Arch Capital Services Inc.; Joe Riesberg,
Senior VP & Chief Information Officer, EMC Insurance Companies
Transcription:
Stuart Rose: (
Well, thank you very much, everyone for joining us today on this session on fueling and I'm gonna be honest, forget this one.
Stuart Rose: (
Data transformation fueling automation. I'll get those mid two words mixed up because I think they can go automatically either way. I'm Stuart Rose. I am a strategic advisor for PNC insurance. We're an advisory and research company out there. And I'm joined by Prashant, let you introduce yourself, otherwise I will crucify your name.
Prashant Nema: (
Prashant Nema. I am the chief information officer for arch capital group
Stuart Rose: (
And then Joe
Joe Riesberg: (
Joe Riesberg. I'm the senior vice president and CIO at EMC insurance.
Stuart Rose: (
So like I say, I will be the moderator for today's session. And so without that we might all get started. I think certainly the topic really is this data transformation fuel and automation. And I think it's fairly obvious that you can't have automation without having good data. So the obvious place really, I guess, to start the conversation is around that. And when you think of the insurance industry, it's certainly no doubt being in a data driven industry for many years, it's used historical data to be able to predict future behavior, whether that's looking at loss, severity, frequency rates, and using that determine the premiums. But unfortunately it's really estimated about 80% of the insurance companies spend about 80% of their time preparing the data and only actually about 20% of the time analyzing the data to get the insights. So for sure, I'm gonna ask you then start off. How do you flip the equation? How do you make it that insurance companies spend more time analyzing data rather than preparing the data?
Prashant Nema: (
So as the title suggests and fuel that word is front left and center for various reasons, these days, especially when you go to the gas stations these days, but this fuel, what we are talking about is worth way more than any fuel you're putting into your car. So what we are talking about over year is literally data transformation. If not done, you almost have a business whose existence is a challenge in the coming few years. And what we mean by that is everybody has realized it everybody has started work on it as Stewart just highlighted because everybody has realized that people were working with it so far, not on it. As they've realized, we cannot keep working with it on the side as, and when needed. This has to be the centerpiece of how actually we do business.
Prashant Nema: (
And which requires for the data transformation to be part of your process automation, data transformation, to be part of every transaction when it's done data transformation, right at the time when you are coding to the time when you are billing to the time when claims are coming, you need to think about data and capturing data and data enrichment and data augmentation across the whole ecosystem. Rather than as a post Factum today, what best people do is something has been completed, dump the data possibly in some warehouse someday. And people run some reports, which will come out in my inbox Monday morning, everyday or Friday at the end of the day. And I'll flip through them, right? Which is working with the data, not on the data. The issue is there's a lot of data insurance industry. It is there are a lot of stats over there that collectively the insurance industry maybe has three to five times more data than what the banking industry has.
Prashant Nema: (
And I wouldn't be surprised actually, the number is way more than that. And the insurance industry has been around for a long, long time. A lot of companies which are existing through MNAs or organically have been around for a long time. Hence you have these data silos and pockets of data strapped in systems and processes and various departments, which is going to take a lot more time. But it has to be through in a systemic way, rather than thinking about solving these in silos, hence the word data transformation, what it applies is you need to be thinking about it holistically so that you can address it. And it could be painful when you start doing such a transformation and it'll take you some time and investment to do it right. But once done, you could actually get to the point where the transformation is complete. And then you're operating that platform as a company and actually spending 20, 30% of the time and actually spending time using the data to run your business. We are some ways away from there because not many companies are thinking about it. Some people who are thinking about it are in early stages, thinking about it, to do it. And some who are doing it are doing it in pockets and parts and in silos before they start thinking about it more holistically,
Stuart Rose: (
I would certainly agree that insurance companies have so much more data like say three to 5%, five times. I think that's very about a conservative figure out there. And I guess then maybe staying on a bit of a theme of whether it's data preparation, data warehouse and whatever else. So who do you think is really responsible for data management, data quality in your organization, Joe? And should it be the business? I know you are on it, so you might be biased when you say that side of things. And how can you, what is it you can tell the audience about how do you get that data in shape to be able to do that analysis and automation?
