AI Your Way: Smarter Data, Smarter Workflows, Smarter Decisions

Hello, everyone. Welcome to our webinar on AI your way, smarter data, smarter workflows, smarter decisions. My name is Aubrey Eyer, and I’m thrilled to be your host today. We have an exciting session lined up, and we’ll have time for your questions at the end of the presentation. If you’d like to ask our speaker anything, please submit those questions through the Q&A function on the Zoom toolbar, and we’ll address them at the end. Before we dive in, I’d like to introduce our esteemed speaker, Chris Bennett. Chris is a senior leader in risk and insurance technology with nearly three decades of experience in client service, sales, and product management. As chief strategy officer, Core Solutions for Origami Risk, Chris is responsible for guiding the division’s growth and development. This includes forging industry alliances and building new capabilities and solution sets to address the dynamic and evolving needs of a diverse clientele. I’ll turn the presentation over to Chris now. Please take it away. Thank you, Aubrey, and thank you everybody for joining us this afternoon. We’re always grateful, to meet our clients and prospects, and have a chance to share kind of great things that we’re doing with you. So excited to present. As we get started, what we’re gonna walk through today is kind of, you know, in response to a lot of questions and inquiries from clients and prospects. We all know that AI continues to grow in usage across our industry and really every industry. And and in insurance in particular, you know, folks are starting to to look at how can AI help me day to day in my job? How can it help me with the tasks that I have to accomplish, make them easier, make them faster, make me more efficient? What we’re gonna talk about today is kind of our vision and where we are heading with AI, what we’ve built so far, what we continue to build, and what we’re really excited, as we roll out the over the next couple of releases to our customers, all the exciting things our clients are going to be able to do with AI. So as we get started, we really built and thought through AI. Obviously, when GenAI first started becoming, you know, kind of front of mind for all of our customers and and all of the industry, really, we started looking at what are use cases we could work with our clients that would help them in their day to day job. And we we worked with some of you. You may have been involved in kind of the early betas. We put together some some very practical use cases on ways that AI could be used, kind of initially. What we learned out of those pilots is while it’s great that we could take a particular use case and build it for our client and the client would be happy, the reality was more and more clients were asking for more and more things because everybody’s job is slightly different and there is no true one size fits all in our industry. And and and we have fifteen years of of experience at origami that tell us each one of our clients figures out the right way for their unique business to make their business function. And so we had to think about a new way of approaching AI. And there were a few core things that we kept in mind as we started. One one, and this was feedback directly from our our customers as we began the the POC process, was security is paramount. People people are nervous about the idea of sharing data with public LLMs and having data that gets out there in the wild, particularly when our customers are capturing sensitive data on their customers, on their employees, on their, you know, other folks that they work with. We didn’t wanna be in a situation. Our customers don’t wanna be in a situation where that’s potentially put at risk. So we knew security was paramount. We also knew that what matters to our customers is actually solving their real world problems. They they want to be able to use AI to make them more successful. And and we couldn’t dictate what use cases were gonna be best for all of our customers. And and finally, people want flexibility. And this goes hand in hand with ensuring success. People wanna be able to use AI to solve their unique business challenges, not to have specifically canned use cases that they could apply. So those were sort of three things that we kept in mind as we were designing our AI solution and our go forward path with AI. And, really, where we landed and kind of what we’re gonna talk about today is really a three tiered approach to AI that origami is taking. And it starts at kind of that foundational level with the idea of smarter data. Can we use AI to better help us ensure that we have data in the system? Can we get data in in an easier fashion? Can we ensure that that data is accurate? Can we leverage that data downstream within the system? The the next kind of layer on top of that for us is the idea of smarter workflow. Now that we’ve gotten the data into the system, what can we do with that data? What actions can we drive? In particular, what automations can we start to deliver based on that data that we’ve pulled into the systems? And then finally, for us, the the last sort of, layer of it for us is what we call smarter insights. And that is now that we’ve gotten the data in in a better way, we’ve gotten the workflows that we need to drive our day to day action. What insights can we start to surface from our data, from our system in an easier