The InsureTech Geek Podcast powered by JBKnowledge is all about technology that is transforming and disrupting the insurance world. We will be interviewing guests and doing deep dives with our own research and development team in technology that we see changing the industry. We are taking you on a journey through insurance tech, so enjoy the ride and geek out!
JAMES: Welcome everybody to episode 4 of the InsureTech Geek Podcast. I am your host James Benham, and I am joined this week, by the illustrious R & D chief from JBKnowledge. JBKLabs. Graham, Leslie. Graham. How is it going today?
GRAHAM: Hey James, I am doing wonderfully.
JAMES: And Graham, you are doing a little bit of a flip flop. You are in the studio and I am in my summer studio.
GRAHAM: I know. I am in the captain’s chair. Big shoes to fill here.
JAMES: Watch out. Watch out, Graham. Taking over, man. That is awesome.
GRAHAM: The pressure’s on.
JAMES: Yeah, the pressure is on. It is good to have you on the show. You and I have been working together for years, but it starts being a regular thing with you being on talking about different topics. People have a hard time wrapping their brains around some of these topics. They talk about them; they listen to them, they make big multimillion, multi eight–figure, nine–figure, even million-dollar decisions around a lot of these topics and they just have a tough time even explaining what it is. And so, we are going to help them by explaining what these topics mean, what it means for them, how it impacts the InsureTech space. And, I can think of no better person than you to help me explain all of this Graham.
GRAHAM: Oh, I can talk about technology all day, and this is a technology podcast, right?
JAMES: Absolutely. And it is a tech company. We love talking about tech and love geeking out. So, let us jump right in and just talk about BI, stands for business intelligence. Graham, what does BI mean when I say the words BI? Business intelligence. What does it mean to you? Just a basic lay definition.
GRAHAM: So, okay. To me, business intelligence is, let us look at it from the perspective of an insurance company, right? So, if I were to be asked at an insurance company what is the most common injury cause for a claim? I would probably know that off the top of my head, but what if I’m asked, hey, what are the most common injury causes for claims and this particular state, and this particular quarter of this year, I may not know off the top of my head, but business intelligence is tooling to be able to figure that out at a couple of clicks on demand without needing to delegate to that to some team or take a lot of time researching that.
JAMES: And historically, the way that companies would answer that question is through some type of report writing process, right?
GRAHAM: Yeah, exactly. Delegating that to some team goes and does the research and more to some IT department than that goes and writes some code to figure that out, write some queries.
JAMES: Yeah, so knowledge of sequel, kind of important back in the day, to run these types of data reports.
GRAHAM: And not just back in the day, but still today for a lot of companies.
JAMES: It is. That and Excel, right?
GRAHAM: Yeah, exactly. Oh, Excel. How we love that.
JAMES: Ah, dear Excel, the cancer of modern business. Excel is wonderful in many ways because it allows you to answer questions quickly, but it is terrifying in other ways because it allows you to incorrectly answer questions quickly. That is always my hang-up, Graham, is that Excel is a great quick data analytics tool, but all too often you and I see it ended up becoming an enterprise workflow solution, right?
GRAHAM: Absolutely, and especially when you have one question you can answer across a thousand different spreadsheets for the data spread of class all of those.
JAMES: Yeah. And then you have a big data integrity problem with Excel because you have tens of thousands of sheets, and how do you know they are all properly synced and how do you know their current, and how do you know which version is the most current? And you can get yourself in a pickle if you are answering critical business questions by dumping it into thousands of tables and thousands of sheets in thousands of Excel files, right?
JAMES: So, what was the predecessor to BI? So, I mean, we have this back in the day, which was a Wednesday, we ran reports by manually compiling information and using calculators and adding tables, and slide rules, and then we jump into software and we have like the first spreadsheet that would, would you argue the first spreadsheet is the beginnings of digital BI tools?
GRAHAM: I think it depends on your definition, but yeah, I think so. Absolutely.
