Key Takeaways:
- Advances in analytics through AI and other technologies
- How to add new data sources to processes efficiently
- New security and cloud best practices in an increasingly distributed world
Travis Bethune (00:10):
Customer and distribution space at our Firm. And for those of you that may not be terribly familiar with Berkshire Hathaway, it is an organization that owns 73 operating companies and it is divided into four or five categories. If we think about insurance businesses, there are 12 insurance businesses, and I represent one of those 12. We have got rail, utility, and energy business. There are 14 entities in that space, 23 manufacturing organizations, and about 24 service and retail organizations. So Berkshire Hathaway has a lot of things. We are going to focus obviously on the insurance businesses. I represent one of those 12 businesses. We are Berkshire Hathaway Specialty Insurance. Ee are a large customer specialty underwriting business. We are about 10 years old as an organization, so we are one of the younger Berkshire Hathaway Companies. But if you are familiar with Geico Insurance, I mean that is probably the most notable brand from a consumer insurance perspective within the Berkshire Hathaway family. But we have got Reinsurance Business, Commercial Business, and the like. So thanks for having us.
Martina Conlon (01:17):
Great. So it seems like our AV issues have been resolved, maybe.
Panel Member (01:23):
Yep.
Martina Conlon (01:35):
So I would like to, before we engage in a conversation and hearing more about Travis's journey with data and things like that, I want to talk a little bit about what is a data strategy and how do you go about updating that data strategy? So over the years we have had many, many customers come to us and say, I need a data strategy. And we would, oh, perfect, thank you. And what we would primarily do then is say, well, what do you think a data strategy is? And many times our clients or prospects would come to us and say, I am not really sure, but my CIO and my CEO tells me I need a data strategy, so I need to put something in place. Well, from our definition, a data strategy really is a plan on how you are going to deliver data capabilities that will allow you to meet your business goals or exceed your business goals and it is data and analytics capabilities.
(02:33)
And there are seven dimensions around that that we are going to talk a bit about in order to understand that you are addressing everything that needs to be put in order to deliver those capabilities, whatever they may be, and order to meet those business goals. Again, one of the top priorities is business IT alignment in terms of business data alignment and the fact that we are not doing the old, I spent the nineties building insurance data warehouses and we very much had the opinion that if you build it, they will come. That is not the opinion and the best practice of today. And so data strategies really encompass an awful lot more. They encompass more than data warehousing and the architecture. It encompasses organizational issues, data fitness, and different things that we will talk about. So Insurance is pretty straightforward and really there are only three levers of value in the Insurance Industry.
(03:30)
You can sell more, you can manage risk better or you can cost loss, you can cost less to operate. And those three things are typically the areas where companies, their business goals are oriented around those things. And as you know, data is absolutely the foundation of all of these three levers in terms of being able to sell more, defining new products, understanding the demands and needs of the market, what your target market is for costing less to operate, being able to understand your processes and fine tune them, put implement predictive models in order and levels of automation that do require advanced analytics and managing risk better looking at analytics and predictive models around managing fraud, understanding loss costs and things like that. So really data is at the foundation of all of these three levers of value. When we talk to Insurers, we do a survey at the end of each year where we ask insurers, what are you investing your money in? And you can see here in this chart that BI and Analytics is the number one thing that the business stakeholders in insurance organizations want it to deliver or want to be delivered, whether that is by IT or standalone data and analytics team or an actuarial team. But BI and analytics is extremely top priority for organizations today.
(05:00)
We also see that that very much translates into building out data warehouses, building out data lakes, putting in place those data-centric projects in order to stand up the data to be able to deliver the capabilities to the business users. But as you see here for mid-size P, and this is just one example for mid-size P and C companies, that the top priority right now is still implementing core systems. And as an organ, as a industry, we are still charging through standing up new modern configurable solutions. But defining your data strategy as part of your core system strategy is going to position you for being able to get the benefits of that modern system through the insights of the data that you are taking out of it. And so very often there is a BI and Analytics effort in parallel with your core systems so that you are leaving this project with a modern solution that you can gain insight from.
