Loading summary
Transcript23 lines
- [00:04]
A
Hi there, it's Jennifer Strong and this is the next innovation. It's quite rare that we think or even talk about insurance within the context of new tech. Most of us, unless we're in the market for a new policy, we don't bother to think about how much the world depends on insurance. It can protect some of the most important aspects of our lives, our safety, our loved ones and ourselves. The fact is, the insurance industry is well over 250 years old, but its operations hadn't changed much until about 20 years ago. Computers really showed how data analytics could provide sharper information in a much more efficient way. And now, as AI continues to grow in application and capability, the insurance industry is experiencing a quiet transformation, providing virtually every aspect of the sector a smarter and more practical way of operating. Much of this transformation was witnessed and discussed at the recent INS Tech Transformation Summit in Dublin, Ireland. It's an annual convention of insurance decision makers and this year they examined how AI data and operating model changes are reshaping the key pillars of the underwriting, claims, service and growth. As a way to better understand what's driving the change, we spoke with three companies who attended the summit and are shepherding this new era in insurance.
- [01:31]
B
My name is Aidan O' Neill and I'm the founder and CEO of a company called Dockersoft. So what DocuSoft do?
- [01:39]
A
It's a company responsible for designing and creating core claims management systems for insurance companies. Claims management systems are the operational software engine used by insurance carriers, health plans and enterprises to manage the entire lifecycle of a claim. It digitizes and automates investigations and evaluations, among other things, replacing manual processes with structured, measurable workflows.
- [02:03]
B
If you take the original document management technology that we were doing, I describe it sometimes like a doorbell on a house. If your doorbell breaks, it doesn't really matter, you can knock on the door, but if you're, that's for a document management system. But if you're doing a court claim system, that's like the electricity in a building, it's fundamental. So if you're an insurance company, the claims system is really where it counts.
- [02:30]
A
Aiden had experience working in insurance in Japan. He was part of a group that developed 3D software for various uses, including analytics. He joined Docosoft in 2008 as part of servicing the broader London insurance market, including groups like Lloyd's of London, which is often considered the world's first insurance marketplace. Having originated in the late 17th century. Automation was still pretty novel in 2008, but was starting to prove useful. For claims upkeep.
- [02:58]
B
At that time, the London market really had introduced the first, they call it an ECF Central Market claim system. Electronic claims file is what ECF stands for. So they had introduced that into London around 2007. Before that you'd see people around, walking around in London with big claims files. I mean that was typical of what the London market looked like. So you'd see your typical bowler hat and black umbrella and a massive claims file binder under their arm. So what happened in say 2007? The market introduced this central claims repository. So what happened on that? And each insurance company would get messages each morning from the previous day's claims that any transactions that would happen on a claim, like we call them claims movements, if a reserve was changed or if the claim increased in value or if some portion of the claim was paid off, that would generate a new electronic message. So basically it was like essentially an Excel spreadsheet that was sent to the company each morning. But over time, when we were asked to solve the problem, it was say 50, 60 messages a day. And what the insurance company had to do was each message had a claims number on it, like a unique claims UC or unique claims reservation number. And what the insurance company had to do was track, you know, if a number came in yesterday with, I don't know, a 50 million number on it and then if a parallel that was paid off the next day. It wasn't rocket science to do it. All we were doing really ultimately at the start was matching unique claims reference numbers from the previous day, from the previous week, store them in a database and link them together on a simple software system. That's how it originally started.
- [04:53]
A
The London insurance market is one of the largest in the world. Marketplaces like Lloyd's of London are often made of 50 to 60 syndicated companies. When Dacusoft started, their key technology proved it could provide clarity and efficiency across the board.
- [05:07]
B
And what happens in the London insurance market is that if there's a risk, say a satellite worth 100 million that's insured in London, what happens that there's a lead insurer and there's a follow insurers. So the lead insurer might take 5% of the total risk. And then there's various other insurance companies would take other portions of the risk. And originally that was written on a piece of paper called a sl. And the slip now is obviously it's a digital slip now, but originally that was a paper slip. In terms of how the London insurance market is different to other sectors is a subscription Base if you're on the, if you're a lies of London syndicate company, you there is minimum standards that you have to, you have to adhere to compared to say if you're writing 100% of the risk like say US insurers that are not in the London market. So in the London market if you're a LYDS insurance carrier you say for example you have to pay a claim inside five days. That's one of the KPIs it's called Lloyds minimum standards. Now if an expert is appointed like if it's an airline crash, you get an extra seven days on top of that. So that's 12 days. So basically if you have a policy, if you have a ship or whatever, a large save something of immense value like a factory, if you insure it with lives of London there's certain guarantees that you get but those guarantees are made up by the 50, 60 syndicated companies. So that's where we play on the claim space that we help those insurance companies keep the lies minimum standards.