JAMES: Yes. So, VisiCalc is the first major spreadsheet that rolls out, still operates. You can still download VisiCalc today. It will still run in an MS dos simulator it was a DOS-based spreadsheet tool that started to show the power of being able to import and correlate data. After the spreadsheet, and before modern BI tools, we had this phase of report writers, like SQL server reporting services and crystal reports. So how would you compare and contrast what reporting services SSRS can do and crystal reports, compared to a modern BI tool like we are going to talk about today?
GRAHAM: Yeah. So, those reports systems are classified today are what we call traditional monolithic reporting systems. They are enormous tools that combined all the different aspects of the analytics process into one big package that required some pretty knowledgeable folks to use. You had to have a lot of knowledge of these systems to be able to do the full process of the analytics, which is, taking all your different data sources, running them through an extract, transform load process into your system, and then building your reports and dashboards. And that was the predecessor to what we have today, which are more modern extraction and data platform services.
JAMES: Yeah and this is back in the day when you had to know a lot about the data structure. And so, let us say you are an insurance expert and you are an underwriting, or you are in claims and you are trying to find what is driving my claim cost. You are going to have to know how your entire claim data table is structured. You are going to have to know to write a lot of SQL. You are going to have to extract a lot of information and then import it. Then massage it, then write a bunch of reports, then output those tables into some type of usable format in reporting. Crystal reports and reporting services, these were powerful tools. And, back when I started my career, they were a hot item. They were a hot commodity. But when you look at how much work went into building them, and you look at modern tools, there was a pretty big gap for the average everyday user was not there?
GRAHAM: Yeah. And that is not something that was addressed until very recently.
JAMES: Yeah. So, let us just talk about, modern BI then. Modern BI is about giving users easily usable data analysis tools that can help them make better decisions. They are making better decisions; they are going in and they can form their dashboards. What you are seeing is instead of like one-time, one–off reports, real–time dashboards, and BI systems, that will allow you to dig into the data behind them. So, it is a little bit of a shift in thinking, to the process of saying instead of creating this report that runs every Monday at 7:00 AM and emails to people on a PDF, we are going to move to real–time dashboarding that pulls in even more data elements and then tries to help us make better decisions. So, what does that look like?
GRAHAM: Yeah. And what it is moving from is enabling IT professionals to create these kinds of reports to enable the actual business user. And today there are 2 categories for these tools. There is what we call extraction tools. These are tools like power BI or good data or Tableau or Sisense. And what they do is extract all of your data from your data sources, and then these tools run the whole process of the TL. The reporting, the ability to create those reports and dashboards, and then, some of that is often overlooked, but pretty important is the authentication piece. If you are not going to print these out as PDFs and distribute them, how do you manage who has access to the right reports?
JAMES: Yeah. And you kind of like ran over it like a little speed bump in the road. Well, let us go back into, wind back and talk about ETL for a second because ETL is a big deal now with insurance carriers and TPA’s, we are doing a lot of ETL work for them. We see a lot of retail work going on for clients. Extract transform load. This is a concept where you are extracting data from systems, you are transforming data by modifying it, cleaning it up, and then loading it into a different system. One of the popular platforms out there is MuleSoft ESB, which is a popular ETL tool, but there is a bunch of ETL tools I mean, fundamentally SQL server integration services as science as an ETL tool too, right?
JAMES: And so good BI systems that we have been evaluating. And just for listener land, we did about a 3- or 4-month evaluation project where we took a bunch of big modern BI systems like power BI and Sisense and Tableau and Looker, and we ran through the paces. We got actual working copies, we built full systems with them. We did full piloted prototypes, and this was one of the ridiculously big to quote Derek Zoolander, ridiculously good looking features to some of them and ridiculously horrible features to some others were the ETL tool they used because there were some systems we use that was really complex to get even the most basic of information into, right?
GRAHAM: Yeah. And if you are an insurance enterprise, you do not have just one database like a software company might. We have lots of different databases and data sources. You have got the information across lots of different types of databases. Your IT department has set up different tools you are utilizing, and that the ETL processes is pulling the data out of all those different sources. Making it fit together in a way that it does not natively do cause these companies do not necessarily integrate, and then loading it all into one place where you can commonly query data across all these different systems in a uniform way.
