This is a shortened version of the Making Risk Flow podcast, episode “Deploying digital risk flows: a how-to guide” with Matt Churchill, VP of Customer Success at Cytora.
You can listen to the full episode here
In this episode Juan and Cytora’s very own VP of Customer Success, Matt Churchill, talk about a practical approach to deploying digital risk flow and the benefits of each stage of the transformation. The conversation covers architectural challenges and solutions, and it lays out the incremental steps an insurer can take to maximise the chances of successful deployments.
Juan De Castro: Hello, and welcome again to Making Risk Flow. This is a podcast that reveals everything you need to know about digitising risk in commercial insurance. I am Juan de Castro and in each episode, I will walk you through how to make it work in your organisation.
In previous episodes, we talked about the business benefits of digital risk flows, how they enable insurers to accelerate profitable growth by automating the data capture of risks, ensuring that underwriters only work on winnable decision-ready risks. As we’ve discussed, this eliminates all the waste that goes into analysing submissions not resulting in a quote, and all the manual activity required to pull in other data and re-keying information into different systems.
Today, we’re going to talk about how insurers have successfully deployed risk flows. And for that, I’ve got with me Matt Churchill, Cytora’s VP of Customer Success. Welcome Matt, very nice to have you today.
Matt Churchill: Good to be here, Juan.
Juan De Castro: Perhaps should we start with a brief introduction of yourself?
Matt Churchill: Yes, of course. My name’s Matt Churchill, I have been in the insurance space for about 12 years now. I began my career at Hiscox as a Project Manager working on various things, mostly operational change, but some digital opportunities as well.
Towards the end of my time at Hiscox, I was a Head of Innovation, running the innovation part of Hiscox, which was ultimately engaging with startups and seeing what was going on in the wider market. About two years ago, I left Hiscox to join Cytora, on the other side of the fence, trying to deploy our technology into insurers.
Juan De Castro: And you lead the delivery of digital risk flows to our clients, right?
Matt Churchill: That’s right. I’m running customer success, which in some respects can be the continuation of successful relationships with the customer, both growing accounts and ensuring that deliveries bring benefits. But also successfully delivering and deploying into the client in the first place.
Juan De Castro: OK. That’s great. Let’s start this episode with what’s driving insurers to deploy digital risk flows, and then we’ll talk a little bit about what are the challenges and how they deploy this successfully.
Matt Churchill: There are probably two primary reasons why insurers would want to deploy digital risk flows. The first would be the desire to decrease costs. Be that an operational cost or a cost of claims. Operational cost is the most obvious one to benefit from digitisation cause digitisation can often mean automation.
It drives the automation of certain functions. Insurers have for many years struggled to grow without also growing the operational cost commensurate to the top line GWP. Any opportunities that can drive a breaking of the link between those two numbers are obviously very interesting.
You can keep your operational costs flat. Then, obviously, there’s more profit to be had. But that’s probably one part of it because that’s an obvious outcome, but I feel that’s at least, from my time engaging with a wider market in the startup space when I was at Hiscox, but also just in general.
It is obvious that there are progressions that are made throughout the world that make life easier, often they are digital progressions. Insurance as a concept and as an industry hasn’t exactly resisted the adoption of some of these things, but hasn’t been able to take full advantage of some of the changes that have happened.
The pace of the change, the availability of information and these sorts of things. I think for a number of years now, insurers have recognised that they need to change and adapt operationally, the way they look at risk, the way they approach it, the way that they manage it and manage claims.
And it’s obvious that there are things in the market now that can actually change the way that insurance is done. And insurers are willing to explore those. But I think the first step is always, well, how do we digitise? What does digitise even mean?
That’s why I think insurers are looking towards digitisation.
Juan De Castro: And I think that those two things that could sound separate in principle are actually very linked, breaking the link between growth and the number of underwriters that you need and making it easier for the underwriter. We hear from many insurers that they are struggling to find the right underwriting talent.
First of all, if there are ways for insurers to grow without adding new underwriters, obviously that’s a direct benefit, but even when they’re fighting for underwriting talent, it’s very different to offer a role to an underwriter where they’re using model technology and technologies helping them in the role, to just offering an interface from the 1980s. And we hear from clients that’s also making it more appealing when they’re recruiting underwriters.
