeconomic infrastructure: the protection we provide to our clients enables them to take more risk in their endeavours, whether it’s homeownership or launching rockets into space. When they reap the rewards from those risks, society advances.

Given our work’s impact and leverage on society, insurers have a responsibility to support and innovate. At hyperexponential, this is a key part of our mission: pricing serves as the primary gateway through which insurers assess and admit risk, so our role in improving these processes is vital.

Over my nearly two-decade tenure in the insurance industry, I’ve witnessed a growing chasm in between clients’ expectations and the value delivered from their cover. Insurers serve cutting-edge companies that are at the frontier of innovation, and as the world becomes increasingly data-driven, so should our industry.

We’re seeing MGAs such as Floodflash and Gaia begin to push the capabilities of what’s possible with innovative approaches to IoT sensor data and customer experience—yet we aren’t making the most of the opportunities this gives us to better manage their risk.

A key driver of this gap is our inability to harness technology at the pace that the world needs. If we strip the problem back to its essence, it’s really our inability to leverage both internal and external data that’s holding us back. A reliance on siloed systems, outdated spreadsheets, and solutions that fail to capture data at critical junctures limit our ability to leverage data in our decision-making processes.

Take the examples above: these companies are driving innovation in product design and customer experience while relegating the carrier to “dumb capacity.” This is happening even though the established players have a potentially significant advantage at their disposal with respect to data, underwriting expertise and industry experience.

Potentially is the key word. AI can significantly unlock this potential—however, to make the most of it, some key foundations need to be laid.

With all the undeniable excitement surrounding AI, the industry is overwhelmed with noise. In this article, we cut through to the challenges that insurers face, the ways AI can improve those processes and outcomes, and how it will enable teams to do useful, impactful work.

How decisions are made

At hyperexponential, the promise of AI lies in how it can improve pricing decisions. Unsurprisingly, can be mapped directly onto our framework for Pricing Decision Intelligence:

1. The data we leverage:
   This refers to the raw information and evidence we collect, which forms the basis of our analytical efforts.

2. The insights we extract:
   It's the interpretation of data that uncovers valuable information, combined with our judgemental beliefs about the world we are modelling, guiding us to comprehend the implications and underlying patterns.

3. The decisions we make:
   The culmination of the process, where our data and insights are used to drive informed choices.

Ultimately, the quality of decisions is a function of the quality and quantity of our data and the depth of understanding we glean from it. With a shared appreciation for these foundational elements, let’s get to the subject at hand.

Fundamental characteristics of insurance and the challenges of applying AI

So how can AI impact the way we make decisions? Before thinking about that, it’s instructive to think about a few distinguishing characteristics of insurance. These provide important context that impose powerful constraints and dictate where we should focus our efforts to maximize impact.

1. A game of outlier management

Insurance exists to protect us against unforeseen events. By definition, these are extremely difficult to predict and factor into our decision-making process.

This challenge is exacerbated when we observe that many lines of business are "fat-tailed", where the impact to a given portfolio is determined by outlier events. In this instance, how can we rely on models powered by historic data to make decisions?

If these factors aren’t in the data, how can any models we build make accurate inferences on them?

That’s not to say that we won’t be able to build models which factor in such outliers (e.g. via retrieval augmented generation or fine-tuning), but those will necessarily be judgemental inputs to the model, not outputs. The companies with the most relevant and useful judgements will maintain a comparative advantage.

2. A very mixed bag

We operate in an industry characterized by a very diverse mix of risks within a portfolio, both in between lines and within lines of business, ranging from personal lines to complex cyber risks.

This heterogeneity requires drastic variations in the techniques, technology and underwriting workflows - which can be extremely difficult to manage. Consider the use of detailed bespoke engineering reports in construction/energy vs the practically zero-touch underwriting of modern personal lines risks.

To accommodate this variability, the decision stack for different lines needs to be configured and "wired-in" differently, and indeed this is why we sought to make hx Renew the most versatile, integratable platform on the market – a one size tool/workflow does not fit all, and AI needs to play a different role in different places. 

3. A syndicated, dynamic and competitive market

Risks are often “sliced and diced” as they are priced and underwritten. More often than not, multiple carriers compete for business while also sharing out the risk between them. Though it is most common at the commercial end of the spectrum, most insurers are exposed to some sort of risk sharing mechanism—even if only through their own reinsurance programs.

This market dynamic makes for a very interesting challenge when utilising models at the “pointy end” of the underwriting decision. A fundamental aspect of model training, which applies no less to AI, is measuring the effectiveness of a model’s prediction against historic actual outcomes.

