- Apr 18, 2023
- 15 min read

Navigating AI usage in the insurance sector

Updated: Sep 14, 2023

“Have you heard about ChatGPT? It’s amazing! It can generate any type of content about any subject after just a few instructions, you’ll LOVE it!”

“Why would I love something that overrides the years of experience in my profession, and will no doubt see me out of a job?”

This exchange is a common conversation that is happening across the world in relation to AI’s hottest new tool. It appears everyone has a love or hate opinion.

While ChatGPT is generating extensive amounts of media coverage (and Google’s launch of Bard is sure to increase the chatter) and public opinion let’s not forget we’ve been surrounded by AI for years. Remember when we started noticing how social channels such as Instagram started filling our personal feeds with images and videos because we watched a five second clip of a cute dog? Not an issue when you like looking at images of cute dogs, much more of one when you only watched the first one by accident and have a fear of them.

Whether you personally ‘opt in’ to the benefits of AI or not, there’s no doubt it has a big influence in our lives. All ‘it’ wants to do is make our lives a little easier and help give us more of what we want, or what it thinks we want (depending on which way you look at it.)

Image Source: Unsplash

But AI is much more than receiving recommended songs after you’ve finished an album on Spotify. It’s not only making huge advances across our personal lives but across the business world too. We know that AI tools and systems can mine vast amounts information and return it back in a way that the user needs delivering notable benefits such as:

· Reducing human error

· Automating manual skills and processes

· Applying known data in new ways

· Learning new data and finding patterns in behaviour

· Streamlining processes

So, how can AI usage in the insurance sector work? In this article Procurato explores how insurers are currently using AI and takes a peek ahead to see how it could be used in the future.

From Chatbots to risk assessment….

Back in the late 1990s / early 2000s GPS became commonplace, and that advance made telematics possible. Insurers and brokers believed it was the technology of the future, a ‘black box’ for drivers encouraging them to drive more safely and be rewarded for it. A tool to provide added-value concierge services to help manage the activities of fleets and essential-use drivers. It didn’t and, to a great extent, still hasn’t had the predicted impact 25 years on. However, the continuing advance of technology means it’s now not only about keeping to the speed limit and taking a few pounds off a premium. Consumers want more personalised offerings, behavioural-based telematics if you will. And insurers can and are adapting to that. The point is, that what was once the darling of insurance technology is now a little passe; the world has moved on and insurers need to move with it. But in the background, for those promoting and using telematics, artificial intelligence is very much at the core of the product – the algorithms are churning the collective and individual data to ‘recommend’ a driver score, premium, course of action – to nudge the driver, to warn the underwriter.

This example is just one of how AI has been used in the insurance sector over the last few decades and as technology continues to advance so does its uses.

Underwriting

In years to come will personal lines underwriting operate in the way it does now? According to McKinsey’s Insurance 2030: The impact of AI on the future of insurance report:

“[In 2030] The number of agents is reduced substantially as active agents retire and remaining agents rely heavily on technology to increase productivity. The role of agents transitions to process facilitators and product educators. The agent of the future can sell nearly all types of coverage and adds value by helping clients manage their portfolios of coverage across experiences, health, life, mobility, personal property, and residential.

Agents use smart personal assistants to optimize their tasks as well as AI-enabled bots to find potential deals for clients. These tools help agents to support a substantially larger client base while making customer interactions (a mix of in-person, virtual, and digital) shorter and more meaningful, given that each interaction will be tailored to the exact current and future needs of each individual client(1).”

They’re bold statements but in our opinion the likelihood for them to come true is high. Insurers first used machine leaning (ML) for underwriting in the 1980s. ML’s ability to make predictions using customer data started to revolutionise insurance risk models. But it would be around 25 years later before AI came into its own in UK financial services, most notably enabled by the rise of aggregator services.

In the UK, consumers were quick to embrace the use of aggregator channels and by 2009, they accounted for more than half of total private auto insurance sales in the UK and 36 percent of home insurance sales(3). Once deemed as a potential challenge for insurers this approach is now firmly embedded in the omnichannel experience. Customers benefit from the level playing field and ease of transaction whereas insurers benefit from much bigger reach of potential new customers and better conversion rates that ultimately can reduce acquisition costs.

We have previously talked about labour shortages and fewer people entering the underwriting profession. AI addresses these challenges when underwriting less-complex policies. Insurers hold a vast amount of information on their systems about individual policyholders and the use of AI enables them to create risk profiles which allows the purchase of personal lines policies to be completed in a short space of time. In addition, extras can be added on at the point of purchase further adding data to that profile.