Joe Riesberg: (
This one's real for us here at EMC, because frankly we're going through a large transformation that's built around data. And for us, it's unlocking that data out of an old mainframe. So truly trying to get it to a point to where we can use it, everything Presant was just talking about. And frankly, I think saying the business is a bit of a misnomer because I think the intersection of business and technology is one in the same. We used to have to differentiate that concept and kind of building on what you were talking around around data management. Well, data quality and data management are a heavy lift for organizations. So for us at EMC insurance data quality sits within the actuary function, the strategic analytics. And we also have data stewards, embedded throughout where those business owners own those elements of the data to wait to ensure quality.
Joe Riesberg: (
They have oversight around the decision of those of that data. We also have an enterprise data management team. That's essentially in charge of engineering pipelines, the data models, the operation side of it. And that does sit within technology. So, that's how we do it at EMC. But I think if you talk to a lot of carriers, the chief data officer, the enterprise data management, the data science team, those are all things that 10 years ago didn't exist. Right? And so we're all trying to learn and try to navigate how best to make that work. The other thing my team hears me say all the time is operating model Trump's organization model. Very rarely does organization model fix anything. So as long as you've got teams and processes around kind of the data stewards, I talked about the business ownership side of it. I think you can make it work in any organizational model.
Stuart Rose: (
Yeah. I say that's a good point. And I think one of the things, like I say, knowing one of the thought of a chief data officer 10, 15 years ago is yeah, certainly a new phrase is out there, but I like the idea that you are beginning to think business. It aren't too silo, different departments in the past. They have been now they're working very much aligned. So sort of again, continue a little bit, you talked around. So the next question, what is a bit of two part scenario on that is certainly this idea of digitalization, as well as the word fuel. That seems to be the theme O of this conference out there. To me, it's created a new generation of data for the insurance company. So the first part really is what are some of those sort of, you talked about how your focus on some of the systems frame, what are some of the new non-traditional data sources that your organizations beginning to use? And the second part is how do you actually then go out and collect a lot of that new data? Because it doesn't sit within those mainframe legacy systems out there.
Joe Riesberg: (
Yeah. We don't use a lot of nontraditionals what I'd say. So it kind of depends on your definition but what I would say is we do use aerial. So you look at drone footage, you look at geospatial, 3d imaging inside. We also do internet of things. Data collection around make you safe, I think is one of our vendors and partners that we work with. And that's a human wearable device in a manufacturing space. And that can help with essentially kind of driving down premiums, risk improvement processes. We also look at moisture readers, refrigeration, internet of things, devices, but then on, from a risk improvement perspective on the underwriting exposure, that's more the aerial. And then on the claim side of the house we've started implementing a solution called Fri and we look at Experian and Carfax and those things, but I don't know, they'd call them nontraditional.
Stuart Rose: (
They probably are nontraditional. I say, okay, they're new data sources to me who would've thought IOT, the data you are getting through from a wearable device or some of those even how you would use that in insurance perspective. Yeah. I'd say five years ago. That to me is I would consider as nontraditional, I think.
Joe Riesberg: (
So the only other thing I would highlight is just social media on the claim side and fraud detection. Right. I think I would definitely classify that as non-traditional is we're always looking even on the commercial side of the house, looking at social media.
Stuart Rose: (
And again, maybe a imagery is not new data. It's been out there for years, but insurance companies haven't really, it's not so much they haven't taken the value is they just, haven't had to some degree the capabilities to use that data. So, I feel as I'm always trying to build off the last question here as I said, it's created this new generation of data how to discover new data sources out there. Cause there is so much more that's being generated and then determining, what that data is extremely useful. I think you get some examples, but there's lots of data out there. And we could question whether social media is useful or not useful, but some of it probably is doesn't really bring a huge amount of value to the business.