That’s why they are deploying these digital risks flows. But I’m sure many of our listeners and insurers will think, okay, but where do I get started? What are the major challenges that they face when they’re thinking about evolving to digital risk flows?
Matt Churchill: Yes, knowing where to start can often come down to just defining and understanding what is meant by digitisation.
It would be wrong to think that insurance isn’t in principle digitised already because the risks exist in a system. They’re generally stored as data, an email to all its intents and purposes is digital. It’s not like receiving a letter through the post and having to read it. It just so happens though that the application of what sits over the top of what is digital information. So, how does somebody look at a risk? How is it treated? How is it managed or maintained? It’s still a very manual approach. And I think that’s what insurers recognise as being digitised, not just turning information into something that’s stored in a database, but actually taking advantage of the fact that that information is now stored in that way.
I think a prime example is the fact that a lot of information is transmitted over email. So it exists fundamentally as a digital reference, but that information by the way that it’s presented and communicated is unstructured. It’s almost meaningless because there’s no structure behind it.
So digitisation can actually be, well, how can we add structure where there is none? Even though the information itself exists in noughts and ones, somewhere in a database, how do we actually understand and structure that? So that’s for us and for many clients that are on this journey, is what they now recognise as digitisation.
It’s saying we’ve got the information, but how do we now treat that information? What has to happen to it so that it can be utilised and used in a way that maybe we hadn’t considered before? And this can boil down to, as I say, structuring unstructured information, but also taking advantage of information that exists already and utilising some of the newer opportunities in the market.
For example, if you augment data with additional market information, there’s lots of it. Every day, there’s more of it available. Well, actually, when you’ve got something digital can you keep adding to it in an automated fashion? Well, of course you can. Those things are obvious. And when you get to that point, you can begin to treat the whole thing in a slightly different way. You can approach the way you appraise it, the way you move it, the way you ignore it, the way you prioritise it, you can do all those things using digital tooling. And I think that’s the big progression that people are taking as they move to digitised workflows, really understanding well is not just turning things into digital anymore. It’s treating the whole concept differently.
Juan De Castro: And we see that quite often, many clients store, for example, quote information, or they risk information on a spreadsheet. I mean, you could argue that it’s already in a digital format. I think what you’re saying is it’s not just how you store it, but how you use it. Can you mine it? Can you get insights out of that at the portfolio level? Probably that’s something that you’ve seen in many clients too, is the transition to make data available.
Matt Churchill: Exactly. It is always interesting when you come across, for example, spreadsheets of information, because they are much more structured than just an email that comes through.
But because there’s no general agreement as to how that structure should look, they tend to have to be treated in very manual ways and that, I think, everyone can recognise that if only I could treat this differently, we could understand it, we could move it from system to system.
If only everyone went to my portal, for example, if only everyone went there, I’ve got these lovely fields already to capture information from the market, considered and ready for underwriting. But unfortunately, not everybody works in the way that we wish they did.
So it’s actually a case now trying to find some way of universalising. Or at least universally understanding the information that’s available and treating it and augmenting it and utilising both internal and external information to round it into something totally meaningful for the insurer.
Juan De Castro: I think nobody would disagree with this point and I’m sure everybody’s thinking, okay, point taken, but insurance is not known for having a modern architecture, there are plenty of legacy systems, plenty of different workflows. Where do I get started? This feels like it’s a massive transition to fully digital workflows. Perhaps you can give us some tips on how insurers are approaching this journey in a way that captures value quickly.
Matt Churchill: So, again there are many ways to get started, but you raised the obvious point that change is always difficult. Changing anything, especially an organisation like an insurance company can be very hard and it brings with it a huge amount of heritage and trust in the way that things have always been done. That boils right down to the way that risks are underwritten and the way that people appraise information.
One of the biggest challenges that we face and that many insurers face is “Is there additional information that may be available to me?” So an obvious one is a third-party innovation. Is it trustworthy? Is it good enough to underwrite on? Is it good enough to believe in and move forward with?