JAMES: But just to show you where the rubber meets the road, we had cases where data imports that on one system, let us say, and one of the easier ones to use as power BI, right? They have got a pretty easy to use, a fair dummy proof import system. And the other thing that power BI has is they have a lot of hooks into common public data sources that are already prefabricated. So, you do not have to go and build your data import. You can say, hey, I use this, just point over there, and get the data. So, we would have an import that would take an hour or an hour and a half to process and get into power BI and some of the other systems we tested, that same import would take what, wasn’t there one case where it took like a week to get the data in?
GRAHAM: Yeah. And that is one of the interesting things about these tools. You have got some very new ones and some tools that have been in the market for a long time, very mature, but just have some slower processes for how they do things. It is a lot of pros and cons across the marketplace.
JAMES: Yeah, we did our pilot with insurance claim data too. And so, you are doing some, just fake data, by the way, it was not any real data, but we did a bunch of fake data imports, and the problem became with field names, field blanks, special characters. We had to do a lot of manipulation to get it into some of these ETL tools. So, if you are out there looking at BI or thinking about business intelligence, first consider there is a sea of difference. There is a huge difference we see between some of these BI tools on how easy or difficult, how expensive, or inexpensive, how labor-intensive, or not it is to get data into these systems. It can be a really great day nightmare or it can be a cakewalk, and so that that has a huge impact because if, if it’s easy to get data in this system, it’s easy to parse complex data sets, then it’s much easier to use the system. So, moving on from ETL since we, I kind of went back and like ran over that speed bump again. Let us keep talking about BI. What is a big differentiator between a BI system and a reporting system? Once you get through the ETL phase, your data is in, what is different that is next?
GRAHAM: So, one of the big differences is building out those dashboards. Your reporting systems let you build out a simple report, but a dashboard is much more than that. So, if you want an interactive dashboard showing you the status of a whole bunch of data altogether so, you can at a glance understand how some aspect of your organization is running, that is a lot of different reports that you tie together, integrate and provide, and then you want the ability to drill down into those. So, if you are looking at some status of your claims, you want to be able to see claims within a certain state or, to do with some particular causing injury, the dashboarding tools tie all that together, and then you can delegate access to the right folks, so people see what they are supposed to see.
JAMES: Yeah. And what for me was one of the killer features of BI Graham, I saw it the first time I saw it, I was like, oh my gosh, this is a killer feature. I have to have this. In particular, when you are like analyzing claim data versus policy data versus payroll data, let us just say that. Or you are trying to cross-reference information and see cause and effect, and you’re trying to see, hey, when I change a variable over here, what happens over here, in really good BI tools, there are some queries in power BI and that was the first place I saw it, but not the only BI solution that has this though. You can cross–reference your different widgets on your dashboard, so you can say, hey, if I click on this element over here of let us say, average days out or let’s say whether there’s indemnity only or medical on work comp claims, or let’s say a state, I want to click on a state.
The cool thing is you can cross–reference data widgets so that all the other widgets around it updates. So as soon as you click on–on a state like Texas, it will automatically filter all of the other elements on that page if you configure it that way. You don’t have to configure it that way, but you can so that you’re only seeing that same dashboard applied to the subset of data now, which is where to me, you get into the power of BI because you’re able to ask a question and then get an answer. You can say, I want to just pick this element over here of claimant state and just pick Texas, and then, I want to see all of the other data widgets I built on this dashboard just for claimant States in Texas. I do not have to go write a new report. I do not have to go build a new dashboard. I can just click that element and auto filters everything else. That is a pretty powerful part of BI, huh?
GRAHAM: Absolutely and the dashboard consists of all the questions you want to ask, so, it is not just asking one question at a time, but here are all my questions and let me change how I want to feed data into this to ask all these questions at the same time.
JAMES: Yeah. But then then there is a step further with asking questions. Cause if you had to lay it out a simple elevator pitch for BI, it is an easy to use technology system allows you to ask questions and get answers for. That’s it. Whether your questions in a form of a widget or in some of the BI systems we worked with natural language processing, they have an NLP engine, or you can type natural language queries, not SQL queries, but natural language queries and get answers. That has been something we have started to see creep into a lot of these systems.