fashion that can actually help us drive our own efficiency, drive our own processes. So how do we how do we drive better action using those insights? And what we’re gonna look at today kind of in those three categories and those three themes, you know, we’re gonna take a look at smarter data at our new AI data center. So what we’re working on for transformation, validation, manipulation of data within the origami system. So how do I get origami? How do I get data into my origami system in a better, smarter way? We’re gonna take a look at a few tools that align with our smarter workflow. We’re gonna look at our origami workflow engine with AI, so our AI tile that’s being built into our new workflow flow engine. We’re gonna look at our origami concierge, which really drives us. And what we’re gonna be focusing on there is ultimately a drive towards no code where an end user could build a workflow through simple English language processing. And then we’re gonna take a look at kind of a unique case, our AI risk and control explorer. So the idea that we can start to leverage AI to look outside of the known things in our system and to say, help us identify things we should be thinking about related to risk. And then finally, for smarter insights, we’re gonna take a look at our new origami AI analytics, which is really exciting. It’s a new way of driving, pulling that data, visualizing data in a very interactive natural language process, eliminating a lot of the requirements or the necessity to understand the underlying data structure, how to use the system to create, for instance, a new report or a new dashboard, graphic widget. AI analytics helps to to make that even easier to get data out of the system. So we’ll start, and we’re gonna take a look at our new AI data center. So these this is something we’re currently working on in AI. So we’re gonna take a look at kind of a very early release. This is stuff we are testing internally right now. We’ll be looking with most of what we’re looking at today, some of it is in release. Most of it is in early testing internally. We’ll be going through beta testing as kind of a next pass and then ultimately into the release. I know availability will come up at the end. So with the data center, a couple of challenges we’re trying to help our clients solve. Certainly, the manual mapping and and manipulation of data when you have to take a data source and get it into the system. Right now, we’ve got great tools to help transform data for our customers, but it does require somebody to look at it and say, okay. Here are all of my input fields. Here’s how I’m gonna map it. Here’s ultimately what I want to come out. We certainly wanna look at performance bottlenecks. Anywhere that slows down your your operation when you’re importing data, we want this to be super fast, super efficient. And we also wanna make sure that we are thinking through how do I make sure my data is accurate. If you don’t trust the data that goes into the system, you’re not gonna trust the analytics and the data that that drives your decision making at the end of the process. So those are all sort of areas that we’re focusing on as we build our new AI data center. And a couple of the themes and and sort of functionalities we’ll look at, one is AI powered data transformation. We’ll look at natural language automation. We’ll look at some cross data validation and then, ultimately, the ability to take those three things, so using natural language and AI, building a data import, and then ultimately being able to return it turn it into a reusable workflow. So as we come into the system and I’m gonna start first by thinking about the data model. So we need to be able to add a new employee table, for instance, to our data because we wanna be able to capture some data or import some data into our system. Now using natural English language, I can come in to Origami’s data center, ask it to create the table. I need a new employee table. It is going to build that table based on the most common elements that it would see or expect in an employee table. And then it gives me the option as the administrator to come in and modify as I need to add additional fields if we need to make changes, but I don’t have to start from scratch. Right? The system is doing that interpretation and saying, I’m gonna go ahead and create this table. Here’s my pass at it. Now you validate. And it’s the idea of the human in the loop. Right? So we’re not trusting AI to get things perfect. Of course, it will get better and better as it evolves. We know that. We we wanna make sure that that it is an assistant, that it’s helping us get this started, and that we can finish the process. So once we’ve created data and we’ve got data in the system and now I’m gonna take a look at what it will look like as we import new data into the system. We expect as we go through that for our clients, it it could eliminate up to eighty percent of the time spent on that initial ingestion, cleaning, validating, and transforming data. All of that mapping, a lot of that validating, a lot of the data cleaning, writing the scripts that has to be done, AI can handle many of those tasks, eliminating them from the process and really speeding up the the transformation of data that we need to import in the system. So we’re gonna take and we’re gonna grab a file, and that file contains some basic claim information. So maybe this was a lost run that