Establishing a function that balances the value of co-operation against competition, especially when this can vary by line of business and even over time (for example as CAT capacity is exhausted over a renewal season), is a difficult challenge and is probably impossible to do completely.

It's difficult to ask a model to help you, when defining winning is challenging.

The opportunity with AI

At this point, you might think we’re pretty pessimistic around AI, but that could not be further from the truth. It’s probably clear that at hyperexponential we don’t buy into the concept of the completely automated, zero-touch underwriting operation. Certainly not at the complex end of the risk spectrum.

However, we see two clear sources of a competitive advantage that AI can enable.

1. The supercharged human in the loop

The greatest opportunity presented by AI lies in the potential for substantial productivity gains through automating non-value-added tasks (for example, submission data extraction or co-pilot driven model development). This will lead to some of an insurer’s key resources, its underwriting and actuarial teams, spending more time on value-added work.

People often think about assumption setting and debits/credits here but there’s a lot more to it. For example, model design (including the choice of techniques and datasets to use) on the actuarial side, and working out which unique "soft" risk data and context is most useful (as well as rapidly ingesting it) on the underwriting side.

The "GPT"- like conversational and multi-modal (i.e. processing far more than just text) nature of AI allows us to do this in a more seamless, natural and human-friendly way than ever before.

2. The unleashed data asset

In an era where AI strips away the advantages from those mundane processes and makes it easier than ever to experiment with and implement better algorithms, where will the comparative advantage lie within an insurance company? On top of those judgments we mentioned already, the other clear area is in our clients’ unique, proprietary data.

Speaking with numerous insurers, one of the most common frustrations that comes up is about their data—the vast majority of which remains untapped. Huge amounts of it get siloed in places that are hard to access or even worse, get discarded.

AI clearly holds the key to unlocking this treasure trove, enabling the industry to harness and refine this data for training and fine-tuning models. Through this process, we can shift the comparative advantage towards insurers who’ve accumulated the most useful data. However, before that can happen, a system is needed to make the capture of this data effortless.

It’s a platform problem

We keep talking about “systems and technology”, but what do we really mean? It’s clear that cutting edge models are systems and technology themselves. Ultimately this is a challenge of building for the future: to win, you need to build on the right platform.

The platform choices made today will significantly influence which insurers will emerge as tomorrow's market leaders. Those that free your time to make uniquely better human decisions, and equip you with the unique data asset to do so, will provide you with an increasingly differentiated comparative advantage over time.

As insurers, we can no longer rely on traditional strategies to drive the next stage of industry evolution. Gone are the days of isolated insurance technology monoliths; now we must embrace the concept of leveraging platforms that eliminate the challenges posed by legacy architectures.

hx Renew as the platform to help you build AI use cases

A core principle that’s guided the development of our industry leading rating and pricing platform, hx Renew, is our commitment to equipping insurers with the necessary tools to seamlessly manage, capture, and leverage data at each step of the workflow. Our primary focus with AI is to empower our customers to harness their data asset, continually enhancing it through seamless data capture and refinement over time.

So, how does hx Renew achieve this?

1. A full development and deployment environment: Providing an intuitive environment tailored to the unique demands of insurance workflows, enabling insurers to build applications effortlessly.

2. Your models, data, and interface all in one place: Establishing a standardized framework for risk data, ensuring consistency and traceability across various stages of the insurance process.

3. A growing data asset of pricing and underwriting decisions: Curating a comprehensive repository of pricing and underwriting insights, facilitating informed decision-making and strategy refinement.

4. Seamlessly integrated into your infrastructure: Offering powerful integration capabilities, allowing insurers to connect with diverse data sources and third-party systems effortlessly.

This, in turn, has helped us become the platform provider of choice for 30+ leading insurers that are on their way to building exciting AI use cases that will benefit their businesses as they see best. We empower them to unleash the full potential of their data, cultivate a competitive edge, and seamlessly integrate it into a landscape rich with AI-driven possibilities—from algorithmic data ingestion to advanced data analysis.

Our platform is designed with the agility and flexibility insurers require to position themselves as market leaders in this new era of insurance.

Get in touch

The cost of waiting at this stage of the technology lifecycle heavily outweighs the benefits that your organization could be accruing with AI. The industry is already starting to see the benefits and returns on investments from AI use cases.

Get in touch with us at hyperexponential. We would love to talk through how your company could be setting itself up for success with the upside promised by AI.