Augmented Underwriting: The AI Powered Insurer
The ultimate vision for insurance is to assist underwriters with the firepower of artificial intelligence. We assembled a cast of experts to illuminate the road ahead.
The ultimate vision for insurance is to assist underwriters with the firepower of artificial intelligence. Thus, all routine activities can be performed automatically, and support offered to underwriters by AI engines for higher value roles.
But where are we? In short, when will we arrive at the promised land of AI augmented underwriting.
FintechOS Annual Summit – our virtual event that brought together impartial industry leaders, visionary speakers, and experts – assembled a cast of experts to illuminate the road ahead.
Here are the best quotes from the session:
Progress depends on digitising the wider ecosystem
Andreea Pipernea, CEO, NN Pensii Romania
“AI underwriting is something that is still aspirational and something to come in the future. You cannot think about underwriting in isolation. It has to be part of the digitalization of other processes as well. At NN we have a full digital selling process, we have digital signatures, and 60 per cent of underwriting processes are now automated. But in order to get further we need a discussion about the ecosystem. Links between medical providers and the client [must be digitised] so we have less paper in the documents we use today. Everyday there is progress, and it’s a top priority. But we still have a long way to go because of the inherent way in which we function, even as an economy overall.”
The biggest payoff of AI underwriting
Aleš Tausche, Chief Insurance Officer and Executive Member, Generali Romania
“For me the biggest benefit is it really helps to streamline your business operation. So we spend less time on digging, processing data, and understanding models. We can really focus on what is important.
In the future, we intend to spend less and less time on processing. Less time asking questions, and making it simpler to focus on the importance of decision making. Here I see an unlimited area for implementations in the industry.
In property and casualty insurance things should be very fast. You get the data, you check, and basically you could have a renewal or new business offer in a matter of seconds. In other fields it is different, as the data needs contextualisation. Underwriters will always need to apply common sense to business decisions, even when augmented by AI.”
Three changes we need
Raj Sakaria, director of enterprise apps and architecture, Hyperion Insurance Group
“In the mid-market there’s a huge shift in the way we develop software. If you look at some of the platforms in the industry they have not changed in 20 years.
For me there are three fundamental changes that have happened around building the next-generation platforms for underwriters. I’ll try to keep non-technical, although I feel myself naturally turning into a geek, so apologies in advance! If you look at frameworks in a non-technical way they are comparable to puzzle pieces. Prior to 2000 you had to build and manufacture each piece of the puzzle, which was really costly. The bigger the job, the more pieces, so more time and money. I think now with the cloud, these pieces, aka the frameworks, are already built for us, and all we need to do is paint the picture and put the pieces together.
The second change is around open architecture and legacy platforms. When you have two or more closed book platforms trying to communicate with each other it’s almost like two people with different languages trying to communicate. From a platform perspective, this means a lot of rekeying, and people start making mistakes. It becomes a data quality issue. This leads to underwriters not trusting the data.
And third is around release management. We need to release quality code faster. I remember we asked one of our vendors to add a field and they were giving us times of six to eight months to deploy the change. In comparison, Facebook releases software many times a day. I’m not saying the industry needs to change to be like Facebook, but somewhere in the middle would be great!”
AI? The current software is often 20 years out of date
Manjit Rana, Insurance Innovation consultant
“My son prior to joining his grad scheme did a project in insurance. Obviously, they didn’t know it was short term, they thought they were hiring him. And so they were training him on their new claims platform. He spent six weeks being taken through it. When he resigned they asked him about the system. He said,
‘It’s a really good job you are moving to a new system, because that thing looks 20 years old’.
The guy doing the exit interview said, “Er, actually that’s the new system being implemented.’
It’s interesting that our expectations as people within the industry is often completely different from that of the raw talent we are hiring.”
AI can eliminate needless questions
Mark Andrews, Insurance Director, Altus Consulting
“We work with Legal & General. They have an insurance smart quote proposition for home insurance. We got the questions set from 50 questions for a quote to less than 10. We used a whole load of data which they had within their business, but also used third party data.
It’s trickier for personal lines, as ultimately price comparison sites own customer acquisition and therefore unless they move to quick question sets it’s potentially a wasted investment.
Looking at Direct Line, their questions set for motor insurance is still at 60. Aviva has invested more and is down to 20.
In the personalised space, where we work the most, AI is still rarely used in the quote and bind journey. We are seeing lots of insurtechs put themselves up, but the main AI and machine learning bits are in the claims space.”
Start with AI at the core functions
Karl Lawless, Insurance Director, FintechOS
“Yes, absolutely. There’s the giant vision of the AI-powered insurer. But people in insurance are pragmatic, and we see them biting off small chunks. Particularly in triage. Basic AI underwriting tools enable underwriters to focus on underwriting.
Underwriting is a data-driven activity. And the challenge with more and more data is how does the underwriter concentrate on what’s relevant? What is going to impact the quality of the risks they select? What is going to be important in terms of pricing a risk, and being able to aggregate all the data into a single place and make it easily digestible for the underwriter? Again, we want them to focus on underwriting, rather than wading through reams of data.
So I think bite by bite insurers are improving. And over the course of the next few years I see AI and automation becoming increasingly important. But it is a journey and will take some time.”
The most effective and immediate deployments of AI in insurance
Aurelia Costache, EMEIA Intelligent Automation Leader, EY
“Virtual agents are bringing a lot of benefits. They reduce customer facing interactions and improve customer satisfaction, whether it relates to underwriting questions or claims management. We also see adoption in the document intelligence space meaning combining Optical Character Recognition with machine learning to transfer information from a paper format to digital. Then using software robots, RPA, to perform different activities and digitalize processes.”
MAIN PHOTO Credit: Unsplash