Prashant Nema: (
No that's an excellent point. So, the way I would like to think about data identification and data classification is almost like a matrix of data, which is serving financial needs versus underwriting needs versus loss management needs. And couple of others could put a couple of other areas, right? It's not to have a prescriptive, but that's one dimension. The other one is data, which is organically generated in your organization, which is a cross section across each of these and data, which is provided to you by your partners and data, which you add on besides what you created and got from your partners for enrichment, right? Just to give you a flavor of that is agnostic of your functional parts of your organization. It's more horizontal. Where, where did you get this data from and for what purposes doesn't matter, which function and when you look at it through that lens of that matrix, then when you're working with data, you start looking at, okay, we are doing this for financial reporting, but this is only data which we generated in our transactional systems, but what was other augmented data, which was added in this third bucket, which we used during the process for data enrichment or what came from the MGAs or in delegated authority, what came from elsewhere, right?
Prashant Nema: (
It helps you think through to make sure you're not missing out. And what are the opportunities you have, which you could be capitalizing on data. The other thing is there are two schools of thoughts. I'm not sure what is right or wrong as yet, but there's one school of thought that says collect all the data as much as you can, because you never know what data you'll need when, which is more of a inside out way of looking at it saying, just gather the data, the need will come. And we will figure out the need. The outside in is what are the needs is our financial reporting efficiency, which you're driving, or you want to improve your profitability, or you want to drive growth that drivers based on which data is collected and segregated and analytics created or predicted.
Prashant Nema: (
And then when you do that, you realize, oh I've got gaps. I'm missing this and I'm missing that and let's go collect that, right. Is there one right and one wrong? It depends on your organization. What you're doing, how large you are with geography you are. And I think it depends, but there are these two schools of thought in terms of collect what you can if you've got a good data governance in your organization and strong data management principles, as well as data analytic principles. And the company understands that I think the first school wins because now you've got a steady way of doing this in the company. Everybody's educated. Everybody knows that, but I think most companies are not there. The second bucket is the more obvious one, how it is practiced, and there's nothing wrong with it, where you go collecting what data you need based on your business drivers. And then those efforts and investments are funded because of the goals of meeting those business drivers.
Stuart Rose: (
I think that last point is to me, the important bit is you got to understand what is the problem we're trying to solve first of all, and then go out and get the data. Don't think you say, let's collect all the data and then see if we can analyze it to find the problem. You have a business problem first, and you go out, I think you talked around frisk and whatever else they go out there, they're looking at fraud detection. They will bring that data in to be able to resolve it. And I think even Carpay data, again, they're looking at what data sources can I help insurance companies improve that commercial underwriting side of things.
Joe Riesberg: (
There's a lot around that whole master data management concept too. It's your source system creates the data, but then the trail of it after I think that's a bigger challenge than most organizations even realize is the authoritative resource has now changed 2, 3, 4 hands down the road and a modern SDLC system development life cycle process does act accurately, correct that. So you gotta do SDLC plus essentially an ABC framework, an audit balancing can control. And that just creates. It's just interesting cuz as that data continues to migrate, it's, who's using it. Who has that? When we talked about data stewardship before, how is it being used? So I think to your point, you know, there's a lot around, that's why I thought that was a good concept.
Stuart Rose: (
And not to know, I think we could have a whole conversation around data, but I think certainly data, there's no doubt that it's bringing a huge amount of value to the insurance organizations, but there's also the case of you need the tools now to analyze that information, to make those informed decisions, as you point out to resolve those business problems out there. So Shaun, what are some of the tools or technologies that you think the insurance industry is now using to help with that automation?
Prashant Nema: (
There are two categories. The way I would like to look at it is one is more data management tools, right? And with the advent of cloud platforms versus in the past, how you thought about it. And in the past, people thought about warehouses and immediately, the next thing you've thought about is ETL tools, right? What is the ETL tool and how I'm gonna write this ETL process and when is it gonna run and how long is it gonna run all night long, all day long, all week long versus now you can stream data, you can do CDC and you can stream data as, and when it's happening in your transactional systems you need to think about that differently. And now it's possible because cloud based data platforms and ability to stream data allows you to do and think collect data in a whole different way.