- [06:49]
A
About 10 years ago Docosoft started on AI. Initially what they noticed was the disarray of data. Some of the information was held in emails and other on spreadsheets. So the challenge was how do we get all this data to be centralized and easily accessible. This kind of unstructured data as it's called was the key piece to designing an AI model that could provide a one stop solution for claims evaluations, investigations and much more.
- [07:16]
B
The central market systems have structured data that's like claims messages with actual financial information on. But if you take emails coming in that has, I don't know, broke broker information on our loss files sometimes a lot of that is unstructured data. A lot of companies with AI they focus on saving on operating costs like say automatic claims handlers, stuff like that. But that's only saving say 1% or 2% of 100 million. But we're more looking at how can you move the dial on the larger number on the 10 billion say with claims leakage or claims portfolio analysis. So that's kind of really the docusoft how we're looking at it. How can we move the dial on the larger number? Not on efficiencies in terms of saving headcount really on the books of business that insurance companies have. An example would be claims portfolio analysis where we can look at, we can look at trends in the book of business that insurance company has. We have the usual AI models that turning an unstructured claims file comes in on an email into a structured claims message. We have the triaging of claims coming in where if it's high urgency, we have the usual, if you summarize documents in, it's a 20 page or 50 page document summarized into a paragraph. But as I said, if we can look at claims portfolio analysis and see trends in the book of business, it's more important. I'll give you an example. We've seen one of our insurance companies in the US actually that they were insuring trucks and there was a lot of accidents with the trucks and they couldn't figure out what it was. But when they looked at the data over time, they could see that the sleeping space on these large trucks that travel overnight, the sleeping space had been reduced by one foot, the newer truck models so the drivers were sleeping less, which meant there was more accidents over time. So there's an example of something that if you analyze the claims over with a large data set, how you can look at trends on that. The holy grail in this area is obviously predictive stuff. What can you predict in future? What's going to happen now from that point, docasoft is looking at, we call it an early first notification of loss. Where we can look, say with the Californian wildfires in the US in say January last year, how could you give those folks even five, ten minutes to get out of their building quicker? So it's not just there is technology at the moment that can pay, you know, if your house is wiped out. You can see insurance companies now, we can pay the claim even before the loss has, you know, if the building is totally destroyed or not. But we're trying to take it a step further. How can we give those folks, like as I said, time to get out of the property to save some belongings if there's a flood or there's a fire and stuff like that.
- [10:38]
A
Part of DocuSoft's AI model is preventative prediction, which can be a useful tool in areas prone to natural disasters. Some insurance policies require sensor installations which can help alert homeowners about potential disaster consequences, like if there's a flood or a fire. It works through satellite technology which can use fire smoke to detect the direction from which the fire is coming and where it may be going next.
- [11:04]
C
I think what AI is bringing to the table is a fundamental change in the business value and value generation.
- [11:13]
A
This is Aidan Brogan, chief commercial officer at a software company called codeast. They also work heavily in the insurance industry, working towards delivering more tailored data driven solutions. Their work is primarily for MGAs or managing general agents, a type of insurance agency focused on analyzing potential risks and handling claims. We asked Aidan why data was such a big part of understanding the future of insurance and why he thinks AI is essential.