It’s also important to recognise here there are a number of terms – Big Data, Artificial Intelligence, Machine Learning, Predictive Analytics – which are often casually interchanged, when they are quite different things. All share a simple core principle, most precisely articulated in Bill Gates’s recent seven-page ‘letter’, “The Age of AI has Begun(11)”:

“Compared to a computer, our brains operate at a snail’s pace: An electrical signal in the brain moves at 1/100,000th the speed of the signal in a silicon chip! Once developers can generalize a learning algorithm and run it at the speed of a computer—an accomplishment that could be a decade away or a century away—we’ll have an incredibly powerful AGI. It will be able to do everything that a human brain can, but without any practical limits on the size of its memory or the speed at which it operates. This will be a profound change.”

The key point being that we’re not talking about replacing workers with automation, we’re talking about capability already in existence which can do things people would never, realistically, be able to do. The AI already in play today has made a leap equivalent to the difference between horseback and air travel.

What the AI of today / tomorrow will do is take the tools at the hands of the insurance sector to supersonic levels. Rather than spending hours trawling data before applying their expertise, insurance technicians can spend more time on the more complex tasks (that they prefer to do) while adding value to their business.

Risk and the claims process

The ultimate beauty of AI is its power of insight. Insurers already capture a significant amount of data – some argue they are more akin to knowledge-management businesses than financial institutions. They know their customers’ demographics; they know if they live in an area with high levels or crime or a higher risk of flood risk. They know their customers’ claims history; they know whether the jobs they do may have an impact on their policy. They know that someone who has a dog may be less likely to be burgled but could be more likely to have accidental damage.

There is so much more that AI can bring to the table. With its power to analyse multiple data points in the complex real-world AI models are able to do a better job of predicting outcomes. As well as the structured data that comes from insurers’ in-house systems AI can also analyse unstructured data which we class as those that don’t fit neatly in the system fields. This unstructured data could include emails, video files, audio and dash cam footage. Assuming that the models are built to understand the data (see AI challenges later in this article) AI can prove invaluable to predicting insurance risk.

There is no doubt that being able to predict insurance risk will improve decision making.

It can also boost productivity through more efficient modelling and these models should reduce the frequency and severity of losses which, in today’s competitive aggregator market, is where organisations are going to achieve financial advantage.

Protection from financial loss

In our recent update on home inflation we advised that insurers are telling us that fraudulent claims are on the rise. While there is no indication that policyholders are deliberately damaging their possessions there is data to suggest the value of the claim is exaggerated. This is not a new trend.

Aviva reported 11,000 instances of claims fraud worth more than £122m in 2021 and at the time of writing their report 16,700 claims were under investigation for fraud(4). Whiplash fraudsters are switching to vehicle repair claims and home owners faced with a cost-of-living crisis are exaggerating their claims.

AI is increasingly being used to help insurers detect and prevent fraud. From analysing social media and scanning images to reviewing data from previous claims AI can analyse data to flag a potential fraudulent claim giving call centre operatives an opportunity to probe further and may even adjust pricing for future policies.

Natural disasters – past experience is not the predictor

In April 2022 the ABI announced that insurers could expect to pay out nearly £500 million to support customers hit by damage from Storms Dudley, Eunice and Franklin(6). Later in the year subsidence claims started to increase following the unprecedented hot weather spell in the UK. Climate and natural disasters continue to have an impact, and this isn’t going away any time soon.

Unlike many other risks historic data isn’t often a good predictor for climate events. Who could have predicted 40-degree temperatures in the UK on the back of those storms earlier in the year? Climate change is altering the world’s weather patterns and to combat this many insurers are turning to InsurTech firms to find alternative solutions to better predict the weather and model the risk accordingly. According to PWC’s State of Climate Tech 2022 report climate tech funding in 2022 represented more than a quarter of every venture dollar invested that year7 so it’s being seen as a priority by the big players.

Thanks to the predictive qualities that AI and machine learning has, we are seeing much improved models to predict natural disaster events. According to TechTarget(8) ‘traditional physics-based models are still outperforming AI models, but they operate at tremendous computational cost that AI models can mitigate when used correctly.’ They also cite AI is already being used to prepare for disasters using aerial imagery technology, location data and satellite imagery to look at building and roof characteristics, surface permeability, roof condition and nearby construction and building footprints. This technology allows users to see ‘predictive insights like levels of tree cover in proximity to buildings that could result in increased vulnerabilities caused by fire spread or damage from fallen trees.’