we we received, and we wanna be able to import these claims into the system. So Origami AI is gonna take a pass at mapping out all of that data, figuring out where it needs to go, what data it’s looking at, and, ultimately, where that data is gonna go in the system. It also allows us to start to do transformations. So in this case, what the system just did is it looked and said, you know what? We only have a paid we have several paid fields, but we lack a total paid field in the input file. So we wanna go ahead and sum all of the paid fields into a total paid field. And AI did that for us, created that new created the formula, built that new total paid field for us without us having to come in and do that transformation. So all of that initial mapping and that transformation happened using the AI. And now, again, I, as the the administrator, is the person who’s doing this import, have the ability to come in, manipulate, make changes if I need to, confirm and validate that what what AI saw and did matches with what we expected, and then we can proceed. Now we could also look at it, and we could look at that file and say, you know what? It came in with a first and last name field in the same field. I wanna go ahead and separate those out. Today’s world, I would have to write a routine that would go through each of those data elements, figure out how to parse it, extract the first and the last name, and to put them into separate fields. In this case, I was able to take and use AI to actually process that. Similarly, I was able to use AI to do things like calculate average weekly wage based on the wage that come in or to to look at birth date. So we can start to manipulate data very easily, and it eliminates the need to write the routines that historically we would have had to have written to do that data manipulation. And finally, data validation. So I want the system to look. And and if the age field does not line up with the date of birth, birth so today’s date, then let’s go ahead and fix that and correct that data error that we see in our system. And so, again, rather than me having to to write a routine to do this, I can use simple English language instructions and allow the system to make that data manipulation change for me, to go ahead and and transform, confirm, validate that that data is correct. And and I’ve saved a huge number of steps in my process. And then when I’m ready, I can say, you know what? I wanna go ahead and run this job. Things look great. I wanna run it. And then, ultimately, I wanna set this job up as a recurring import. We’ll feed it in, and that file will get processed. So really, really saving time in the manipulation of data for our customers who ingest a lot of data into origami across our markets. So the next thing we’re gonna look at and talk about is the idea of smarter workflow. So now that we’ve we’ve gotten data into the system, we’ve manipulated data, how do we actually start to work with that data within the system? And there are a lot of of, you know, ultimately, a lot of outcomes, a lot of workflows that we can drive based on that data. But what we wanted to think through was how can we leverage AI to allow our clients to automate the things that they do day in and day out or need to do day in and day out with that data once they have it in the system. And for us, we really started with the idea that that we wanted to democratize AI and the way that AI worked in the workflow in origami. Meaning, our customers can use AI in any way that they want to leveraging the new workflow tools. We certainly wanted, you know, the the configurability to remain for our customers. We know that our customers all use the system differently. We know that they wanted, you know, secure native AI integration, not not a third party tool that we’re going out to, something embedded within the system. And we know that they want to be able to drive the workflows that matter to them. So we really focused on building something that could be used by all of our clients across all of our markets to meet their particular needs leveraging AI. And I’m demonstrating here. So this is a simple claim summary. So I’m using the the AI tile in our new workflow engine to summarize a claim record and to put that on the claim screen. And I can drive that that summary. I can trigger that summary similar to the way for those of you who work with our administrative tools, who build workflows today based on any action that happens. So I could you know, anytime a claim is updated, I could have AI generate a summary and drop it on the screen, or I could have it set so that I could actually add a button and say, I wanna generate an AI summary. You’ll see the sample of that here on the screen. And I could click that button, and it would generate the summary. But the idea is it is looking across the entirety of the claim record to summarize what’s going on status and what the next steps are for this particular claim record record. For all of our customers who work with claims, it it’s a very, you know, simple, repeatable process. They all need the idea you know, coming in and looking at a summary without having to read through every single item and every single attachment that’s on a claim. So super powerful for our customers. But this is just one sample of what we can do using the power of the AI tile in our workflow. And so, really, the AI tile, what it will allow our customers to do is, you know, any of the actions that we see here. So if if I come in and I wanna build a workflow that allows me to generate a task or generate an email and and I want AI to go ahead and write that email as a part of the process, I can do that. I can use it to summarize inputs. So if I’ve got documents that are coming into the system and I want for instance, medical records and I want the AI to summarize those medical records and perhaps put that into a note, I can use AI that way. I could really summarize any record. So I could say I want a summary of a location record, or I want a summary of, you know, a policy record. And and you’ll see we broke policy out separately in claim because our AI will allow us to train models for unique cases like claims and policies where it truly is based on on the the data that our customers are using and how they wanna see those summarization. So these will be trained models for for claims for policies, for instance, as well as more generic models that will take any type of a record and write a summary of that record. Searching doc documents, generating lists. You know, some of the way that that we see our customers wanting to use this and that we’ve started to build use cases for, you know, wanting to use this and that we’ve started to build use cases for, you know, things as simple. We looked at claim summary, getting a claim, summarizing that rec record and writing out a summary so that whenever anybody opens that claim, they’ve got that record right in front of them. For some of our clients, they they may want to upload a policy document. So they get a a copy of a policy document from their broker. They want that document to actually generate policy details in their system so that they can look at the claims against those policies or look at exposures against those policies. And rather than having to take that and manually type it, they could upload that policy document, run that analysis, and ultimately have the system write out the details of that policy for them. This can work the same way for our customers who do, for instance, claim intake. If I have a first notice of loss document that I receive and I wanna create a new claim record in the system and begin the routing of that claim based on that intake. I could I could be looking at that first notice of loss, analyzing it, and ultimately generating that new claim in the system. And then, of course, you know, simple output efficiency based tools like generating emails. So whenever we have a status change, we we may wanna automatically generate an email, have it go to the adjuster to review, and then ultimately send out that email, so that they they they can keep the brokers. They can keep their their clients apprised of changes in the status of that claim. So these are just some of the examples, obviously. It it really is limitless. The idea of of embedding it within the workflow allows our customers to use it for any particular use case that they want to. It it’s really removed the limits. So we’re not saying here’s the five ways you can use AI. Instead, we’re saying here’s an AI tile that you can embed into your workflow and use for anything you wanna do to help save time. And and, really, for our customers, you know, take a second and think about what could you do with this in your day to day use of origami. Right? How would you use AI to automate your workflow? And I’m sure that there are hundreds of ideas being generated right now. It it it is it’s a it is really gonna be a game changer for our customers. Alright. So next what we’re gonna look at is something we call origami concierge. Some of you may have seen the early, the kind of the first iteration of concierge. It’s available now. And what the concierge is is really a guided walk through help that exists within origami. And it started as a way, what we call tell me, to look through all of the supporting documentation, to look through all of the information that exists on a particular question and provide answers to our users who can type a simple question. And we’ll take a look at that. The next iteration of it will be what we call show me. And show me is is similar in concept, except you can ask the system to show you how to do something. And in essence, you have a a teaching assistant that exists in the origami application that will navigate to the place where you need to click the button to begin that process and walk you through the process. And then the end stage where we’re really heading with this is what we call do it for me. And do it for me is the idea that instead of me having to go into the administrative model module to create a workflow, instead I could just interact with the concierge and say, I wanna be notified anytime a claim looks like x. You know, anytime we have a high dollar claim or a claim over x, anytime we have a certain type of loss, any type we have a certain type of risk or a finding that that is found, those can trigger those actions and you can build workflows automatically. So we’ll start by taking a look at tell me. And the idea with tell me is I can just start to ask simple questions of the system. So, you know, how does CMS reporting work? And Origami AI in English language is going to give me information about CMS reporting and then ultimately provide any links for follow-up data that I need. And I could ask another question. What can you tell me about endorsements? Similarly, it’s gonna go through and tell you everything it knows about endorsements in the system and provide those links. So the idea is is interact in English language. It’s gonna drive you to a simple answer and then provide links to the underlying documentation that you can click, open, view, and dig deeper into. So think of it kind of as a an embedded helper, much better help and and navigation of of help than historically has existed in in system. So day to day productivity tool. The next step is something we call show me. And the idea with show me is I want you to show me where to go to do this thing because I just don’t remember. So how do I, for instance, create a new t core model? And Origami will actually open up the system. It’ll tell me what I need to do and then walk through and let me say, create it, and it will actually take me to the place where I need to go, tell me to click the button, and what I need to do to drive that action. So it guides me, lets me click it, and I can see and start to enter my new t core model and enter the details and and do what I need to do there. And then, ultimately, where we’re looking to get to is what we call do it for me. And the idea of do it for me, like I said, you know, I want to build a workflow. So I asked the system, let’s create a workflow. And we’re gonna go ahead and start the process. I want a workflow that looks at claims. And if it’s a high priority, I wanna automatically email the account owner. So instead of me having to build the workflow myself, AI is actually interpreting the question or the request, and it is going out to the workflow to tool and building that workflow and the steps that will exist in that workflow. And then it will actually open that workflow up. Let me look at it. Let me revise it if I need to so I can see it. I can see the workflow built. If I wanna make changes, I can. And, again, this goes back to the idea of always we wanna use AI as a helper, but we always wanna make sure that we can look at it. We can validate that, ultimately, it’s driving the action that we want to, and then we can put that in place. So a real time saver to let us create those workflows automatically. And finally, in the in the workflow category, we’re gonna look at is a new tool called the AI powered risk and control explorer. And, again, this is a a simple example, but the idea is tell us a little bit about your business, and we are gonna tell you the things that you should be thinking about as risks you need to mitigate for your type of business and then help you create the within the system, the ability to monitor and manage those risks so you’re not having to create them from scratch. So in this case, it’s a taxidermy company, ventured some details about the company. We copy that in, and we say, hey. What should we be thinking about? What recommendations do you have? What risks need to be on our radar? What mitigation steps should we think about for those particular risks? An AI is is looking at the types of business, looking at the types of risks that potentially exist, and then providing us a list of all of those risks that need to be on our radar and allowing us ultimately to create and then to create plans to help mitigate those risks. So a very, very, you know, unique, really cool application of AI, to to help us identify. And this is where if you didn’t know that you had that risk, you might not think to mitigate it. And if you you don’t have somebody who’s coming out and looking at it for you, you might not think to mitigate it. So really a proactive risk mitigation and identification tool. And so now we’ve we’ve we’ve talked about data. We’ve talked about some of the workflow tools we’re building, which are certainly important. That’s gonna drive a lot of efficiency. But let’s talk a little bit about insights. You know, and one of the things we’re working on at Origami and that we’re really excited about is making the the ability to get data out of the system easier for any end user. Right? So I don’t wanna have to ask somebody in my administrative group to help me create a new report. I just wanna be able to ask a question of the system. Because, ultimately, if you think about the way that analytics are used today and Origami has incredibly powerful analytics and dashboarding tools that our customers use all the time, and they build great dashboards that do a lot of things for them. But a lot of times, what happens with with a dashboard is you get a piece of information, and you say, that’s great. What does it mean? Why is that happening? And you need to drill down more, and you may not have those reports set up, and then you have to ask somebody or you have to know how to set those reports up or set that widget up to sort of dig one level deeper to ask follow-up questions. We wanted to really change the way that that customers interacted with and pulled analytics from Origami. And, you know, as we said, dashboards are great, but they can generate more questions than they actually answer. Why? Why is this happening? Tell me the details. Where do I dig in? What do I need to look at? What should I be paying attention to? So as we thought about origami analytics, we know that our clients want to be able to interact with their data, ask questions of their data, get to meaningful insights from that interaction, really see what matters to them. Ultimately, you know, much, much simpler for customers to interact and pull