One way of actually beginning digitisation in the way that we described it here, is to start small, to start in a way that can begin to build confidence, build trust, build belief in both the processes and the method that sits behind it. So this can be as simple as saying: what about if every risk we saw was automatically appraised on a few data points? The obvious ones, the ones that aren’t ever really gonna be terribly problematic, but maybe the ones that currently we struggle to even get in an automated way. That could just be, where does it come from? Which brokers do they come from? Do we like them? Are they top-tier, mid-tier? Is it from a company whose trade or profession is something we like or are uncomfortable with? Is it in or out of appetite? You can actually capture those data points and add some inference models to them very quickly, come up with a profile on just a few pieces of data. You could call it a clearance profile or a kind of selection profile too.
We’re now in a position to automate a decision on this first intake before anything’s really got down to the risk and actually underwriting. Just do we like it, don’t we like it. You know, should it be on top of the pile? Should it be somewhere near the bottom? But the automation of that kind of thing can deliver huge benefits because step number one looks like it’s not gonna have a huge impact on my internal resources.
It’s not going to require a huge transformation program or high degree of integration but actually would begin to identify benefits very quickly. And one of the key things that come from this is of the 50-100 data points that ultimately make up a risk, you’ve picked out five key ones and you’re getting trust, you’re getting belief, you’re getting feedback. You’re getting everything that comes from that, but in a very rapid way that lets you then move into those other stages in a deliberate but understood way.
Juan De Castro: That makes sense, start small with capturing the first set of data points.
What is the benefit that you can capture in that first stage? I think you touched on starting to build underwriters’ beliefs. If you can talk a bit more about the other sources of value, just that first step, that would be useful.
Matt Churchill: So, they’re not universal cases, but what we’ll often find is that out of a hundred risks that get submitted through, probably 70%, it can be lower than that, or slightly higher are actually probably gonna be out of appetite.
So not actually to be looked at, or even underwritten or even quoted. That’s a huge amount of what comes in is actually not for the organisation. The appraisal and the understanding of those is a burden. So number one benefit is you can actually start to remove those burdens, the burden of appraising risks that you know are a huge proportion of them you’re not gonna want after all; there’s a time-saving in being able to do that.
But there’s actually a more long-term benefit that can come from having done this in a structured and deliberate way, because you are automating the extraction of key information and it’s being stored, and it’s being appraised. It’s actually available there for a CRM system. A lot of insurers are taking CRM more seriously than before. That’s not to say that it hasn’t always been a serious point for insurance companies, but often it’s been at the risk level. So there’s a customer and they have their risk underneath them. But now it’s actually, well, there’s a customer and they may or may not have a risk that we know about at the moment, but at least we knew we saw a customer once upon a time who had a request for a quote or something.
Quite often that can actually be lost in today’s processes. If you’re not going to underwrite it, you may not capture it at all. You might just look at it and just send it off. It might go to an admin team and get captured somewhere. But that can often be slightly undefined.
What insurers can do with a digitised workflow is everything that can then be stored. So even for the risk that you don’t want, you can capture quite a reasonable amount of relevant information about the customer and about maybe what their risk profile is and have it in your CRM. Have it in your CRM for a later date, for chasing down new opportunities for if the appetite ever changes, you can identify that you can go back and have a pool of customers that are worth chasing down there. That’s another benefit: unifying the entire intake under a data schema structure that lets you capture information in a way that you couldn’t have effectively do before, without investing quite a lot of time in the manual function of doing so.
Juan De Castro: And historically there was a tension between capturing data and cost, many insurers would say, well, if I need to capture all the data, even for those submissions I’m not gonna quote, it will require a lot of effort from the back office team or even underwriters capturing that data, so I won’t do it. Whereas I think what you’re saying is that with digital risk flows, where you can extract that automatically you don’t need to incur more cost, but you can have all the intelligence about the full submission inflow.
With this first stage, extracting the first few data points and doing basic triage, you can also start analysing the type of submissions you receive, right? And is that something that could be also used to inform things like appetite expansion. What other uses of that data have you seen in the market?
Matt Churchill: Definitely yes. To the point of, can it be utilised to support new insights, which is what that boils down to. It’s fair to say again that finding a documented appetite for an insurer can be quite hard, like what is in and what is out? Because there can be a huge amount of ‘greyness’ there both for the brokers and customers, but also for the insurer themselves to actually clearly articulate, well we don’t like this, but we do like it under these circumstances.