Prashant Nema: (
But there is this whole set of tools which sit in this bucket around data management in the cloud and being able to stream data. The second bucket is the ability of creating data so that you have the needed data. So you are not going through an honorous exercise of people, either punching it based on what they're seeing and extracting it from excels. It is all about data extraction, where they're using intelligent data processing using OCR from emails and various documents coming in, or from APIs, which you're using for data enrichment. That's all part of the data quality and data enrichment, right? Making sure data when it's landing in your organization through your own systems, through your partner systems or through vendors who you're partnering with, who are giving it to you, that there's appropriate quality and appropriate enrichment. So you have good data so that it appropriately, when it lands, it is ready for use.
Prashant Nema: (
And now you don't have to go through a second round of cleansing and third round of cleansing or people have doubt and they don't trust that data, right? So data capturing and making sure you have good broad pools of trusted data and data management. And besides all of these the most important one, which we should not be forgetting is your data model of whatever you wanna do and how you want to represent your data. There's a lot being said. And in the last decade, a lot being done, people talked about big data and had, and got a little carried away thinking that is the panacea and how it'll solve things. But I still think data, you want to have analytics and finally want to analyze it doesn't matter where it came from structured unstructured. It is still, the analytics are very structured. You cannot take videos and videos can be analyzed.
Prashant Nema: (
The videos need to have metadata. The pictures need to have metadata before it becomes data. And then you can analyze and have appropriate analytics on which you could have insights and decisions. So having good data models, of course, there's been an advent in how those models were, there were relational models earlier. Now you can have data walls so on and so forth, but never take that for granted because eventually all data collected segregated has to be sitting in some kind of a data model, which will then feed your analytics, which can be then analyzing and getting the needed insights.
Stuart Rose: (
Okay. And I think that goes back to that first point, having that ETL is that data model helps with that data prep to be able to increase and allow you more time to do that analytics. So I think one of the comments you made was OCR, unstructured data etc. And I think that is an untapped opportunity out there for the insurance industry in that part of the automation. So, really when we think about it, there's so much of the insurance industry really does still rely on that manual process, whether it is a physical document, that's coming into the organization that someone has to read. So if you want automation, you need to have the tools to be able to do that. So, Joe, then where do you see then the greatest opportunity for automation in your organization? Is it in claims underwriting? Do you see it certain lines of business?
Joe Riesberg: (
Yeah. I mean, I'll talk about claims and underwriting here in a second, but I think that the most important, at least for most carriers that I talk to is that need for that customer 360. So when someone calls the service center, what policies do they have? A lot of times you'll call a claim center and they'll know your claims history, but they don't know the policies. They don't know the renewals. They don't know the premiums that yet they don't know the history of the account. Do they have a life account? Do they have a bond with you? I mean, I think that's one of the things that, that we've gotta do a better job of is really pulling all of that together. And frankly, a lot of that data is in different parts of the organization. And one of the things we're talking a lot about within technology DMC is around you.
Joe Riesberg: (
We have to be a data management, a workflow, and an API machine, right. Dawn are the days we're really kind of fully soup to nuts, right. In all the code, it's those three elements that we've gotta really take advantage of. And that gets us to that customer 360, because now you're able to pull together all those disparate areas O of customer behavior, customer information. So I think that's first and foremost. So on the automation side of your question, I think that's an easy one for a independent commercial carrier for us. That's straight through processing of underwriting. And that's small commercial that low touch, no touch where we can get really good in an industry or a six code segment that we know really well. We should be able to prefill. We should be able to kind of ask enough of those questions that pushed that right path an underwriter.
Joe Riesberg: (
And I think there was a conversation earlier around does that minimize the need of an underwriter? And I really don't think it does, right. The analogy I've seen that resonates with me, it's almost like an airline pilot, right. When you put it in an autopilot. Right. And I don't remember who, so that isn't my idea. That's just something I'm repeating and it's that repetitive task. It's getting it off their desk so they can focus on that middle that complex accounts. So I think that's the easy one, small commercial. But I also think on the claim side of the house you're looking for automated by default, right. And manual by exception only that's where we've got to get to. So I think those are the two big areas that I see from a carrier space.