data. We know that if you are not technically proficient, sometimes it can be harder to explore data. If you don’t know the data model well, you’re not really sure how to create a new widget. It’s not something you’re trained on or you do frequently, then you’re having to go to go to somebody for support, either internally, your own administrator, or you’re coming to the origami service staff. And we’re always happy to help our clients create those reports. We just wanna make it easier for our customers to self-service. You know, we know that reporting and analytics required a great deal of training, and we know that the, you know, sitting and adjusting those visualizations and getting it just right takes time. These are all things we wanted to help fix with the new analytics tool. We wanna make it simple to interact. We wanna make it very smart so that it can it it knows based on simple interaction the type of data it should be looking at, and it takes a pass at pulling that data. We ultimately wanna drive smarter decision making from our data. It’s what it’s really about is enabling our customers to use the analytics to drive better decision making. And so we take a look at at the new origami analytics tool. Instead of having to create a custom widget, I can actually come in and in simple English language, start to interact with and ask questions. And the origami analytics tool is actually taking a pass at interpreting what I’m asking, going out and setting the dataset, and then saying, you know what? What is the best format to display this type of data in? I’m gonna go ahead and display it. Now I, as the user, again, keeping the human in the loop in mind, can come in and manipulate as I need to. I can ask further questions. I can look at data differently. I can slice it, dice it, look at just a subset, maybe change the way that I’m outputting the data. So I still have that full control over the ability to edit both what’s being seen and how it’s being seen. But I’m not having to think through and create these from scratch. Right? AI is creating that query for me. I can see the query it created. I can see the underlying data. I can confirm it, but I’m not having to think through writing the query or writing that that block of of criteria myself to pull the right data to take a look at. And then I can just continue to ask more questions of the data. I can look at that data in different ways from a visualization perspective. I can drive to the underlying, you know, query, the underlying data. I can look at it in grids. And then, ultimately, as I as I continue to manipulate and I arrive on something, you know what? This for me is really, really useful. This is something that I could take, and I actually want to put this on my dashboard using the analytics tool. Once I’ve gotten that data to a state that I want and I’m like, yep. This is how I wanna see it. This is perfect. I can actually then take and drop that that new widget that ultimately AI created. I didn’t have to think through the the process. I didn’t have to do any building myself, and I can drop that right onto my dashboard. So so this is really going to unlock a lot of of power in analytics for customers who otherwise may be limited to what they’re provided by the administrator who set them up. Right? So they’re looking at their dashboards day in, day out. They’ve got certain KPIs, certain analytics, but but that’s what they have. And it’s hard for them, and they don’t know how to create new ones. This is really gonna open up that that idea that any user can come in, interact with the system, and drive to data that is meaningful and that matters to to them as a user. And with that, I know we’ve got fifteen minutes left. I wanted to open it up for questions for you guys. So Okay. Yeah. Thanks, Chris. We are opening up to questions. So it looks like we have quite a few in the queue already for us to get started with. But if you’ve got them, you could type them into the q and a section. So, Chris, I wanna start with our first question, which comes from Michael Stacom. With respect to data import export, can you speak to managing the risks related to AI hallucinations on the reliability of the import and the export? You know, it it’s a it’s a great question. So one of the things this is something we have to be cognizant of and think about in in any use case with AI. And there are different techniques and tools. I know we’ve experimented with the idea of QA ing AI with other AI so it could actually look for hallucinations. But at the end of the day, I think the key for us and where we landed on this is is this is a tool that’s meant to help, so Eve can drive the beginning of the process. But, ultimately, we wanted to keep this as a human in the loop. In In other words, we don’t wanna rely entirely on a AI and say, hey. Convert this data and load it into the system. We wanna rely on AI to say, hey. Take a pass at mapping this data and getting it ready to load into the system, and then I’m gonna look at it and validate, and then we’re gonna go ahead and load that data. Okay. Thank you. The next question we have is, how are you calculating the AI return on investment of eliminating up to eighty percent of the time spent ingesting, cleaning, validating, and transforming data? You know, great question. I don’t know that we’ve really thought through the the ROI calculation in terms of the dollars. What what we’re really looking