The (in)human touch: AI and fairness in insurance

The (in)human touch: AI and fairness in insurance
People don’t trust insurers. We’ve probably all had to contact an insurer about a problem only to find it’s not covered in the small print. Rates are usually fixed, and sometimes unfair. There may also be hurdles in approving payment once a claim is made.


Such factors can explain why a 2019 YouGov report found almost 70% of policy holders believed providers will do whatever they can to avoid paying out in the case of a legitimate claim. A 2018 TrustPilot report meanwhile found that insurance providers were the least trusted companies in the least trusted field, which was, perhaps unsurprisingly, financial services.

Insurance has a legitimate image problem. Luckily, artificial intelligence (AI) could be the gift consumers are waiting for. Startups around the globe are already stepping up to transform the insurance industry in the same way banking has been changed by fintech, using AI technology to automate everything from risk assessments to auto insurance. This automation is more advanced than that which generates an instant quote from an insurer, or a variety from price aggregators on assorted products. In fact, AI can help insurance products finally keep up with changing lifestyles and consumer demand,. As a theme itself, AI is impacting on sectors and businesses in an aggressive fashion, as noted by analytics firm GlobalData in their recent report on AI in insurance.

But automation could prove to be a double-edged sword for the insurance industry. Those who may already perceive insurance to be unfair to customers may reasonably wonder how much fairer it can get once the human element is completely removed. This fear of cold, faceless calculation could apply to various points of the customer experience, from applying for a quote to making a claim. Those rare insurers savvy to customer perception may worry about the reception to automation. It’d be just as savvy to explore what AI has to offer the industry.

A risky climate for AI insurance

Automation is already commonplace in risk profiling, and the insurance equivalent of fintech, unsurprisingly known as insurtech, is leading the way. Startup insurtech brands like Sweden’s Greater Than have partnered with Zurich on providing AI-driven risk assessment for the global insurance provider. California-based Zesty.ai also joined forces with another incumbent, Aon, to create a wildfire risk product called Z-Fire.

Changing climates necessitate new challenges for insurers, and automation is there to fill the gap. Z-Fire was created in response to recent wildfires across the US, a machine learning solution designed to gauge the risk of wildfires, something which Aon was finding tricky to model. This was mainly due to a lack of detail regarding property characteristics. Z-Fire managed to solve that by using computer vision (CV), in which a machine can accurately identify, classify and react to objects as if “seeing” them. In this case, Z-Fire extracts property data from satellite and aerial imagery, formulating a risk score dependent on weather patterns and building materials, all the way down to vegetation.

It makes financial sense for insurers to delve into granular detail whilst the climate crisis intensifies. According to Aon, Californian wildfires resulted in insured losses of $16bn in 2017, and a further $18bn in 2018. No wonder then that insurance companies will likely spend $3.4bn on AI platforms worldwide by 2024, as forecasted by GlobalData.

Putting the AI in bias?

This all may be good news for the sector, but what about customers? Insurance as a whole is still in debate about AI’s potential for discriminatory bias in risk profiling. Insurance analysts say it could become a regulation issue: See a 2019 circular from a New York regulator which stressed insurers using automation need to establish their methods are “not based in any way” on race, creed and other factors susceptible to prejudice.

Wired has argued how “supposedly ‘fair’ algorithms can perpetuate discrimination,” pointing to how US insurers in the 1960s would “engage in overtly discriminatory practices…while selling insurance to racial minorities”. The fear is that discrimination is baked into the historical data curated for AI to pull from.

Others though have the view AI can help reduce bias; David Moschella, research associate at Leading Edge Forum, told Verdict in 2019 that “algorithms, analytics, and machine learning will, over time, create much more fairness than harm.” In the short term, insurers are strongly encouraged to be transparent about their data and methods. So-called “Explainable AI,” the sort being used by more and more banks to explain loan decisions to customers and regulators, could greatly assist insurers.

A usage-based future

An exciting future for the sector can be seen in what digital car insurer Root is doing. New customers are required to download Root’s mobile app and perform test driving for several weeks while the app monitors driving behaviour. Premiums are then mainly based on the driving score calculated by the app, rather than factors that often discriminate against lower-income customers (e.g. credit scores). This transparency is probably why Root is one of the bigger insurtech brands out there, having gone public in late 2020, raising $724.4m in IPO.

Transparency is a big feature in insurtech and one that incumbent insurers should take heed of in order to appeal to younger customers. Other car startups like By Miles are scoring points with millennials with their Usage Based Insurance (UBI) which charges customers for what they use rather than a flat rate. Too revolutionary? Well, GlobalData analysts note that UBI and behaviour-based policy pricing can actually improve profitability by reducing loss ratios for companies. Zurich are already offering UBI to new UK-based mobility clients, proving that incumbents are also capable of leading the way in the insurance field.

If that isn’t incentive enough, then the recent successes of insurtech point to a revolution on the same scale as banking’s fintech upheaval. KI Insurance, the digital syndicate from Lloyd’s of London, raised $500m last year. Next Insurance meanwhile netted $250m in a recent funding round, doubling its valuation to $4bn.

Ensuring insurance’s future

Insurtech brands have yet to become household names in the same way fintech banking stars like, say, Monzo and Revolut have. Either way, it’s clear AI is already changing the world of insurance.

With better data and better ways to utilise it, insurers can keep one step ahead of a changing world and the new risks it brings, such as with climate change.

A changing world also means changing demands, and insurance brands must keep in mind customer perception and needs. Millennial consumers are becoming less and less satisfied with flat rates and unfair methodologies. Using AI, the insurance industry can solve both problems without losing profit margins. Whilst doing so, companies need to keep accounting for any potential biases in their data, vital to ensuring fairness and objectivity in insurance.

Source: Verdict

Share this article: