Adopting AI in Insurance
On a balmy, September evening in London I attended a round table dinner, joined by a motley crew of actuaries, underwriters, CEOs of venture capital firms, data scientists and innovation-in-insurance experts. The event was focused on discussing the struggles Insurtechs face on breaking into the centuries old Insurance space.
Below, I have outlined the four key takeaways that may help Insurtechs convince the Insurance industry to adopt their AI technology.
Takeaway 1 – Trust:
Isaac Alfon, an astute senior risk practitioner – along with the majority sitting around the table – agreed, that the biggest problem at the moment with incorporating new systems into the old insurance market is trust in the data. A level of trust can be gained with transparency (how old is the data/where did it come from/when was it last filed) but the suggestion that gained the most nods was to win approval from the key decision makers within the big companies, so that trust in the process your company is selling will filter down throughout the organisation. In the modern world, technology is moving faster than large companies can keep up with, and so harbouring and nurturing client (human) relationships is still vital to attaining success in the competitive Insurtech space.
Takeaway 2 – Language:
Too many data companies advertise their services as if they are talking to other data companies who know what they’re talking about. Having spent the best part of a decade sitting in product meetings as a designer, where data science experts, developers and designers alike scratch their heads at the complex problems of the inner workings of their own tech, it is easy to see how anything but down-the-line, straight-talking copy on a website would convey a tech company’s purpose. So, any wording on a website which is directed at potential clients needs to be as understandable as possible, and needs not mention anything too technical, but should focus on how your platform will solve their problems instead.
When you buy a car from a dealership, you don’t want to know about the chemical reactions that start the engine, you just need to know if it will work when you turn the key, how comfortable the kids will be on the school run, and whether or not it’ll save you money on petrol consumption.
Takeaway 3 – Disruption vs Collaboration:
Collaboration is the quickest way to market. The general consensus was that the big insurance companies, so far, have been good at making small changes to optimise their data systems, and have avoided being left in the lurch by the speed of the modern tech revolution. The problem with the ‘change-as-we-go’ model, is they will run out of these small changes and will eventually have to make drastic ones. Companies like Monzo and Starling are prime examples of Fintech’s who decided to disrupt the market and force the big banks to invest in modernising, and quickly. And although these new ‘virtual’ banks directly target the individual user, they have benefitted greatly from starting afresh, speeding away from their bulky predecessors, weighed down by hundreds of year’s worth of legacy baggage and a reluctance to modernise.
On this topic, David Reynolds, CEO of Venture Agenda, believes that the ‘Big 4’ will soon be left behind by midsized insurers, who are forward thinking enough to change the way they conduct business now (by partnering with innovative Insurtechs), who don’t have the legacy baggage to contend with. Examples of this sort of movement can be seen in the Chinese and African ‘pay-as-you-go’ insurance economy, which flourishes as they’ve honed their processes alongside the adoption of the phone, and are flying high without legacy baggage to drag them down.
Takeaway 4 – Displaying the Data:
David Trefusis, CRO at NHBC, believes an underwriter doesn’t want to see six-hundred data points at a time. He has heard many pleas from underwriters for a high-level analysis of a book of business, based on the data found relating to those businesses. The ability to check the data’s validity can be secondary to the initial analysis of the book, and in time, that high-level analysis will be what is necessary to make a decision, when the trust for what is under the tried-and-tested bonnet is in place. Every person is different and so every person processes data differently. Any new tech must be highly customisable in what it displays so as not to serve data that isn’t directly relevant to the user.
Some other titbits related to the adoption of new technologies within insurance:
- The historic precedent is there for tech to be adopted, but it takes a long time. (Electricity was invented in the late 19th century, but not rolled out commercially until the early 20th century. The reason for this is because companies were built around coal. From training their staff, to the layout of the buildings, everything centered around the now controversial fossil fuel. But it all changed eventually.)
- Make sure to promote the benefits of speeding up old processes, as that seems – naturally – to be a constant pain point for companies with a vast legacy.
- How do you bridge the skills gap when your tech is introduced into the market? This is an important consideration. Do you establish a solid on-boarding team? Or do you send your engineers to the establishments using your technology?
- The ability to condense information is underused and underrated in the race for data – always give the option to minimise data display. Getting data off the internet is like taking a drink from a fire hydrant.