Why AI in Insurance Improves, Not Replaces, Human Decision-Making

Why AI in Insurance Improves, Not Replaces, Human Decision-Making

In Uncategorized by Roger Lewis

Full article with thanks to: forbes.com/sites/adrianbridgwater/2024/09/01/insurance-technology-how-insurtech-is-assured

Tech is rich in blends. The use of portmanteaus and lexical blend word splintering is as prolific in technology as it is in show business (e.g. Brangelina and other fusions) with terms like DevOps being among the most well-known (developers + operations teams as a unified single entity) pieces of terminology.

Aside from all the Ops extensions (FinOps, AIOps, SecOps etc.) there are the industry-specific connections where we add “Tech” onto what is often a shortened version of a business discipline – hence MarTech (marketing technology), FinTech (financials), GovTech (government, obviously) and perhaps even the potentially non-specific AutoTech (for automotive manufacturing), although the latter could arguably apply to any use of automation.

Then there is InsurTech for the insurance industry.

How InsurTech Developed

Chief technology officer at on-demand embedded insurance services company Inshur is Chris Gray. Explaining why he thinks many InsurTech organisations have a bad name thanks to years of proclaiming technological innovations that can produce automatic claims payments with AI, with a focus on paying out cash to claimants in under five seconds, Gray says that loss ratios have risen sharply, making the insurance industry anxious about working with them.

NOTE: As defined by Investopedia, the term “loss ratio” is used in the insurance industry to represent the ratio of losses to premiums earned. The investment portal notes that losses in loss ratios include paid insurance claims and adjustment expenses. The figure itself is worked out via the following formula: insurance claims paid plus adjustment expenses divided by total earned premiums.

“The issue is that InsurTechs are failing to understand the ‘insurance’ element which is leading to policy pricing inaccuracies,” said Gray. “This is resulting in a mass exodus from reinsurance partners and, although the technology may be working wonderfully, without insurance capacity to pay out on claims InsurTechs only have a swanky tech platform to offer.”

In the niche that Inshur operates in – commercial auto insurance for on-demand drivers in big cities – capacity issues in the insurance industry make this a challenging environment to operate in. The company says it has over 40 years of loss ratio data specifically for fleet, taxi and delivery drivers meaning it understands on-demand driver demands. It is working to develop new rideshare and courier insurance products.The video player is currently playing an ad.

The Future Is On-Demand

“The future is on-demand. The way we access services like taxis and how we purchase our groceries and pizzas has changed forever. Incumbent insurers need to adapt their models to become more flexible and embed insurance products into the platforms used by drivers. If they don’t, more nimble insurance players with complementary technologies will enter the market and satiate the seismic demand,” suggested Gray.

The global on-demand economy has created the most profound economic shift in four decades and, according to PwC research, it is expected to surpass $335 billion USD by 2025.

Out With The Old Breed

The Inshur team say that the old breed’ of InsurTechs have burnt too many insurers’ fingers with their focus on growth at all costs, using AI-first pricing and claims handling as a way to lure in customers and capacity partners. It seems that through trialling the technology to automate pricing and claims, these InsurTechs completely forgot that insurance is part of the financial economy and therefore requires specialist knowledge and data in order to automate hundreds of years of insurance experience. In this industry, there is a great deal of data that needs to be managed with care and diligence – from PII, to health (in claims) data, to financial information.

“Because of the sensitivities, we have focused on the technology, data, data models, databases and the fundamental requirements of insurance and claims handling to build a platform that is viable not just for insurance, but for platform partners and drivers,” said Inshur’s Gray, speaking to press and analysts this September in London.

AI-Augmented Assistance

The Inshur platform makes use of artificial intelligence and machine learning primarily as an augmented assistant rather than a replacement for insurance expertise, such as ID verification, fraud detection and assistance with claim triage and handling. It embeds insurance into applications so that it’s accessible for drivers.

“We listen to our insurance team and use the technology we have to benefit their situation. For example, our claims department needed assistance to deal with the magnitude of incoming claims and how to prioritise them, so we built an AI assistant that summarises each claim and its current status, categorise it into type of claim (vehicle, personal injury etc.) and then prioritise the claims for the claims handler to manage based on a variety of proprietary factors such as recent communication with the claimant and other parties involved in the claim. AI complements our team’s day-to-day and enables them to do their job more effectively,” explained Gray.

The company reminds us how important it is in this market to be global. A scalable software stack in this sector should be capable of being implemented globally whilst meeting local insurance regulations and policies – particularly if we are thinking about scaling across all 50 United States, or indeed any other country in the world.

How InsurTech Actually Works

On-demand commercial auto insurance requires a myriad of insurance data, such as location, weather, vehicle type, how the vehicle is used, where it is parked, miles driven, hours driven, driver history, driver work location, driver insurance claims etc. A good platform should also use data from telemetrics to assess driver safety and speed for instance, alongside data from the apps on-demand drivers use. Then there are associated biases that need to be factored in to ensure the technology is supporting the underwriting team to issue fair policies for on-demand drivers and the platforms they use.

“Let’s take the heavily regulated US market as an example. Many insurance products operate in the ‘admitted’ space which essentially means a state’s regulator signs off your pricing and underwriting – and is generally resistant to subjective-based pricing, which makes it almost impossible to add AI to the pricing side of the equation,” said Gray.

To address this, Inshur instead focuses on using machine learning to help refine data models before they’re used in real-time. For example, it models data inside Google Big Query using AutoML as part of its pricing strategy to identify pricing factors, such as historical driver behaviour, environmental or geographical factors and seasonal or temporal factors, that it may have not spotted before. It also helps identify trends with fraud and higher claims volumes. These insights are analysed by an actuarial team to allow them to apply their experience to adjust prices and underwriting criteria, as well as remove any biases.

Inferred Location Data

“Gone are the days where you fill in 100 questions to get a price. We work very closely with our embedded partners such as Amazon and Uber to automatically gather bespoke data about our customer’s driving experience – for example, with Amazon we have access to information about our customer’s block bookings and the shifts they work. This allows us to combine claims data, inferred location data as well as information provided by the customer to ensure complete coverage, as well as price the risk fairly for all parties,” concluded Gray.

With all the digitisation happening here – and with Inshur reinforcing its stance on on-demand automated technologies as the future – when the company’s AI engine offers some advice to a person dealing with the policy or claim, it ensures that its advice is a recommendation and not a decision. Highly trained claims handlers make the final decision meaning that the AI tools are used to enhance and help rather than to control. This is also complemented by utilising Google Explainable AI frameworks which helps understand why a decision has been made and ensure that as much bias is removed from decision making.

That’s some comforting validation from the specialists working in this field perhaps i.e. when we do all start buying all of these types of services in a completely digital and automated form, at least we have human operatives at the centre of the (final) decision-making process. InsurTech is assured to be with us going forwards… and it feels like the shift to on-demand in-app services being underlined here is also validated.

Full article with thanks to: forbes.com/sites/adrianbridgwater/2024/09/01/insurance-technology-how-insurtech-is-assured

Did you enjoy that? Why not share this article.

We specialise in supporting change across the insurance sector, helping insurance businesses (and other businesses like distribution and logistics) adapt seamlessly.

Our services include Change & TransformationM&A IntegrationBusiness Process Management and Executive Coaching. We help make organisational change and technology change much easier.

Got a question about our services? Get in touch and leave a message.