These models can analyse building conditions over time to see if for example a roof can withstand a weather event or if it should be replaced. Being able to predict potential damage before it happens could be beneficial to insurers who can add this as extra coverage when underwriting a policy. Although parametric solutions are already well-established insurers should continue to explore emerging technologies and scale them because as climate-related events increase so will the pay outs for customers and clients.

However, insurers may do well to consider the recent technology developed by the Met Office and Bristol University. They’ve developed climate prediction and flood models which they are ‘100 per cent certain’ where and when heavy rainfall will occur It’s a ground breaking flood model and combines historical risk and events while combining them with climate model projections. These predictions have been analysed against ABI flood loss data and hold up well. As insurers will know, where there is certainty, there’s no risk so who would underwrite a location which is guaranteed to make a loss. To use an analogy, it’s the equivalent of insuring a vehicle you know is going to crash. Insurers will have tricky decisions to take to support customers in the face of technology such as this.

A step past telematics

The so called ‘In home’ Internet of Things (IoT) brings many opportunities to insurers through personalisation, improved risk predictions and prevention.

In a similar way to how insurers offer telematics solutions to drivers they can offer connected devices for the home which can monitor its security and safety. Should an issue arise the IoT system would alert the insurer which would allow them to step in to prevent that damage, providing a safer home for the customer and less risk for the insurer. Traditionally insurers do provide better premiums to policy holders who have better security of their home and vehicles but if this were to be expanded not only would customers benefit from cheaper policies they’d feel more protected by their insurer.

We’re already seeing that the cost of IoT devices is falling as their use become more widespread. If insurers were to offer the device in conjunction with their policy this could be seen as a competitive advantage and will lead to less claims.

As with so many of these “opportunities”, one of the challenges for insurers is to have enough time and expert resource to know what to do with them. It takes a lot of thinking about, which is exactly why insurers should partner with new technologists.

Image Source: Unsplash

Customers conversing with bots

One of the most common uses of AI in the insurance sector are chatbots. It’s not uncommon on any website for a chat box to appear with a chat bot asking if they can help. It’s an easy technique to reach a customer or potential customer within seconds of them reaching your page; you don’t want lurkers you want sales.

While 60 per cent of consumers say they prefer to speak to a real person when dealing with a query5 a significant number are happy to speak to a bot when they need a simple question answered. Bots are a perfect solution to give answers to simple, non-urgent requests and to provide them out of office hours. Insurers have been using chatbots for considerable time; it’s freeing up the time of customer service representatives who answer many similar FAQs every day. But what else can a chat bot do in the insurance world?

· Automatically process claims.

· Explain insurance plans in simple terms.

· Provide cover policy details if lost.

· Automatically qualify customers.

· Assist customers with insurance payments.

· Receive feedback on plans and customer service while simultaneously measuring the sentiment of online discussions which can provide ongoing feedback.

· 24/7 availability reducing time needed in call centres.

Further support for contact centres

Aside from ChatBots AI offers many more benefits in call centres. Research from Qualtrics(9) has indicated that 63% of consumers said companies need to get better at listening to their feedback with 60% saying they would buy more if businesses treated them better. AI can enable this by:

· Understanding a customer’s full history at the touch of a button: Customers expect their insurers to have access to all their previous correspondence and don’t expect to have to repeat themselves to multiple agents. AI can scour huge amounts of data to bring together one version of the customer truth.

· Providing empathy through natural language processing: Some AI solutions have the ability to understand voice and text even if the customer is not speaking in their first language. This enables call centre staff to understand the emotions of that customer and can even provide suggested commentary. This along with understanding a customer’s history can be further enhanced via the provision of personalised real-time scripts for the call handler.

· Providing quicker responses: Customers can be frustrated when they don’t hear back in suitable time on their correspondence. AI can enable a receipt notification and subsequent ones such as being read, understood and who it has been re-routed to.

In addition, AI can provide business support in the contact centres through:

· Automating coaching and quality management AI in the call centre can automatically track agent performance, successful resolutions, and script compliance. With that information, AI-powered contact centre software suites can offer up real-time insights on exceptional performers and where there are coaching opportunities.

· Providing new insight. With AI tracking 100% of all customer and agent interactions it provides even more insight which can benefit the business. This could be through determining new trends or requirements or by highlighting any issues.

· Improving workload and reducing employee burnout. If AI can take away the administrative and less complex tasks away from employees it will free up their time to focus on important tasks that add benefits. This in turn is also likely to improve employee satisfaction.