GRAHAM: And it works too. That’s what’s cool about it is it’s not just some vaporware, you can ask it, hey, show me, show me the number of claims in 2017 by state, and then it will do just that if you have it set up in the right way.
JAMES: Yeah. So, you have to hint your data properly and you have to name your fields properly. There is some setup that is going to be done, but when you do it, it gives you the ability to ask natural language questions. This is something we are seeing with some of the cloud-based search solutions like Amazon elastic search, are they give you an NLP search engine as well. The very first time I used natural language search, and I am going old score right now. Do you remember ask Jeeves?
GRAHAM: Yeah, barely.
JAMES: So that was an early competitor to Google. I know you are not a Google fan. You are your duck–duck go, right? I am a little thing called privacy, but that is another conversation. Yeah, you can Google Duckduckgo and Google will return results for it. They do not block it, believe it or not, but Duckduckgo is a search engine that does not track you. And, Mr. Graham is one of the only people on the planet I know that uses it.
GRAHAM: It works.
JAMES: It works! So, when I started using ask Jeeves a long time ago because it was a natural language search, it gave you the ability to search by just typing in plain English. Like, how many widgets did I sell on December 3rd of last year? And that is what it allows you to do is answer those without having to structure your question in a way that is not natural to the English language. And this also lends itself extremely well to voice search through solutions like an Alexa skill, because when you are speaking to Alexa, you are not speaking in SQL queries, right? You are speaking in natural language.
GRAHAM: Yeah. And that is one of the integration points that these new platforms allow for. Talk about ease of accessing your data. What previously took a team of IT professionals, now you can just ask it over your phone or Alexa enabled device.
JAMES: Yeah. So, there is a bunch of stuff that has to come into place to make that happen though, right?
GRAHAM: It is not a silver bullet, right? There is a nontrivial cost and effort to build these types of solutions, but what you get in the end, it’s just when you put the work in upfront, you get something that’s so flexible that you can access your data in a natural language way like that, which is incredibly powerful for the decision making process.
JAMES: Yeah. So, a good BI solution has natural language query ability. It is got good data visualization right, charts, graphs, any type of way to visualize your data. And we are seeing some new creative visualization tools come out through BI, aren’t we?
GRAHAM: Yeah. I mean, more charts than you would think existed. What is neat is some of these solutions will even suggest, hey, this is, of course, you can do this as a bar graph, but there is a particular graph that is suited to this kind of case of what you are queering for. And they will recommend, this is a better way to visualize your data. power BI will recommend graph types to try to improve your visualization experience.
JAMES: Yeah. So, you are getting into data intelligence, not just data reporting. When your software starts making recommendations to you, your software has transformed itself. It is saying, hey, look, there is a better way to show this information that people will understand better and then, the result is better decision-making. That is what your goal is, is for adjusters and underwriters and insurance exacts and brokers, carriers, TPAs, PBMs, to make better decisions more quickly using data, not their gut.
GRAHAM: Exactly. And across all the different parameters that that data provides. It is not just underwriting or anything like that but even thinks about like fraud detection. You have got the data there to be able to start realizing the trends and claim fraud. Very interesting opportunities.
JAMES: Yeah, and in some of the cutting edge InsureTech firms that are not vendors to insurance companies, but our tech companies have become insurance companies, we are seeing a lot of that being put into play. In particular, and you just hit it the nail on the head, fraud detection. Especially fraudulent claims for personal lines. We are seeing some fascinating new technology be deployed that says, hey. 99% of the time when someone says this or does this, even the way you collect data on a claim, you can ask questions in a particular way and ask the same question in multiple different ways that will very easily root out fraud. And so, you are seeing a much lower fraud incident rate. I mean, there are some fascinating, consequences of that. So, let us dive into the individual solutions because we talked about, and we read, we researched a bunch. The top solution and we will just go ahead and bottom-line upfront, top one that we liked, and this are for embedded analytics, by the way. Embedded analytics is when you take an analytics suite and you embed it in your application. It is not standalone. And, and power BI has both standalone and embedded. Some solutions are embedded only, and then other solutions are standalone only, like their desktop, right?
GRAHAM: Yeah, that is right.
