Download our latest State of Pricing Report here

Written by Jamie Wilson

AI for Actuaries: Three key use cases

Pricing

4 minutes

Jamie Wilson, Head of Pricing and Innovation at hyperexponential, on how AI could unlock new possibilities for actuaries

According to the 2024 State of Pricing report, 50% of insurers are already investing in AI technology. Only 9% have no plans to invest within the next 5 years.

Coming from an actuarial background, including time spent building out and leading the predictive analytics team at Allianz, I've been exploring the implications of AI in insurance for many years. For those familiar with my views, you’ll know I’m not a blind optimist. While AI holds immense potential, much of it remains just that—potential—and the current state of AI is more constrained than its biggest proponents might suggest. I also recognize that many insurers are still grappling with technical and data challenges, which are essential for maximizing the benefits of AI and ML. Fortunately, these are issues most are actively working to address.

That said, I’m genuinely excited about the possibilities AI and ML present for our industry. These technologies will empower insurers to maximize existing resources, harness previously untapped or underutilized data, and automate time-consuming tasks, to allow for more focus on value-driven work. Actuaries will be able to develop more robust pricing models, and ultimately drive business success.

Even better, the barriers to entry for these innovations have never been lower.

In this piece, I’ll explore three key ways Generative AI in particular can help pricing actuaries overcome long-standing challenges.

1. Accelerate actuarial productivity to focus on building more effective pricing tools

When it comes to accelerating actuarial productivity, one of the key ways AI can help is by making the coding process faster and easier so we can focus on building more effective pricing tools.

For many actuaries, coding can be a steep learning curve, and there's a real concern in the industry about not having the right technical skills—83% of actuaries worry about not having the right tech skills for the future.

The solution with AI

This is where AI steps in. For those who aren't as confident in their coding skills, AI can help them get up to speed, both writing code for them or acting as a code reviewer to the first attempts at a new language. This significantly speeds up the historic process of try, debug, ask stack overflow! It can also help actuaries translate code between languages such as R to Python, making cross-platform work easier and more accessible.

Even for seasoned developers, AI-powered co-pilots can take productivity to the next level by managing simpler coding tasks, allowing them to focus on solving more complex and strategic problems. It’s like having an assistant who takes care of the groundwork, so you can dedicate your energy to innovation.

That’s not to say learning to code or understanding your technical frameworks won’t remain important. AI is there to assist, not take over the entire model-building process, and that’s not likely to change any time soon. But as an accelerator and enhancer of your workflows and knowledge, it can provide a real edge.

The possibilities in this area really gripped me when I first started experimenting with language models. While at a previous company, we ran a test with ChatGPT to see how well it could generate predictive modeling code with minimal input.

The results were impressive—strong, functional code in both R and Python, with only minor tweaks needed. It still required oversight from someone who had a good grounding in Data Science to ensure it wasn't 'hallucinating' methods (which it definitely did when pushed to more cutting edge approaches)—but it showed instant potential for speeding up some of the more mundane coding work.

The real surprise came when we asked it to add some humor to the code commentary, and it delivered both solid code and genuinely amusing comments linked to the nuances of various Python packages. That may not be a key use case for AI in the insurance industry, but it made for a fun afternoon!

2. Equip underwriters with more sophisticated models to improve pricing insights

Building the strongest possible models is one thing; ensuring that they are used, and used to their best effect, is another.

One of the key challenges for S&C pricing is that underwriters either don’t trust the pricing tools available or don’t use them effectively, leading to poor adoption. According to the 2024 State of Pricing report, less than a third of underwriters consider pricing models to be “essential” and use them every time they write a risk. 41% rely on these models only when they trust the output, 21% view them as a 'necessary inconvenience,' and 8% say they 'don’t trust them' because the models are often ‘outdated and inaccurate.’

Finding ways to leverage AI to improve the underwriter experience with Pricing Tools and provide greater insight at the point of pricing is a great way to address this risk of poor adoption.

The solution with AI

AI can help by:

  • Improving risk assessment by analysing unstructured submissions. With AI-powered tools, everything from photographs to scattered PDFs can be incorporated seamlessly and automatically into your models, and your pricing workflows.

  • Allowing underwriters to delve deeper into the context of the risk they’re pricing with natural language. For example, an underwriter could “ask” their pricing tool “What rate change have I achieved on this account for the last 4 years?”. Presenting this information is something that can be solved for today without AI, however it requires the Pricing Tool developer to build out that insight and interface. AI could enable underwriters to access this insight without increasing the developer burden on Actuaries or IT.

  • Automatically providing points of comparison. AI-powered tools that automatically surface the most relevant information are an effective way to enhance pricing workflows. Flagging “risks like these” is one example of how this can look in practice.

3. Data preparation

Data preparation has long been one of the most time-consuming tasks for actuaries. Issues like incomplete or inconsistent data, siloed systems, and manual data handling can create significant bottlenecks.

It’s vital work—portfolio analysis and predictive modelling both rely on clean, well-structured data—but actuaries are often left spending more time on data wrangling than on meaningful analysis.

The solution with AI

AI can streamline this process by automatically extracting and cleaning data from both structured and unstructured sources, such as claims descriptions, loss notes, or even external datasets.

This can aid in merging claims with exposure data for modeling purposes. AI can also generate peril-and segment-specific insights that are often missing from traditional datasets.

This type of work would historically have required a summer intern to come in and manually work through multiple claims descriptions, policy files etc or for a modeler to spend significant amounts of time writing code to parse the various data sources. Simplifying this work is a significant win and allows Actuaries to more quickly move onto the fun part of the work—the analysis.

Ultimately, leveraging these tools can give actuaries richer data to work with, faster and with fewer manual interventions.

Further applications for AI

Pricing model development is just one facet of AI for actuaries. Accelerating productivity across all work responsibilities

Here are just a few of the possible use cases:

  • Explaining actuarial techniques. Actuaries who want to develop professionally, stay on top of industry shifts and embrace new techniques—here they can benefit from using AI to simplify complex actuarial methodologies. Leveraging LLMs to generate summaries of technical papers can make new methods easier to grasp.

  • Day-to-day administrative tasks. AI can assist with routine tasks such as drafting emails, conducting peer reviews during busy periods like year-end, or even writing actuarial opinions, freeing up time for more strategic work.

Next steps

AI seems to be sparking excitement and apprehension in equal measures. From my perspective, actuaries who can leverage AI will find themselves coding faster, summarising complex analyses faster, communicating more effectively and building more tools and applications that would have been impossible 5 years ago.

Far from being replaced, this means they will be adding significantly more value to the business.

While AI is easier than ever to get started with, the competitive edge is going to belong to actuaries who best understand and leverage it. Embracing training opportunities, staying on top of industry news and innovation, and building for the future will be key.

To learn more about the foundational requirements for AI and ML (and how hyperexponential helps you meet them), get in touch with us here.

Loading...
AI in hx Renew