· Optimising supplier relationships. One client we have spoken to has shared their experience of testing AI is to better support opportunities for salvage through improved keyword use and searches extracted from insurance reports which should identify more accurate salvage detail and prospects. This improves responses and creates less unnecessary land fill.

Procurato recently conducted a piece of original research on behalf of a major Insurer client who wanted to understand what their broker customers actually wanted. Over five months we reviewed the market, interviewed many brokers, talked to other industry experts and the outputs were reassuringly straightforward:

· Keep it simple and make it easy for me to use i.e., authentication, interface, recording of chats.

· If you’re going to do chat, make sure it works for me, and it doesn’t just put me in a different queue from the phone queue you’re replacing it with.

· Make your digital engagement do what I’ve been asking you to do for years; if you give me bells and whistles but I can’t access the data I need, send the files I want, or update the records I’m working on, I won’t use it.

Of course, we learned more richer insight into what people will and won’t use, what is good and bad, and what is essential versus optional. But the key point is the automation or digitisation has to transparently add value to the current position, not simply move the problem.

AI in insurance – challenges

It’s clear to see there are many benefits to implementing AI in whichever form to a business but with so much available it’s important to be able to cut through what is potentially ‘hype’ and what is really needed to achieve an organisational strategy.

A challenge we often hear from CFOs is the pressure to balance the significant investment of these (often still) emerging AI technologies against the potential payback. Many technologies have not been proven and insurers must decide whether to get ahead of their competitors or risk poor spending decisions.

Investment is just one issue, there are other considerations when choosing an AI tool:

· Aging IT systems can stand in the way of delivering digital solutions for its customers and employees meaning investment is not just about the new tech but the existing.

· Bias and Ethics. Analytical models need to be built and ‘trained’ to find relevant data. If it doesn’t this can lead to biased predictions or outcomes particularly if the models learn from human-labelled data. Most of the regulation in place today for insurers is founded on fairness, which needs to be evidenced transparently and there needs to be accountability for data bias. Understandably, those personally regulated individuals are nervous about entrusting their reputations to computers.

· Weaknesses in the data AI understands. While AI / ML models can turn complex structured and unstructured data in actionable insight the data needs to be current, accurate and well maintained. If a model is built with poor quality data, you can expect to see poor quality outputs.

· Data security and fraud. Cybersecurity attacks can be more prevalent when accounts are linked. This could be in the form of connected devices which could put not only your business at risk but also your consumers. Additional care must also be taken when integrating with your customer database.

· Customer concerns: Some customers may be reluctant to share their personal data with insurers particularly when using connected devices. They will also want reassurance that their data is only used for the purposes that it’s intended for. Breaches of data privacy can lead to huge fines and loss of reputation for an organisation.

Image Source: Unsplash

AI in insurance – the future

Looking back on AI’s history it’s unlikely we’d ever predict where it is now compared to where it was when it started and the same applies if we look too far into the future. But we are seeing emerging technologies.

· The Metaverse. Consumers are already immersing themselves in the metaverse when buying property, cars, furniture, clothes and artwork. We have also seen the rise of digital properties such as non-fungible assets (NFT). All of the above have insurable value and insurers will need to adapt.

· ChatGPT and Bard. Still in their relative infancy for public use ChatGPT and Bard (Google’s version) is limited in some abilities such as handling multiple tasks and not being able to understand some human context such as colloquialisms or sarcasm but there are some potential uses. In addition to being able to simulate conversations or providing realistic helpful responses, both are designed to augment existing search facilities.

Roi Amir, CEO of Sprout.ai suggests that the technology could be used by insurers to start their own innovations, “[Chatb is a great generalised model. What I think we will see is many companies using it as their baseline model and add layers of intelligence and specificity for their individual domains(10).”

· Increased prevalence of physical robotics. Autonomous features are playing an increasing role in technology – self-driving cars is one such example. Insurers will need to understand how the presence of robotics will shift consumer expectations and enable new products and channels to match them.

AI is continually evolving, and insurers should be willing to evolve with it. It brings tremendous benefits but not without its challenges, so why not speak to Procurato about the best solution for you.

(This article was written by a person)

--------------------------------------------------------------

The Procurato Way

Procurato’s consultants have over 30 years’ experience helping clients to navigate their way through emerging technologies to ensure they choose the right tools and systems to meet their objectives and gain competitive advantage. We know how to run robust tenders, selection and contracting while also having the ability to move quickly so that client needs are efficiently met when they need them. Our team of experts can work independently to free up resource, or they can work in conjunction with your teams to augment their knowledge.

Please contact us if you would like more information or support in identifying the right technology tools for your business.