at is purely time. So we look at how long does it take us when we get a file, for instance, to go through the process of field mapping that file, of taking every single one of those fields and figuring out where does this go. Right? What’s the table it goes in? What’s the data element this fits into? Is it the right format? Do I have to do some sort of a transformation on this data so that it will feed into the field I want it to feed into? And then sort of next level down, if you think through the transformation aspect. Right? So where you have fields that need to be manipulated before you put them in, you’ve gotta write those routines. You’ve gotta build that logic that says, I wanna create a new total field, and I wanna add these fields up and subtract these fields, and this is my new total field. Or I wanna separate name fields, or I wanna automatically, based on dates, populate ages or do other things in the system. All of that can be done for us using AI at the front end of the process, and all we’re doing again is validating. Right? So that’s what we’re really looking at is how long does it take us to do this if we are doing it ourselves, end to end using, you know, just our own processing power versus if I use AI to speed up that process, what what does that save us? Okay. Thanks. Our next question is, will Origami AI be able to take a user’s natural language to find data fields within the database and create customized reports? So for example, could you say something like, please create a transaction detail report that includes all claims for a given company and group all transactions by the claim number and claimant number? It it’s a great question. I I would say, maybe down the road, what we’re focusing on at first is the idea of visualization. Right? So it’s the idea of creating widgets and graphs interactable, less about the structured report with the groupings and the various things. I can see the underlying data from the visualizations, but that’s certainly a great suggestion as we think about next step for that is, can we create what would what would traditionally be a printed out report? And just say, here’s how I wanna group. Here’s how I wanna sort. Not something we’ve tackled in the first iteration. It’s really more focused on data visualization. Okay. Thanks. We had a couple people ask about this. So when you say that these are trained models who learn from each client, do you mean the system is learning from the client’s data? And are there walls in place to make sure the program is not learning one client’s data and somehow using it in another client’s program? It it’s within the client’s data. Okay. Thank you. Our next question is, for summarization of our record, does it apply only to core domains or also CDEs? It can be any domain. K. And then they had a follow-up. What fields are considered in the summary configurable? What fields? Not sure I understand the question. Meaning meaning the output. Like, what’s the summary gonna do? That great question. I don’t know that we’ve set the ability to configure what gets what the output of the summarization looks like. AI is taking its best guess at at summarizing that output. Now that’s different obviously for claims and it’s different for policy. Because in those areas, we’ve sort of said, hey. Here are things that we wanna be able to summarize as a part of a claim record or a policy record because those are the most common area, and that’s where we’re starting with trained models. Okay. Thanks. We had another question. Some people asked, can AI look at numerous files to determine trends, or can they use related record fields for a summary? So it will look through the entirety of the record. So it can look through all of the attachments. It can look through all of the transactions in order to summarize. And kind of along those lines. So it’ll be able to summarize five hundred to a thousand page medical documents, which have a lot of redundant data. Why would you wanna do that? Those are fun to read. Yeah. The idea is that that it can be used to summarize anything. Now medical documents obviously are a bit unique, and that may be one where down the road, we think about more of a trained model for medical records. But, certainly, you could feed medical records too and say, summarize, you know, summarize the what’s included in these documents, and it’ll do its best to to make a summary. But we don’t have a trained model for medical records yet. Okay. Thanks. Next, is the AI for claims still focused on auto claims, workers’ comp, and GL claims, or does it go beyond that? It it’s it’s any claims. It’s it’s not particular lines of coverage. K. Thanks. Then we had a question about the AI created widgets. Can they be added to home or territory or location dashboard types and domain forms with the ability to tweak and filter after? Eventually. Okay. Eventually. Not in the first not in the first pass. So when we when we do first release it, it’ll it’ll just be the dashboard. It won’t be individual page dashboards, but down the road, certainly. Okay. Someone asked, where can we go under admin today to get started on what is available related to AI? So the only thing that you would have today because the only thing available for general release is the first iteration of the concierge. So that’s not in the admin tools. That’s you know, you can work with your service folks on on the enabling the concierge or making