(06:06)
Also, property and casualty insurers are generally taking advantage of InsureTech opportunities that we have out there. This is our research on emerging technologies and the emerging technologies that insurers are leveraging. On the top couple of bars of the chart, you will see that there are a lot of AI enabled capabilities. So text AI with instruct unstructured text such as text ingestion, different things like that, imaging AI, so interpretation of images for claims for inspections, those types of things. So those categories of AI that there are fairly good adoption rates there. If you notice for mid-size in P and C insurers in general are, a lot of those are being deployed in conjunction and partnerships with Insurtech companies. There are some insurers that are going out and leveraging AI, horizontal AI technologies in order to do innovative and interesting things. But there are a lot of companies in the Insurtech space that really can deliver a lot of value to the organization in order to help you manage lot manage risk better, improve efficiencies and understand your market better.
(07:29)
So let us go back to talking about for making these business goals and for achieving a data mastery. You are a strategy, as I mentioned before, really is the plan on how you are going to deliver these capabilities and be successful. We work with a lot of our clients around data strategies. We come in, we do a current state assessment, and then we build out blueprints and roadmaps in order to do that. A key piece of that is that assessment at the very beginning in understanding where you are before you define where you are going to be and how you are going to get there. And that is the strategy piece of it. We go through, we have something called the insurance data and analytics maturity model, which is an assessment framework to determine where you are strong, where you are behind, where you are in different categories that will then help you or help us, help you define what that data strategy is, which generally covers in our experience generally covers these seven aspects. So with the data maturity model, the data strategy really is made up of more than just, let us see if I can, is a little pointer.
(08:58)
Okay, there we go. Right, thank you. What most people think about when they think of data strategy is the architecture and technology piece. Is there a data lake? Is their data marts and a data warehouse? But even beyond that, what is the technology that is being used? Our cloud platforms being used, and this is an important aspect of your data strategy, however it is involved. Even the architecture has more to do with is validation occurring in your core systems? Is your data in a format that it is granular enough? Are there different things like that that have to be considered around the architecture and technology? So that is one piece of it. But additionally, data innovation. Are you including data initiatives as part of your innovation programs, assessing data opportunities on an ongoing basis? Organization and leadership is a very, very big part of a data strategy. Do you have a chief data officer?
(10:06)
Is your actuarial chief actuarial person the chief data officer who is the leader of data in your organization? Is your data and the analytics team reporting into it? Is it reporting into operations or some middle ground or some business unit? Do you have business data and analytics teams embedded in the different business units? And so it is really understanding the impact of that organizational placement of data responsibility and data analysis and data creation. Additionally, there is the aspect of, I do not have my glasses on. Can someone read that to me? Data Governance. Oh yeah, which is an incredibly important aspect of it in terms of who are the data stewards, who owns the data, what is the structures that you have in place such as a data council or a data governance team, what are the processes you use to adopt and change the definitions of data?
(11:16)
And finally, what are the tools that you are using to manage the understanding of what data is where in your organization, what has been classified and not classified as PI? What tools are you tracking those kinds of information in? And then finally, data fitness, which has to do with data quality. This is where the level of granularity, the timeliness of your data, the whole aspect of is your data fit for purpose, for what you are going to use it for? And then finally, information utilization, which is your data consumable by your organization. Are there people who, do you have self-service? Do you have other aspects of the way that delivering data to your actuarial teams, your business units, your underwriting teams, the teams that manage your distribution? And so these are the aspects that we consider part of a data strategy and that pulling these aspects and dimensions together in a plan in order to determine how you are going to put all of those aspects in place in order to achieve that data mastery level within your organization.
(12:34)
And so with that, I just wanted to go through one very quick example of how data has data and analytics has really impacted our industry in the last 20 years. And this is actually a slide from the next session that we are going to be doing on intelligent decision making. But if we look to the left, the ty this is all about underwriting and how underwriting has evolved over the years. If we looked to the left, it used to be that underwriting teams spent a lot of time gathering data, a little bit of time validating that it is right and then a bit of time analyzing it in order to understand the risk, the pricing, all of those types of things. And it was a much more manual process because we were getting all of our data from our agents. We found once third party data started to become available and automated in our processes, we found that we were starting to use less time gathering data and a lot more time we were getting more data and suddenly we had to spend a lot more time analyzing the data that we had gathered.