JAMES: So, the top solution we liked from banner embedded analytics was a solution that we had not heard of before, our research project, Sisense, S. I. S. E. N. S. E. What made Sisense stand out as the top solution?
GRAHAM: It’s a tool that worked well with the 2 folks who are going to be using your BI software, your developers who are the IT professionals who are going to be setting this up, connecting all the data, and your business users who will be running the queries and building reports and trying to make better decisions based on the data. What Sisense is, they position themselves as a competitor to Tableau’s the name you will recognize in this space right, they have been around for a long time, but, Sisense came in and tried to do what Tableau was doing, but with a better user experience for both your developers and your business users and, some really good performance.
JAMES: Yeah and by performance, you mean query time? Like it was fast to import. It was fast to query. It was fast to build.
GRAHAM: Yeah, exactly. Cause these are not running a couple of hundred thousand rows through them. You want to be testing this out on, on millions of actual records to try to process a lot of data and see if you can answer a question quickly when you are using real amounts of data.
JAMES: Yeah. Now in a fairly close second place for us was Tableau. Tableau was good, more expensive, and, they just got acquired by a little company called Salesforce for a small amount of $15.7 billion. There have been a few licenses of Tableau salt, huh?
GRAHAM: And just a couple of those.
JAMES: Yeah, so, Salesforce is defending their ground.
GRAHAM: I think it is all about that vendor locking.
JAMES: Yeah, so they have acquired Tableau and who else?
GRAHAM: And MuleSoft as well.
JAMES: Yeah, among a bunch of other companies, but the big ones for insurance companies, because almost every insurance company we work with is on Salesforce and there is a lot of Salesforce dev, and a lot of them use MuleSoft and they use Tableau, well now it is all under one roof. The whole BI stack extract transform load, the CRM system, the force.com platform, and then your BI and, and platform with Tableau. So now it is all under one umbrella. Tableau was not our number 1. It was not number 2, but it was a close number 2. We liked using it. It was good to use, a little more expensive. We did not, there were some things that we did not like about it, but still a great solution, widely adopted, and now it is going to be super wired in, to the MuleSoft and Salesforce ecosystem, right?
GRAHAM: Absolutely. It is a safe bet. They have been around for a long time. They know what they are doing. They have a little bit of technical debt being an older solution, but it is the most mature solution for BI.
JAMES: Yeah. And then Google recently acquired, and I will be honest, this acquisition surprised me. Only because I did not feel like they were that far along down the road, but they got acquired for $2.6 billion by Google Looker. It is likely to fall into the Google cloud platform. And Looker was super interesting because it was like BI as code?
GRAHAM: It is different. Different is the best way to describe it. What I can tell you is our developers enjoyed working with it. And I think it is a developer focus BI tool. They let you, what you typically deal with point and click and drag and drop, which is much more accessible for business users. You can do easily with code and Looker, but that does restrict it to a smaller group of users. But for your IT folks, for your developers, this was a really refreshing, refreshing tool to use.
JAMES: It is a geeky tool.
GRAHAM: Yeah, yeah.
JAMES: I mean, you are deploying it with scripts and code, and so it is different. It is not going to be your every man’s solution.
GRAHAM: For the organization that can support that, it is a cakewalk, but that is a unique organization.
JAMES: Yeah. Now, you used an interesting word for power BI that I appreciate in your report. And, in our prep for this meeting. It was fish what?
GRAHAM: Whole Fisher pricey!
JAMES: What do you mean by that, with power BI?
GRAHAM: Power BI is easy to get the hang of. It is relatively easy for users to get on board with, but it is a little bit Fisher pricey, you picture those multicolored playhouses that you have got in your backyard for your kids. Probably I can feel a little bit like that sometimes. We often recommended it as an internal tool. It is, it is great if you are not embedding it. And the UX is pretty good for business folks, but it is a little challenging for developers. It is on the IT side. Some of the code you write to integrate it like working with excel. You and I both understand how much of a pain that is. That is helpful for some folks and it is painful for others.
JAMES: Is power BI built on the Excel engine?