Stuart Rose: (
Yeah, no I've heard the same type of analogy around the autopilot and the pilot. And I certainly agree is like, you are a part on the plane you don't think about, or a pilot in the middle of nowhere, but you wanna make sure when you land, I'm not sure if you really wanna trust the machine to be able to land that plane. Certainly if there's bad weather out there where you need then to focus on a very specialized underwriter pilot, to be able to do that.
Joe Riesberg: (
I think a lot of the new systems that we're utilizing the Workday for HR, you got Guidewire on the policy administration, duck Creek, they all have a tremendous amount of automation built in, but a lot of that is getting to a point where you've got enough of your product space on it, enough of your business process built out to be able to take advantage of that. So I think that's an impediment that I've seen is just trying to be able to get to that full platform capability and RPA solutions in general, robotic process automation is another area that for quick wins and that we've seen some limited success in.
Stuart Rose: (
So I know I got one final question for both of you, which own are my final one, but I don't know whether we want to open the floor up or any questions you have for Joe and Brant before we get to that section. So all gone quiet out there. I know it's probably the post launch session, so
Stuart Rose: (
I wouldn't say negative final question. We can go back to one of, but we talked a lot about automation, innovation out there for the insurance industry. So what do you actually think is driving this? Is it the technology? And you've talked around cloud, we talked around, well, we haven't talked, but there's artificial intelligence out there. Do you think it's the business that's asking for this for more automation? Is it the it department or do you think it's actually being driven to some degree by the consumers? They're saying what the insurance industry, they tend to be laggards compared my experience of going into a bank. I need to have a lot more automation. So gentlemen, which one wants to start off that?
Prashant Nema: (
Sure. I'll go first is, I think the insurer whether it's on the personal side or commercial side is looking for better ease of engagement or ease of use. Insured is not asking for automation. Insurer is saying, I expect ease of uses. It's just like all of us expect ease of views when we use the Amazon app to order the next thing every day. But Amazon cannot do that until they have automation. Otherwise there would be a 2 million employee organization versus whatever they may be. Right. So the ecosystem the end consumer whoever's the final one on the right side is asking for ease of use, ease of engagement the brokers, the middleman, the brokers, the MGAs, they're all asking for more data enrichment. They're also asking for ease of use. They work with all the carriers on one side, they Don want to do several things.
Prashant Nema: (
They wanna keep their costs down. So they're also asking for simplification as the demand is changing in turn from on the retail side and their consumers, the carriers are looking for more profitability, right? It's been a long time that the insurance companies had small profitable margins. And I think they are looking for more, we've seen a hard market after a long time which is almost tailing off. Now, they're realizing what is possible to some extent, which was triggered by the pandemic. But now they feel a lot more possible and they're seeing a soft market coming and they want efficiencies to make sure they are profitable, if not growth or with growth. So their drivers are, let's get this done and to get it done with keeping the profitability in mind is the tool in the arsenal is automation.
Prashant Nema: (
You do that with automation, you do that with data in your pocket. So you're more data driven as you're making these decisions for which lines, which are the good customers which will give you the needed profitability. And how do you score your claims and how do you fulfill your claims quickly and not sit on them too long. They all wanna be data driven. They want to improve these of views, but that is only possible by doing more automation. So the market dynamics are driving the need for automation, which is driving the need for better transformation, because finally everybody is seeking ease of views on one side and being more profitable businesses on the other.
Joe Riesberg: (
Yeah. Well said,
Stuart Rose: (
Can you answer full of that?
Joe Riesberg: (
Yeah. What's interesting. I would tell you is COVID was a growth stimulant for a lot of this too. I mean, that's the easy answer. But I also think and we do surveys of all of our independent agents and ease of doing business, which is what you're a lot of your talk. A lot of what you're saying is, is hitting on, but there's also an element of relationships. Because even in the small commercial space, 80% of the small commercial insurance in the United States is still owned and managed by an agent. Right? And we think that that's should be, to me. I always thought that number should be drastically different. You should be able to go onto if you're a florist, right. And go onto a website and pick your coverage. But insurance is a complicated product. And as a result, there's always that need for that agent interaction.