- [11:41]
C
Insurance is fundamentally a data driven business. You collect a risk and try and understand as much as possible about that risk and then underwriters and actuarials model that risk and try to design an appetite, let's say, on a premium to, let's say, ensure or mitigate that risk. What's interesting is the insurance industry is fundamentally designed to operate with historical data. If you look at sort of ironically, insurance is a data driven business, but it's been operated off historical data. So our focus now in Code east is not to specifically code, let's say it's more design and focus on architecture and then use other tools like cloud code, et cetera, to automate that. The results are fascinating. What we're seeing is 5 times or 10 times better efficiency in our, let's say, speed to market. But also, and as importantly, the quality of the results are far greater. Quality of documentation, quality of quote and quality of testing. So I would say in the last year our internal capability has fundamentally changed. Now if I look at our customers, and specifically, let's say customers who have handle on their data, we're now seeing data and AI agents transforming their business. If I give you, let's say one example, one of our customers has a renewal book of business and typically most delegate authorities have about an 80% of their revenue comes from renewals. So that's pretty predictive. The challenge is it's all historical and most companies use Excel to analyze that and forecast it, et cetera. With a real time data framework and AI agents, you can, let's say automate that. So the forecasting is done in real time and then as the business occurs and renewals either are accepted and renewed or rejected or declined, that is also brought into the mix. So that one area specifically of renewal management is being transformed by data, real time data and AI. If now you then go to the stage that you start building agents on top of that, and that's what we're doing now, you have experts, digital experts, constantly looking at your renewal book of business and then when they see something that needs to be advised, advising the CEO or the chief underwriting officer of business that may be at risk or maybe is probability of declining, stuff like that. And if you think, for example, at a basic level, insurance is data driven by experts and that is the limit to the scale of your business, what AI brings along is the opportunity to engage all of these digital experts to assist your team and your business. And I think therein lies the opportunity. So this inflection point, let's say that we're going through, is really exciting and I think it's, let's say, presenting huge challenges, bit scary, to be quite honest, for everybody because of the pace, but presents huge opportunities.
- [14:56]
A
Code East's core AI product is called OneView. It's an AI native underwriting platform for MGAs and it's deployed around the world for companies in the United States, Canada, the EU and the uk.
- [15:09]
C
So first of all, we have an underlying very sophisticated data and AI governance framework. We call it a canonical data model, which is a bit technical, but it's basically a data model of all of their data entities. And there's tens of thousands of data entities. But what it means is that they have a granular view on their business and that is all aspects of their business, their brokers, their underwriters, their declines, their submissions, their gwp, their commissions, et cetera. That then feeds a suite of intelligence services that are directed at specific parts of the business. So finance, the CEO, sales and marketing, how are brokers performing the underwriting, team submission, ingestion and triage, et cetera. And then on top of that, we have built a suite of AI agents. And the idea of AI agents, agentic agents, you know, whatever the language, the way we look at it is very basic. It's like having the best expert in the world in a digital form. They don't get tired, they're consistent and are available there to help you as an underwriter or as a finance person to do your business. And the idea is that you can focus on your actual high value, you know, business tasks and these agents will do the mundane work, but also then in real time analyze and triage the whole data payload, your business, your operation, and then make advice, analysis to you and your business. And our view is that's the future model. So AI is presenting not sort of a small change of how a business, let's say, operates. It is a fundamental change in the business operation. So let me be specific. We run MGA CEO and he spends probably a week pulling together all his business and putting it into Excel and then analyzing his forecast for the year. And there's a number of objectives of doing that. One, he wants to understand what his gwp, what his revenue target is for the year. Two, to ensure that he has the proper level of personnel available to service that, because it's not linear. There may be periods where it's very busy and not so busy. And thirdly, to ensure that he has the right level of capacity to service that renewal and he can report into his capacity providers. The challenge is that prior to Code east and our OneView platform, he was doing that manually. Now all of that is automated, but also and more importantly, it's in real time as business is transacted either positively and is bound or declined. He sees that in real time and then we can start to analyze that and direct the business to opportunities that we believe may be at risk for whatever reason. So we're seeing what's interesting is I would say with digital sort of transformation, the focus was on making things more efficient and effectively taking cost out of the business or doing things, doing more with sort of the same. However, AI agents are now bringing revenue opportunities and helping the business to focus on, you know, where there needs to be focused to maximize the return for the business.
- [18:33]
A
Most of AI's involvement in the insurance industry is largely behind the scenes, providing companies the leverage to ensure they can operate smoothly, creating workflows that work 24, 7. But can AI models be helpful for new policy cycles and new customers? Niall Crawley is the CEO of Inaza. Like Code East, Inaza automates underwriting and claims management. But it integrates AI into almost every step for new business applications.