GRAHAM: I think it borrows some aspects of it, yes. And that is what makes it so familiar for a business user, but it does cause some of those limitations as well.
JAMES: Yeah, they share the same Dax language. That is data analysis expressions. And so, they tried to bring a lot of the familiar queries and language from Excel into Power BI. We found some scaling problems with power BI, didn’t we?
GRAHAM: Yeah, and I will not name them, but a rep from a Microsoft partner that was helping us test this out told us, it just does not scale up to these numbers. When we were loading in our mock claim data, it was a pretty large data set right, and he said, what it comes down to is, is power BI is packaged with some of those different aspects of the BI process we talked about. The ETL engine. What it comes with is great for smaller data sets, but once you load a lot into there, it bogs down and you end up having to ditch a lot of what comes packaged with power BI, replace it with your infrastructure, and at the end of the day, you are using power BI and another tool to manage all that data versus a competitor that might have all that rolled into one all in one tool.
JAMES: Yeah all in one. And one of the last solutions that that begs mentioning for insurance companies is Meta-base because this is a free open–source solution for BI that is fairly easy to plug into your in-house app and start using pretty immediately.
GRAHAM: It is a great tool. It is managed by an open–source community and you can get with open–source software. There are some tradeoffs there. It is quite Fisher pricey. Even more so than power BI. It is a simple tool, but it is a tool that you can ask your IT department to spin up. They will have it run, and by the end of the day, and you can connect it to your data sources and-and start generating reports. I mean, talk about how easy that is to start, kicking the tires with BI.
JAMES: Yeah. And that is something that really enjoyed using it for is when you have an insurance organization or a risk organization that has not leveraged BI, they do not have a huge budget, it is the middle of the year, they cannot get out. There is no occasion where like, hey, let us just snap in Meta-base and let us take a crack at the database and see what you like. And what I like about it is that you can permission the data set so it pulls your tables in, you can permission the data set and you’ll say, hey, my claims adjusters get read–only access to these tables with just this type of data and then they can go to town on building their dashboards easily. And I have seen people who are noobs, I mean complete newbies at getting some of these things done, being able to build a dashboard and make better decisions with Meta-base straight out of the gate without busting their budget.
GRAHAM: That is got to be the most important part. Is how easy is it for you to bring somebody in who’s nontechnical, give them their data and say, now you can go ahead and write your reports, answer the questions you want to ask without needing to involve or delegate this to someone else, and to do it quickly and easily.
JAMES: It could be a phenomenal starter BI system for an insurance organization.
GRAHAM: Yeah. It is like you said, we did not armchair test these out. We use all these tools and Meta-base is still when I come back to you all the time, we cause it is so easy to bring some data in and test it out without having to pay for anything.
JAMES: Yeah. And we have ended up snapping it into some other solutions of ours, which is great. We have used it for internal data analysis, we use power BI, we use that, we use a bunch of these tools. Not every tool is the right tool for the same different job, right? I mean, so sometimes you have to pick a different tool. Check out Meta-base. Now, a buzz phrase that we have talked about in a previous episode of the InsureTech Geek Podcast is predictive analytics, and it is an evolution of a good ETL process. You extract, transform, and load all of your data. You get it organized, you put it into a BI system, and then you can start trying to predict probabilities of future outcomes. How does that express itself?
GRAHAM: Predictive analytics, what it is at the end of the day is a statistical modeling problem. This is a math problem and the simplest example of that everyone comes to first is linear regression, which is just fitting a best fit line to, some properties of your data. So, determining, hey, does this set of predictor variables that do a good job of predicting some outcome. It is nothing fancy. It is a math problem.
JAMES: Yeah. And that is why we often hire mathematics graduates in computer science departments and companies.
JAMES: Because so much of this is a complex math problem.
GRAHAM: Yeah. There is a programming language called R, and that is what is typically used to snap into these BI solutions for your statistical model and work. Just R
JAMES: Just R? Just the letter R. Yeah, I like that. Kind of like C, but different. It is R.
GRAHAM: You got a lot of computer scientists and their naming.