sure you have that set up. The rest of it is stuff that we are currently in our initial round of testing internally with our employees. We’ll be moving to beta testing some point in the near future here, and then starting to roll these features out. So not nothing that we looked at with the exception of the first part of the concierge is in general release yet. This is all this is all brand new stuff. Okay. Yeah. There were a couple questions about the time frame and when the first release will be. So it sounds like we’re still we’re gonna have some of our Honestly, a lot of it it’ll be over the next, you know, probably three to four quarters that we release all of this functionality. And and I would say work closely with your your customer success and service teams. You know, they’ll have a lot more information on kinda what’s becoming available when, what will be in each release as we roll it out. Great. We had a question. How does the Show Me tool work with custom data entities? I believe right now, it only works with the base entities in the system for the show me. Okay. Great. We have another question. Will the AI functionality have the ability when generating an email from the system to read and apply internal company policies and make us aware of deviations from policy or make us aware of risk concerns around the emails before they are sent out by the system out of the system by end users? So great question. I I would say, initially, I don’t know that there’s a plan to interact or allow AI to to merge sort of company policy or client specific, information into into the way that it’s working. I think it it’s gonna be based on on the model itself and on the data. I think as we progress down the road and this this is a lot of places. Say, similarly to the concierge, we’ve had the question before. Can it read our own policies, our own internal help files, and display the answers? And and initially, no. It won’t. But as it continues to evolve, it is certainly something we’re considering is how do we start to to let the client’s AI get even more specific for that client’s usage. Thanks. We did have someone else if you comment a little further on the the data security. Are we hosting this model directly? So we are using, Amazon’s Bedrock. So we’re using the the large language model that exists within Amazon. Each client’s data will be in their own, their own version, their own LLM, if you will, separated and isolated from other data. Great. Thanks. Someone asked, can we turn AI on and only allow it to access allow access to specific data according to our company policies? So they have restrictions on using it with particular type of data. I don’t know that we’ll be able to segment or say what it can or can’t work with. It certainly can be turned on or off. So if you’re if you’re concerned about turning it on, you don’t have to use it, you know, absolutely. I don’t think that that that control won’t exist at least initially. Okay. Thanks. And we are almost out of time, but we’ve got just one or two questions. So one is what is the limitation here in terms of processing documents? For example, if a claim may contain police reports, medical receipts, and they can range anywhere from one to a hundred pages, What would you say is the accuracy for longer and more complex document extraction? Great question. I I don’t know that I know the answer to that one. That’s one that you can certainly stay close with your service team on. I know from a processing perspective, you know, perspective, you know, the we’ve architected this, and this is designed to function very efficiently even across very large data sets. Right? So I’m not so worried about, processing speed per se. You know, in terms of accuracy, do I have metrics on how accurate it is on specific types of reports? You know, I I don’t have that yet. No. K. Thanks. And then we have someone who asked, how is the summary limited? Some roles might need a summary relative to select aspects. Can what specifically be used for the summary be configured, or is the only option everything? So, initially, I I don’t know that it will be as con you know, super configurable. I think the idea is we we’re telling it, here’s what you typically wanna return. It’ll be somewhat limited. I think as we go, it will continue to evolve. Okay. Alright. Thank you. Well, we’ve got just one minute left, so we’ll take one last question. And that just asks, when you think the control over that data might exist. Control over the data. And segment it out. Oh, Definitely, we see farther down the line. Or Yeah. That that’s further down the line. I don’t think it’s something we’ve necessarily even anticipated or start to think through yet. Okay. Alright. It’ll continue to evolve, but I don’t know that we have any plans. Alright. Well, with that, we are right at the end time. So, Chris, is there anything else you’d like to say? Otherwise, I’ll wrap us up. No. I I just just to express gratitude to to all of you for joining us today. We’re always excited to share kind of cool things we’re doing at Origami. We learn a lot from all of you. Your your feedback, your participation, your interaction with with your service and support teams means a lot to us. And, ultimately, that’s what drives a lot of the ideas that we end up focusing on. So so thank you to all of you. Okay. Well yeah. Well, thank you again for joining us today, and, I hope that you have a great afternoon.