There’s often nuances that reside within there. Now, one thing that can be done by having digital profile very early on in the process, is as that risk goes through its subsequent stages, be they either decline or to full underwriting into bind. You’re actually capturing and adding more knowledge to that first part of that profile, what you can do with that is begin to identify, well, actually there are areas where we think we don’t write it, it’s technically what we class is very low appetite, but we actually have quite high success with it. The MI that we’re seeing through says, even though it was a low priority or almost at the bottom of the appetite pile, it turns out there’s a niche there that we do very well at. Whatever that might be. And you can start to uncover those things because you have a slightly different take on the insight you can gain because instead of being quoted one, quoted lost, and then basically declined; you actually start to get a bit more nuance behind. Well, actually this was almost a decline. It was almost within the kind of appetites of things that we weren’t that keen on but here’s a huge opportunity. Or alternatively we’ve identified that we are actually declining or not winning a lot of business that actually we think we could move into.
So an insurer could analyse this backlog of information that’s available and identify new appetite areas to go into. Almost test and see what it would’ve meant to have underwritten some previously out of appetite risks, because you now captured the volumes, the flows. You can apply some logic to the likely conversion rates. It definitely gives you a different stream of insight that you can play across the entire business.
Juan De Castro: At the end, this is about treating commercial insurance where the sales cycle is much like in a retail business where you understand in what segments and what products are you having more success, less success? What type of clients do you bind more or less?
So, I think all these benefits come from the first step of getting onto the journey. What can the end state look like? What’s the end value that insurers can capture after a number of stages.
Matt Churchill: So, there’s kind of two end states, actually there’s one that’s more immediate and one that’s very much future focused. We’ll touch on both, but the immediate one, and when I say immediate, I could mean, you know, a number of years for some companies to get there, is actually that all of the risk data is captured at the start of the process in an automated way. So we’ve talked about maybe five to seven data points being captured because you can do an awful lot of analysis and information on just five; increase that to 10 or 20 and you begin to be able to say, well, where’s the most appropriate team for this? Who’s the most appropriate underwriter to work with? What’s the most appropriate course of action within the insurance company to see that this is won? You can start doing that as you add more data in. You can augment it with other relevant data sources. That can, in their own aggregate, give you a huge amount of insight, not necessarily directly at the point of underwriting but definitely at the point of analysis. When you come to analyse a book and see, well, why aren’t we winning this? You can sometimes uncover certain things about, about the profile of the risk that may be lost in the capture of the information, because an underwriter might have appraised it and seen something they didn’t like and declined it, or didn’t take it in that instance. That’s actually then there, because the data was augmented.
So more data means that you can actually start adding even more to it, that becomes exponential. You eventually get to the point where the entire risk object is identified and extracted and created at the very start. The whole thing is there, ready for an underwriter to just underwrite.
Now, that’s the first immediate place where we would like this to get to, is a risk from its receipt or even this can apply to a renewal, being able to catch the information and find out all the other things about that risk that might exist in a third party system somewhere. You are trying to give the underwriter the full profile of the risk to underwrite. So it really does support underwriting. It would also support automation of certain underwriting decisions. So especially at the lower end of the markets where you could almost call it the direct approach to business where maybe the premiums are kind of sub £5,000 on a commercial class.
There’s actually enough information there to start making automated decisions. You’ve got an awful lot of risk information, a huge amount of profile. Some of those decisions can be automated. Even if it’s only the offering of an indicative quote, you’ve actually saved a huge amount of time and you can see the turnarounds on those in minutes, not hours or days, which is often the case.
What I think is exciting though, and this does move on to the next level of all of this digitisation, is that it allows insurance companies to start looking at underwriting and actually saying, well, is underwriting moving at the pace of the available information? Is there information that means that we should reconsider the way that we appraise some risk?
There is definitely risk information available that if utilised, maybe in the aggregate view to write large books of business could be taken on its own without actually needing to have a huge amount of additional stuff added to it, to understand the risk profile of a client. And that I think is where it starts to get very interesting. When insurers prepare to actually look at the way that they underwrite and say, well, the information available at the moment from third party sources, it may not be accurate all of the time.
