As part of our IT Insights InsurTalk series, Future Processing sat down with Artur Niemczewski, Non-Executive Director at the Chartered Insurance Institute, to discuss how AI is both benefitting and causing a few headaches for insurance leaders across the world and looked in detail at how the insurance landscape is being affected by this new technology.

In this article, we take a deep dive into the current state of insurance underwriting and look in detail at what might be achieved through the successful adoption and integration of modern AI and Machine Learning tools and how they will very likely benefit the entire industry as a whole.

The current state of AI tools in insurance underwriting

2017 Predictions

Back in 2017, the insurance press was swamped with doomsday predictions about the possible impact of AI replacing human workers in the near future. At the time, publications such as Insurance Business were predicting that as much as 98.9% of insurance underwriting jobs were at risk of being replaced by AI, leading to some understandably concerned sentiments in the world of insurance.

Present Day

Fast forward to the present day and it is clear that these 2017 predictions of the obsolescence of the insurance underwriter role have yet to come to fruition due to a slower adoption rate of new technologies than many had foreseen.

While AI did make some inroads, the fundamental underwriting tasks remained heavily reliant on manual processes, enveloped within a labyrinthine of spreadsheets.

The current sentiment towards AI technologies has shifted somewhat, with the World Economic Forum now feeling that insurance underwriters are actually among those who have the most to gain from AI augmentation.

While at first glance this is very promising, there is still a way to go before we fully understand how AI augmentation and other AI tools can be fully utilised to realise their potential in the world of insurance.

AI uptake has been a hot topic in insurance in recent years. Even before the public release of generative AI tools such as ChatGPT in late 2022, insurance companies had long been experimenting with their own AI software in an attempt to solve a number of key issues in data handling and analysis, as well as data augmentation in the field of underwriting and risk analysis.

The current issues faced in insurance underwriting

Ethics

Ethics have always been a key issue, with many companies having developed policies to guide them in how they would use the huge datasets that AI models require in order to provide fair, accurate and professional insurance services and offers.

A key question for businesses has always been centred around the usage of AI digital tools in the decision-making process, something that is hotly debated.

It’s crucial to understand how these tools might work, what challenges they might present, what their outcomes might look like and how to balance this efficiency with effectiveness and fairness to the customer.

Judgement

Insurance companies are concerned with the concept of ‘risk’ while insurance customers are concerned with ‘trust’.

Regardless of which side of the fence you sit on, there is a common denominator: judgement. Consumers and insurance companies alike want to know which brokers to trust and what the risk is of choosing one over another.

Until now, there have been numerous methods to make this judgement, which is confusing at best. Before AI, companies would look to create long and complicated spreadsheets filled with data and other information, held together with complex code.

However, businesses have found that when, for example, the member of staff who created that particular piece of code leaves, they are left high and dry, unable to adapt or evolve this code. This forces them to leave it alone and makes it impossible to make any meaningful updates, largely rendering the old code useless.

A lack of integration

As more and more generations of these types of codes emerged, they were successful in creating limited solutions to individual problems but failed to solve the problem as a whole.

The COVID-19 lockdowns pushed the whole industry to use infinitely more electronic data solutions, but this has largely not helped or sped up the underwriting process.

Why? Because these solutions used in isolation do not function well together, they do not integrate.

The problem is not a lack of time and financial investment on the part of insurance companies when it comes to creating new digital solutions, the problem is that there are ‘too many’ solutions on the market, all of which solve a particular problem but do not solve all of the problems that need to be solved.

These ‘in-house’ solutions are not the answer, as Artur Niemczewski explains.

Companies tend to build solutions in isolation; ‘building a workbench over here, an underwriting workbench over there, a data augmentation tools that will pull data from social media and third-party databases over here’, and so on.

But as an underwriter, Artur explains that with multiple data streams coming from all directions don’t get the underwriter any closer to tying them all in together and making a good, well-informed decision.

Multiple separate tools don’t currently integrate, which results in too much time being taken up by simply working out how each tool works and learning what the data in front of him even means!

Companies are making inroads on this solution currently as it stands, but the solution is not quite there yet. Artur explained that during his time at Pro Global, they built a tool based on an earlier version of AI that was seeking to achieve the automatic cleansing of data, but at the time, it wasn’t possible so the solution unfortunately didn’t materialise.

The future of insurance underwriting with AI tools

As we acknowledge the current shortfalls in insurance underwriting when using AI tools and practices, it’s important to have a clear vision for the future.

Currently, there is an ‘arms race’ between around five industry-leading companies who are working on solutions and how to successfully integrate AI data-augmentation tools into their operations in order to revolutionise the industry.

To the winner will go all the spoils, but currently, no one is quite there yet. But what could this ideal future look like?

Artur Niemczewski paints a clear picture.

One tool, one data stream, one solution. The key here is ‘integration’ – all tools should integrate perfectly together, talk together, understand and process their data as one, finally producing a single data stream of information from which the underwriter can make their final decision.

The whole point of integrating AI into insurance underwriting would be to assist the underwriter in assessing the associated risks of each individual case by crunching those hard numbers and offering simple and easy-to-understand results.

AI should do all the ‘heavy lifting’ by effortlessly analysing the data at hand and offering a very clear result.

This could be as simple as a traffic light system – ‘green’ could indicate the low-hanging fruit that offers low risk and a high chance of successful policies. ‘Red’ could indicate high-risk situations that the underwriter should steer clear of.

However, the ‘amber’ situation could be the most interesting…

Here, the AI data augmentation analysis could flag the underwriter to risks that are filled with nuance.

It is in these middle-ground decisions that human underwriters could involve themselves by studying the data presented in order to arrive at a decision on whether the risks are worth investing in or not.

Armed with all of the necessary information, they could study the case and make a good, informed decision.

The AI has done all of the heavy lifting in siphoning out the green ‘low hanging fruit’ and warning against the red ‘no-go’- areas, leaving the amber analysis clear and comprehensive for a human to make a final decision on.

All datasets and systems would have been crunched and analysed by the AI, allowing the human underwriter to spend their valuable time judging the nuanced situations where the high-value decisions lie. The AI would have already taken care of all of the areas which were previously difficult for humans to navigate due to the complex errors brought up by non-integrated systems, leaving them worry-free and focused on real value-driven tasks.

Summary

The problem is not investment, but rather integration.

Many companies are heavily invested in pursuit of this type of solution but the answer will lie in how well each individual system integrates into the next, and how well they can harness the power of AI in order to arrive at a single source of data to make a well-informed final solution.

With this ideal solution on the horizon, the key question remains: who will get there first?

If you would like to watch our full interview with Artur Niemczewski where we discuss the current state of AI in insurance underwriting and his vision for the perfect future, please follow this link to visit our IT Insights Hub.

About our guest

Artur Niemczewski has a huge wealth of experience in the insurance industry gained over a 25-year career at the highest levels.

Starting out, Artur qualified with a PhD in Nuclear Engineering and a Master’s in Public Policy of Technology from the Massachusetts Institute of Technology (MIT) in the USA. From there, he transitioned into insurance as a broker and enjoyed many varied positions in the industry such as Head of the London Market Operations for Willis and Chief Executive at Marsh, a branch of the London Market Speciality Division.

Now working as the Non-Executive Director at the Chartered Insurance Institute, Artur concerns himself with the current state of insurance in this ever-changing world and works towards developing and implementing modern AI solutions to help drive the industry forward in an attempt to revolutionise the use of technology across the whole insurance world.