JAMES: Exactly yeah. At least it is not named after a Greek god, right? That is another common one that a computer science guys like making. So, when you can leverage these statistical models, you can see trends, outliers, hopefully, causality, and correlation. Things that show you things you might not have, or probably did not come to on you own with your quote–unquote gut feel, right?
GRAHAM: Exactly. And it a manual process that requires some knowledge in the statistical modeling field to be able to do this.
JAMES: Another byproduct of more and more people adopting BI, is that early adopters machine learning, they are leveraging machine learning, which is when you teach a machine how to learn by brute force repetition, right? It, you teach it what a good and bad outcome is, and then it goes through brute force repetition until it finds the causality for that outcome?
GRAHAM: Yeah. You got it.
JAMES: So basic definition of machine learning. We are starting to see BI solutions; we have seen machine learning. Now we are seeing all of them use the two letters that everyone is overusing right now, AI in their marketing, but there is a whole bunch of fake AI out there right, that is just not AI.
GRAHAM: AI is a marketing term. Machine learning is a computer science term.
JAMES: Yeah, exactly. So, they are using it as a marketing term, but it is just a giant, if–then statement behind the covers.
GRAHAM: You got it.
JAMES: They’ve functionally built an expert system using conditional statements and that’s not AI. That is not even a subset of AI. That is just conditions. It is just a condition statement. So, there is some actual use of machine learning in BI where they are analyzing outcomes, they require you, the human, to tell them if it is a good or bad outcome. And then they are using machine learning. And some of the solutions that we did would automatically recommend, widgets for your dashboard saying, hey, you have got some problems in an area you did not ask about. Create this widget, you have got some outliers there that you got to check out. That was like my favorite thing and the first one I got to use in that area was power BI as well, I forgot what they call it, but it was like a suggestion tool. It allows you to quickly identify areas that you should have a widget for but do not, which, takes it way beyond that old reporting tool. I mean, typically we say, and as computer geeks, we say, computers are dumb, right? They, they do exactly what you tell them to do, they do not learn. But this is taking it to a different level. These are software programs that are recommending criteria in widgets that they were not programmed to add to your dashboard, right?
GRAHAM: Yeah. And what you are describing is a subset of machine learning called anomaly detection. And that is this machine learning algorithm can look at a set of data, determine here’s where some reasonable boundaries are on the values for this data, and that these particular values fall far enough outside of those boundaries to be an anomaly, and that is all without a statistician analyzing your data and doing those boundaries by hand. That is a computer recognizing that just by reading tons of data and beginning to understand.
JAMES: Yeah. And then at the end of the day, it expresses itself in, hey, here are some widgets you need to look at cause you have outliers or here a new report you should add to your dashboard, and you are like, oh man, I should do that. I have enjoyed getting to use BI tools with a lot of my software vendors too because what we’re seeing with common BI systems is like pre-baked integrations, so you do not have to build the integration and then import the data and then do all the ETL work, it is kind of done for you. And so, I enjoyed that aspect of all of this. And then on the data side, that has also been super exciting to get to peel into my datasets, and then to get to go into a suggestion tool and be like, hey, what should I add? What should I add to this? And that is that that is also to me where it ends up doing it, and in power BI, they call it quick insights, where it adds significantly more value than just a reporting tool. So BI is a powerful thing. If you are not using it extensively, you need to get there.
To me, like Microsoft Excel was the business tool of the ’90s, that if you are going to go into business in the ’90s, early 2000’s, you needed to know Excel. Then it was kind of the Access database, became another critical one. But to me, it is now power BI. If you want to be a power business user, go learn power BI cause that will probably come with your office 365 subscription and it is a good introduction. That, or Meta-base, deep dive into Meta-base as it is a free solution you can use. And then if you can get some licensing money from your IT department, you can jump up into “Sisense” and Tableau and Looker and all these tools that allow you to build even more powerful data sets and more powerful reports. These are not cheap solutions though, Graham. We looked at pricing and this is not on the low end on the paid side.
GRAHAM: Yeah. That is, that is what it comes down to. The total cost of ownership is not cheap for these, but consider the alternative, which is having your IT department build the same sort of thing, I mean, you can ask an IT guy, hey, can you build this, and they will say, yeah, let me spin up a server and write some scripts, and what you end up with is some tool that will give you those exact reports you asked for 6 months ago when you first requested it. When you ask for one more report, it will take them another 3 months, but when you invest upfront to build a proper BI ecosystem and you have those tools in place, it is as flexible as your business needs are at that point moving forward.