Joe Riesberg: (
And so the more you can still build a relationship with those agents and those agents with those policy holders, it may not be their favorite part of their day to go talk about insurance for their business. But at the end of the day there's a need there and so I do think the policy holder is driving some of it. The agent drives a lot of it. So when we talk about our customer, a lot of times it is the agent specifically. And in that case, as you're talking about agency management integrations, and those are still painful, right? So I think those are the pain points that I think are fueling some of these conversations and where it's coming from,
Stuart Rose: (
your point around.
Stuart Rose: (
Extra lines, the relationship with the agent like that. I totally get, I think people underestimate the value that brokers and agents bring to the business because you think you are an automated, you wanna make it better, but insurance is a very complex product out there. It's not everyone gives us the experience or trying to get analogy. That's like gonna Amazon. It's not the same, cuz I think in many instances, the last thing you want to do is you find out when you have the claim that you are underinsured because you didn't take the advice of an agent out there. Great. So, but me just selling that point again open the floor up. Any other further questions we have out there got one at the back there put the hand up challenges.
Joe Riesberg: (
The question was around frisk and some of the benefits we've seen around. So we're using it primarily in our legacy systems for claims fraud identification. We have had challenges implementing with our newer systems. We've had challenges with some of those external integrations. So I talk about integrations and workflow. That's been some of the challenge. Overall fraud identification is just such a important part of business. And I do think they have a good solution, but there's a lot of good solutions in that space. So I don't wanna be a frisk commercial, but it's been a good solution for us
Prashant Nema: (
If I can add a point to the question is in addition to what Joe said is
Prashant Nema: (
When we are doing these kind of transformations, it is almost touching the day to day people who are doing the day jobs, right. And never take that lightly because people have to take time out of their day jobs outside of the data management team, whose job is to do this. But if you work with the finance guys or the actual guys or underwriting operations, whoever they are, they have their day job to do. And if this is gonna carve out anything more than 5% of their time, which in many cases in the early days, definitely it is that is someplace where people shortchange themselves and almost trip almost always and people can keeps on getting kicked down because we don't have time right now. We're busy with that and busy with that. It sounds a very non-technical aspect, but it is the single most biggest aspect where either you'll do a sloppy job at it or you'll not get the initiative done. And in general, the initiative itself looks like, oh my goodness, it's been running for too long. It's difficult to do difficulties in making sure all the people in the company can spare their time to participate in it.
Joe Riesberg: (
And it is goes back to that whole workflow, right? So how well is it integrated? So if every time they're going through a claims process and they have to kick to an outside different software solution, and now you're running a Lexus nexus report, and now you're running a Carfax report behind the scenes to your point, that's adding time PR and it kicks you out of the system of record that you're working within within that workflow. And we're seeing it on the underwriting side too. So we're not using it on the underwriting management on that fraud side which is I think a big component of a lot of those fraud solutions. And it's the same concept as you're kicking out, looking at better view, you're kicking out, looking at aerial images. And so every time you get out of that workflow process, it you're slowing down.
Joe Riesberg: (
You're making things and technology should be about efficiencies. And yes, it's probably a more holistic picture, but if you're taking that from a 10 to 20 minute assessment to a 20 to 30 minute assessment, and now you're not able to process that new business or that renewal the way you were, that's not improving lives of your underwriters. That's not improving lives of your agent or your policy holder. That's actually slowing down the process. So you have to be very diligent around looking at those integrations to make sure you're utilizing them in the way that truly adds value.
Stuart Rose: (
I think probably we should rename this session. Data and integration is a transformation to automation because I think that is the companies that can get that integration. The APIs is probably, I think probably gonna be the bigger game changer out there. That's gonna be out there. That's gonna drive new automation aspect of it. So I think with that we'll be really to run out of time again. I appreciate Shaun your time, Joe. Thank you again for joining us for this session.