- [19:00]
D
What we focus on is automating the policy life cycle. So new business applications. So hey, I want an insurance policy with you to an endorsement which is my driver's license has changed, it expired, I got a new one to renewal to claim. At every point it kind of result revolves around the policy. So that's very key for how we integrate with systems. We work with quite a legacy systems I would say, so we have to make the integration very easy. So we just say just give us all of the data. You don't even need to care about formats, just give us the files and we'll work out if it was a claim. New business Once you automate key points of the policy lifecycle, we say start with the boring problem first. So those kind of document automations, checking those images, then you can start doing the fancy and exciting stuff, which is fraud detection. So we have a standalone product we call Doclins. We only give it to clients whenever they've actually started automating the claims process with the boring stuff. But what it's able to do is holistically look across the policy, across the claim, pull in third party data and it can provide you with an entire Fraud like PDF, everything from. Is this AI generated? Is there handwriting on this document? Hey, the invoices for a Honda Civic, but it was a Toyota Corolla on the policy. And it just dilutes down all those. All those data points that we've collected during the underwriting process. So we can straightaway just give a score 0 to 100. How fraudulent is this claim? And just dilute it down into plain English for a claims analyst, because they have so much going on in their life that they just want to see what is the actionable items out of this fraud, out of this claim in regards to fraud. So it'll just say AI generation. We've picked up that the website on this invoice isn't real and that the address of this body shop is actually in a different location. This is what you need to look at.
- [21:02]
A
Niall came into insurance not by choice, but almost by chance. He'd been working in investment banking for some time, and he quickly learned of the different ways the insurance industry lacked the necessary means of automation just to keep the wheels turning. As demand grew, he developed an AI model that could automate many of the menial aspects like underwriting, but used market participants as resources.
- [21:23]
D
And because of my background, I kind of accidentally built a AI automation engine for underwriting and claims. And I was more and more involved in the sales cycle and I realized that all of the questions from people we were speaking to were converging about how we were able to automate at such a large scale with AI at such a low price, effectively compared to other solutions on the market. So we decided to spin the automation engine out as the Announcer platform. I suppose insurance, on the face of it, is actually a very simple industry. If you sell policies at the right price and you sell enough of them and the cost of your claims is less than that amount, then you're going to make money. It's kind of one of the few industries that you can kind of statistically forecast how much money you're going to make. Now, the reason that insurance companies become kind of unprofitable or have bad years and stuff is either there could be a hurricane or a cat risk or an act of God. That's very different. But say, in a place like Ireland, where there's no earthquakes, no hurricanes, no nothing, is because they're not pricing those risks correctly, so they're not underwriting the risk correctly and they're not keeping their cost of claims low as well. So if you automate and make your systems more accurate in the underwriting process and cheaper. And if you keep your cost of claims cheaper, meaning if it's a windscreen crack, you know, just pay out the claim straight away if it passes some basic fraud checks. And when someone gets started on a tricky claim, they have all the information they need. So that directly affects what's called the loss ratio, the difference between the premium you take in the amount of money you lose on claims. So that's how automation kind of ties in there. And that's what we're like really, really focused on when we start a project with one of our clients is how are we going to tackle your expense or loss ratio? How are we going to make an impact on your P and L? If that's not like clearly defined, we'll almost not do the project. We need to actually sit down with our clients and say how does this map out to a direct impact in three months time, in three weeks time, and then we're happy to push forward. But that's how automation kind of makes a big impact on the insurance industry. If you're writing the right risk and if you're doing it cheap enough, you're going to do well in the market.
- [23:51]
A
The work that Inaza and companies like docosoft and Code east are doing has demonstrated that automation can provide the upper hand for efficiency. But looking ahead, it appears that its behind the scenes, largely unseen nature could have a profound trickle down effect on current and prospective customers.
- [24:09]
D
I think pricing will get even more accurate, not only from it being easier to explore your data as a non technical user, but it'll be better priced because a lot less things are going to slip through the cracks. It's a very manual process even with high frequency policies. So like auto underwriting is thousands of policies a day, a lot of that is still manual and a lot of stuff just falls through the cracks. So I think things will get a lot more accurate, a lot more well priced, potentially a lot more personalized. I would say that's probably going to be one of the biggest impacts on insurance.
- [24:53]
A
This transition marks a significant turning point for insurance. AI's role in the corporate sector is slowly revolutionizing how business clients and large syndicates operate and how they interact. The through line in these conversations with companies wasn't just that AI and automation are the future. It's actually the saving grace for such an established, yet overlooked industry. And as we prepare to embrace more and more of AI's capabilities, capabilities, we can look towards insurance not just as a case model for transformation, but also an example of what was missing thanks for listening to the next innovation. This series was produced by Situation Room Studios and powered by Enterprise Ireland. Investing in the next wave of innovation. Our executive producer is Christine Barata and our senior producer is Sharon Barreiro. Lysa Pena is associate producer. Additional production assistance by Global Situation Room. I'm your host, Jennifer Strong. Until next time.