JAMES: Yeah. And result in insurance, claims dashboards, policy dashboards, real–time loss triangles. There are all kinds of beautiful things you can build that will help you make better decisions about clients and what we are seeing is the rise of the modern risk manager that is demanding this type of real–time data from their insurance providers, whether it’s carriers or TPS or brokers. And so, you are seeing the real tech–forward carriers TPA is brokers get on the ball with this and provide BI solutions to their clients because their clients are starting to ask for this. I have not had a risk manager that had SQL server enterprise on his desktop, and he was cranking data and building his own stuff cause he just was not getting it fast enough from his vendors. And you certainly do not want to be behind your client on technology and data reporting, right?
GRAHAM: Yeah, absolutely not and it is critical to be able to be ahead of that and have the data you need to make decisions at the moment.
JAMES: Yeah. So that is about it. And that is that. That is BI in a nutshell, as we like to say, we are Austin Powers. This is me in a nutshell. That is beyond a nutshell. There is a lot more to it. We could probably talk for three hours on this, but if you want more information, just go, and look at some of these solutions. We are talking about Sisense, our number one solution, we talked about power BI, Tableau, Looker, go and check these out. And you have to get hands–on and you have to start working. Pump data and start building reports, start building dashboards, and start looking at quick insights and see about natural language processing and what machine learning you can apply to it. There is no bottom to this well. You can just keep going down in the area of BI. We think there is a pretty bright future for BI coming, and in particular, just in an industry that is this data–intensive, like insurance, an incredible amount of value can be added by being a BI pro. So, Graham, thanks for being on today. Thanks for explaining the basics, before I let you go, there is a future to this that looks a little bit different and it involves something like the brain.
GRAHAM: You got it. That is artificial neural networks. AN’s. The future of this is what computer scientists have been working on. That is true machine learning is neural networks, which work a lot like the brain. The brain has an input and it has many layers and synapses on the inside and somehow put the comes out, whether it is some type of stimulus that triggers you to jump if you are scared by something. These same neural networks can be applied to, business intelligence. So, taking some kind of data in, passing it through these things that are like synapses but simulated with technology and then create some kind of output. So, you look at those same situations we talked about before.
And a good example is photos of the damage. Think about an auto policy. You get in an accident and you need to document the actual damage to your vehicle. Imagine the neural network that has been trained, like you said, with that brute force reinforcement training of showing it. Millions of photos of accidents and saying every time this is a picture of a damaged door, this is a damaged quarter panel, or whatever it is, and that neural network can take in any photo provided to an automatically generate a list. You have got minor damage to the bumper; you have got major damage to the side. Imagine being able to shortcut that whole process of the claim to be able to generate that automatically. That is the machine learning that is being worked on right now that will drive beyond the future.
JAMES: That is just next level stuff. I mean, this is creating artificial brains functionally, right?
GRAHAM: Yeah. It is very, very narrow artificial brains. They do one thing, but they can do it in an incredible way that we just have not been able to do in human history before.
JAMES: Yeah. That we just do not have the chemical processing power to get done ourselves.
GRAHAM: Yeah. No, not doing those and a millisecond being able to figure all that out.
JAMES: Wow. Great. Well, great discussion, Graham. Thanks again. I appreciate all the work you do and thanks for being on the show with us today.
GRAHAM: Thanks for having me. It was a pleasure.
JAMES: Awesome, and thanks everybody for listening. Again, we love talking about insurance. We love talking about tech. it is always exciting to cover complex subjects like this for you.
The InsureTech Geek Podcast powered by JBKnowledge. That is jbknowledge.com. It is all about technology that is transforming and disrupting the insurance world. I have been your host, James Benham. Thanks for joining us today. Big thanks to Graham Lesley for joining us as well. And I look forward to talking to you soon.
We are taking you on a journey through insurance tech. So, enjoy the ride and geek out. Talk to you next time.