(13:45)
Finally, with the introduction of predictive models, data presentation tools, those types of things, we are finding that amount of time to analyze the data is certainly going down. And now if you look over way over to the right, when we think about AI technologies, the third party data, predictive models, intelligent text ingestion, image recognition, all of those things where we can get intelligence out of different sources, the amount of time to gather is minimized. The amount of time to analyze is minimized because hopefully in the future there will be more assistance from AI enabled technologies in order to do that. So that is just an example of how data and analytics has made such a big difference in the underwriting space. But this is true across the insurance value chain too. So with that, Travis, if we could talk a little bit about the data journey at Berkshire Hathaway and what where you are from your data capabilities and meeting your business goals and kind of a summary of where you are.
Travis Bethune (15:03):
Yeah, happy to offer that. And I will just offer some context and to remind that the Berkshire Hathaway specialty, again representing one of the 12 insurance businesses, we are a large customer specialty commercial underwriting company and we are on the build. So we have, we have been on a journey for the last 10 years and having been in the business for 30 years, I can clearly state that having the right insights on the front end makes a real market distinction on how we perform in the marketplace against our competitors. I say that because when our doors were open in the early 2000 thirteens, our leadership had keen insight to try to build an organization that was different from the places that they came from. And a lot of the organizations that our leaders came from were a 100, 150 year old organizations that had a traditional mentality about receiving information, processing it, and making it consumable usable.
(16:00)
We had an eye for that at the front end. And so at Berkshire Hathaway specialty, it is a compliment of both internally developed data and externally sourced data from third parties. And we really try to make the most sense of that so that we can be as effective and as useful and as timely with our customer audience as we possibly can be. we are a very customer-centric organization and I think you make a differentiated impression with a customer, be it a new relationship or an existing relationship where you can demonstrate that you have done your homework access to a really good data strategy and the consumable outputs from that data strategy really helps set you apart from your competition, the ability to respond quickly to a customer's need, especially if in the large commercial customer space, if any of you have served large customers from a risk management perspective, you can appreciate that the risk landscape is particularly dynamic. It is moving at a rapid pace. There is high severity involved when it comes to first party exposure. Think about wildfire earthquakes, windstorms catastrophes, think about consumer level exposure. When you think about if you are buying your automobile insurance and what does it mean for you to get the best rates or not based upon your driving record. All those things seed into making a differentiated decision that hopefully the customer can understand. And again, the good outputs from a good data strategy just help you. It gives you one market advantage against your competitors.
Martina Conlon (17:33):
So as you have been building your data capabilities over the last 10 years, well, actually let me back up. You talked a lot about the customer and understanding the customer. Do you view the agent as customer or the pol, the prospect and policy holder?
Travis Bethune (17:52):
Martina, It is a great question. Our team really uses the term customer broadly. We are referring to our intermediary distribution partners as our customers because we serve them on the front end and they provide us access to the end customer. So we think about both audiences in that respect. The interesting part about the distribution customer is that we are also their customer in a way. They have gotten equipped and incited enough to be able to sell us data. And the value of them selling us data is that we get access to better understand their customer portfolios. And so they are our customer in a sense, but we are theirs as well. And them selling us that good data, again, helps us be more prepared when we are calling on them, asking them for business and being very deliberate about the opportunities that we are most interested in, distinguishing those from the ones that we are less interested in.
(18:49)
And it just makes more efficient delivery for lack of better terms. The customer audience is the end buyer. The interesting part of that dynamic is with good data and good intelligence, oftentimes they learn from us in certain respects with respect to how they manage their risk and delivering insights that help them make informed decisions on what to do. A lot of customers at a pivot point on whether they want to buy more insurance or self-insure to a higher level, depending upon the line of business, the product, whether a certain product is firm in terms of rising rates or whether another product is soft in terms of reducing rates. All of that combines just helps them make some more informed decisions and they are accountable to somebody, whether it is a treasurer, a chief financial officer, legal counsel in terms of that risk management repro that how much they want to assume on their own, how much they want to buy the data outputs and the decisions that follow those, I think really help them make more incited decisions.