But actually, if we look at our underwriting process and we maybe reconsider the way that we do certain things, actually some of this information could be utilised to streamline the whole thing without having to fill in the prop form, which is often the approach at the moment, just fill in the current prop form. Actually, maybe the prop form changes to be filled in with data that’s not the same as is used at the moment. And I think that’s quite interesting when we start to get into those spaces.
Juan De Castro: That is very exciting. I think we’re seeing a number of insurers already halfway through that journey.
Perhaps one last question, because we talked about challenges in terms of how underwriters look at risk. But one of the challenges we hear quite often too, is about the technical architecture. How do you create these digital workflows in an environment where, as you said, submissions are received via emails, some insurers also have a legacy policy admin system, some insurers are thinking about deploying a CRM?
So, perhaps if you can touch on almost the top two most common architecture patterns, I think that would help the audience.
Matt Churchill: there are lots of architectural patterns, but we can talk on the principles of two that get you going very quickly or two that seem to always work.
So the first one is if you want to begin digitising a process at the start. So it’s going be submission based. How do you get the submission itself into a place where that function can then take place? Ultimately, we’re talking about an email, how’s the email put somewhere to have automation done to it.
Now that email could, for example, in our instance be sent to us for that task to be done. So sent to us as a provider of the service and it can come over in a way as simple as you simply forward it. The easiest architectural pattern is you send an email to another email inbox. But there are also very quick and simple methods of that integration happening, which can just be almost an API post to our platform or to any platform that says here’s the email that I want you to extract the data from.
So generally they’re not too fraught as an architectural pattern those methods, because they don’t require a huge amount of data integration.
So that’s the first pattern, getting the submission or the email or the bulk information somewhere where it can then be processed to become digitised in the concept we’re describing here. Now, depending on where you send that, a lot of the functions of getting the information, augmenting it, adding certain rules and considerations to it can be done in that platform.
So that can actually be taken care of in there, without any other architectural engagements. A lot of the time you’ll find that augmenting information, calling out to third party services, sometimes they have APIs, that’s great, sometimes not. Any platform that can actually offer the ability to rebuild certain services on itself so that they’re available, can obviously save a huge amount of time and effort.
But then we come the main challenge, really, which is how do you get the information back again? How can you actually push this back into an insurance system? So the easiest pattern of all would be an API ready to receive risk information in its entirety. It has to be said that that’s not often the case because of the complexity of the multitude of products that insurers would write. So, you know, a PI product or a PL product, and then property and motor and all these things. There generally isn’t a ready to go API connection. So what we often do is integrating with a CRM directly in its first instance and a number of insurers, certainly all the ones that we’ve encountered probably without exception, I would say they are using something like dynamics or Salesforce as their CRM tool.
Now, integrating with things like Dynamics or Salesforce is obviously a lot easier than integrating with some of the archaic or custom built policy admin systems. So you can see the benefits very quickly. Because integrating small amounts of data into those systems is a lot easier. And, the architectural patterns for those are defined and quite clear. It’s not that challenging. What can become more challenging, which is why we also start slow, is to take full advantage of an extracted risk, so somebody’s already done the job now, turning it into a digitised profile. You ideally need somewhere to send that.
So in the interim period, we will often allow an insurer to come and view it on our platform. So you can actually see the information presented and make decisions there because they still might be declined. And then you may actually choose to manually re-key because you’re only re-keying the quoted and best risks you’ve got. But obviously the end goal for all of this stuff is to automate the process of taking that information and putting it into your own policy systems. Some insurers are well on that journey already. Some are just discovering it and moving into it, but all of them have the same ambition, which is just not wanting to re-key information from one place to another. But it’s fair to say that is still in development in lots of places.
Juan De Castro: Okay. That is very helpful.
I think we’ve touched on many topics from the why insurers are moving to this digital state, into the challenges. And I think a question that comes up often is, okay, that sounds great, but like, how do I start integrating? I think these patterns that you described sounds like a feasible first step as it also ensures an involved target architecture.
Matt, thank you so much, and for the rest of the listeners, I’ll see you in the next episode of Making Risk Flow.
Follow the show in your podcast app of choice: Making Risk Flow
