Machine learning and the future of insurance – the debate

What are the implications of the use of AI in the insurance industry?

A.CG: The first and most obvious use of machine learning in insurance is that risk and claims will be calculated by software and decisions to insure or to pay out will be made by software - software, not people - and at least many holders of policies will feel that they are being treated mechanically without due regard to the human nuances of their circumstances.

NP: What will be interesting over the next 6-12 months will be AI’s compatibility with regulation. The main issue that insurers will have is how heavily regulated the industry is. Regulations will only increase (with the introduction of AI), we also have the Insurance Distribution Directive (IDD) coming in later this year. The senior manager’s regime will be coming into the insurance space by the end of this year and extending into early 2019. The FCA remarkably said that they had planned to introduce more regulations this year, but they put it on hold because of Brexit, which shows that there are no signs of them slowing down in terms of regulation.

At the same time, you have this technology that is becoming more intuitive and yet in industries that are heavily regulated it presents a quandary because you want to leverage that technology as much as you possibly can, particularly with all of the cost pressures people are under at the moment, but there is a big question mark over how it could ever be fully compatible with GDPR for example.

When you develop a machine learning product that is so intelligent and designed to go beyond what it has been created to do, how can you ever audit that fully? I don’t think anyone knows the answer to that.

There is also an issue of governance, one thing that will undoubtedly become commonplace is the big insurance companies developing ethical data boards. Which Microsoft and IBM are already looking into – the board will effectively fulfil a wider function than a standard data protection officer and would be independent of the C-suite which would mean they can’t be governed by the C-suite. They ultimately will carry ultimate sway over what happens with respect to data processing and data protection activities. This will also come because of a greater demand for data scientists within the market, the demand is already there but it will exponentially increase as the deficit increases.

Does machine learning need to be handled with a degree of caution?

A.CG: For the above reasons, yes. The AI has to be extremely sensitive to the very nature of human experience and happenings if it is going to command the respect of potential clients.

NP: Definitely. It has huge benefits, but as the saying goes “with great power comes great responsibility”. Wielding something like AI without caution will lead to a lot of problems down the line. I don’t think the insurance industry is using this technology indiscriminately, they are far more cautious. I think there is a middle ground, but they are right to be cautious, particularly if it is a regulated industry where there are customers involved who could suffer financially. 

Will AI help overcome the human propensity for corruption (i.e fraud)?

A.CG: It can certainly help, for it will very probably be far better at identifying prompts for a closer look at e.g. claims which might be fraudulent.

NP: The positive uptake on AI that I have seen within the insurance industry is for fraud protection. A great example is Shift technologies who are the UK leader as far as fraud protection and prevention is concerned and they have grown exponentially in the last year. They use artificial intelligence and huge data sets to detect patterns and flag fraudulent insurance claims. It is a software-as-a-service platform designed for insurance companies.

Could this lead to the eradication of job roles?

A.CG: Almost certainly yes, and across a wide range from clerical to actuarial to adjustment etc. However, the main issue is that real human problems and needs will be overlooked or wrongly calibrated because algorithms have a tendency to be too sharp-edged and yes-no.

NP: 12 months ago, I would have said the profession most at risk of disruption were Brokers but actually, where we are now I feel that it is underwriters. The reason being that when you boil it down an underwriter is performing an actuarial calculation, and if you look how AI processes large banks of data using algorithms in a matter of seconds, compared to the hours and days that it normally takes now.

I do think the underwriting market will be affected and Brokers will have to be a lot broader in terms of the services they provide. This means taking a proactive risk manager role to their clients as opposed to just placing insurance cover.

Who would regulate it?

NP: The FCA is technology neutral, so they will regulate the outcome. There will have to be an oversight board set up by the government, whose function will be to be responsible for the regulation and the development of AI. This isn’t just relevant to insurers but for many other industries using machine learning.

Why should delegates attend?

A.CG: AI and machine learning are here already, are changing the world already - and have only just begun: a huge revolution in many, many aspects of our economic, social, educational and personal lives is already starting and can only gather speed faster and faster. Everyone needs to know about this and to be thinking about how to handle it.

NP: Now more than any other time there is a real imperative for the market to take this very seriously. This is because the adage that this will happen no longer applies as it’s already happening. The external capital markets, pensions and angel investors are taking the risks that insurers aren’t necessarily ready to take. If the insurance market doesn’t realise that, there will be a rude awakening, as business begins to dwindle. The insurance market will no longer just be for insurers, there will be many other disruptors coming into the market.