Martina Conlon (19:49):
So it sounds like the insurance industry as a whole traditionally has been buying data from LexisNexis or Verisk. It sounds like you are, have some innovative ways of getting data from your brokers. Do you also get them from your reinsurers, what your, we do not want any special sauce, we do not need to know any underwriting secrets. But have you found that you, you have found innovative and interesting ways to get gather data that maybe parts of the other parts of insurance have not?
Travis Bethune (20:28):
That is a good question. I do not know that I have a good answer. I could say this to say on the broker side, we are getting a lot of customer data and intelligence and we are also buying industry centric intelligence in terms of how certain industry sectors on the customer side are performing, getting insights in terms of the broker community is really good at collecting data and benchmarking data. And the broker side of the business, the distribution side of the business has really evolved to do a really good job at providing insights on where the customer demand is heading. So from that, it is tremendously valuable. When I think about data development on the internal side, we have taken tack with certain lines of business and certain industry sectors to gather as much data over a period of time as possible. And the reason for that is it is helped us develop proprietary rating schemes that help us be more competitive. And those industry sectors are line of businesses where we have got enough insight over a 10 or more year period to understand what the products should be priced at, how we should term it in terms of limits, deductibles, etcetera. And that will continue to evolve the more customers that we get access to, the more submission activity we see we are developing, we are continuing to augment that data set to be more insightful. So it is a real blend of the two that hopefully helps us get to the right outcomes. Right.
Martina Conlon (21:53):
Well, I think we are just about running out of time. Does anybody have any questions for Travis or myself? Sure.
Audience Member 1 (22:04):
What coming up in the, I think keeps coming up in the work that we are doing. We do a lot of data transformation stuff. Is data quality is generally craft, and I am just curious what your perspective before you start processing data, what are you doing? What kinds of tools are you using? How are you cleaning it up? If you use imperfect data and you feed a machine learning rule, you are going to get an imperfect result. AI will be worse. And if you do not, then you are relying on your underwriters to interpret the data the way they will do it. So you are kind of damned either way unless you clean it up before you start playing with it. Anything, any insights or tools or observations there?
Travis Bethune (22:42):
Yeah, you, you have thrown me a softball question and I will try to hit it. I think the quality of what goes in affects the quality of what goes out for us. And I will use the example I have used for the last few minutes. There are a lot of brokers that want to sell you data, but it is not of kind in quality. And so for us, we will take a preliminary evaluation on the strength of that data before we decide to sign the contract and receive the data downloads. And in many instances, brokers that are pitching to sell list, they will take feedback and figure out how to augment that so that it serves what we are trying to accomplish. So we do get a little bit of front end input and assessment on whether we want to buy or not buy. And in some instances we buy and in many instances we do not.
(23:28)
I would say one time out of five we say yes to the brokers that are pitching the data. And with that we have to take a real keen eye in terms of how we use that data. And that decision is not scoped just to an underwriter community's opinion. The data serves not just the underwriting team, it serves the audit, the actuarial team, it serves the claims team. And so we get input from all three or four of those business groups before we make the decision on whether to buy and integrate what, what is coming in from the outside. Does that answer your question?
Audience Member 1 (24:02):
It sounds like it is still a brute force and that
Travis Bethune (24:05):
Call it what you owe, we are just trying to make a good sense of it on the front end. Yeah, that is
Audience Member 1 (24:09):
What most people are struggling with. I mean, there are needy tools out there. You can run Redshift, snowflake, or Kafka tools, but you are dealing with massive amounts of data and data transformations. Just massive problem
Martina Conlon (24:21):
Chat GPT is going to solve our problems, do not worry about it. Not this one. No. Alright. With that, I think we have to wrap up. Thank you everyone for joining us and have a great afternoon.
Travis Bethune (24:31):
Thank you, thanks Martina.