In a recent webcast titled “Actionable AI that Adjusters Actually Adopt: Strategies to Realize 200% Plus ROI,” Property Casualty 360 hosted Cylus Watson, President and Head of Voice of Customer at Voltaire, and Jo Farris, President of TSI Adjusters, to discuss practical AI implementation in claims operations.
A full recording of the webcast can be accessed here: https://voltaire.claims/resources/#case-studies
Read below for a summary of the key themes from the presentation.
1. The Claims Letter Bottleneck: Manual Processes Creating Risk and Inefficiency
Complex claims letters routinely consume 45 minutes to an hour of adjuster time. The problem intensifies during catastrophe events when carriers onboard hundreds of temporary adjusters unfamiliar with specific policy language. As Watson explains during the demo, adjusters “face constant pressure to close claims quickly” and often resort to copying from previous letters—a practice that introduces serious risk.
The financial stakes are substantial. Litigated claims cost carriers 4x more than non-litigated ones. Mid-sized carriers processing 10,000 property claims annually face millions in potential losses from communication errors and claims leakage. The root cause is complexity: insurance policies layer base provisions, state-specific amendments, and endorsements that can modify definitions across dozens of pages.
Jo Farris describes the challenge from an independent adjuster’s perspective: “We’re working for numerous carriers. Desk adjusters haven’t worked for these carriers before. Everybody has a different format.” Even experienced adjusters struggle to track which version of water damage exclusion applies when page 87 modifies page 22, and an endorsement on page 96 further changes the definition.
2. AI Implementation That Works: From Pilot to Hurricane-Scale Deployment
Voltaire’s approach differs from template-based systems. The AI generates each letter from scratch after analyzing the claim file, dec sheet, and estimate. During the live demonstration, Watson showed how the system prevents errors before they occur: when an adjuster attempts to deny coverage without valid policy language, the tool returns “no relevant policy language found.”
TSI Adjusters tested Voltaire with their desk team before presenting it to carrier clients. Within months, they deployed it during Hurricanes Helene and Milton, processing thousands of letters. Farris reports the results: “We were able to improve our QA scores across the board” and became the carrier’s #1 graded adjusting firm.
The system addresses the adoption challenge that sinks most AI initiatives. Watson notes they achieve near-universal usage by focusing on end-user experience: “We built a product that works for both the decision makers and the end users.” New adjusters generate accurate letters immediately instead of spending weeks learning carrier-specific policies. One adjuster reported saving two hours daily.
3. Measurable ROI and the Competitive Imperative
Voltaire documents 200%+ first-year ROI through combined time savings and reduced litigation costs. At $15 per claim for unlimited letters, carriers processing 10,000 annual claims invest $150,000 while gaining efficiency equivalent to multiple full-time employees. The math improves further when factoring in reduced litigation expenses and claims leakage.
Farris emphasizes the strategic advantage: presenting Voltaire to clients “opened up great conversation pieces with newer clients and potential clients” by demonstrating technological leadership. Watson projects that within years, AI correspondence will shift “from innovative to expected minimum” as plaintiff attorneys increasingly use AI to identify weak claims letters.
The webcast poll revealed that freeing up adjuster time for higher-value work ranked as the top desired outcome. Watson reinforces this vision: “We’re not trying to replace your workforce with AI. We’re trying to free up their time” from the task adjusters like least, enabling them to focus on investigating claims and serving customers.
——
For a demo to see Voltaire generate an accurate claim letter in as little as 30 seconds or even on one of your carrier’s templates, schedule time with a click.
Full Transcript
Geoffery Metz
Welcome, everyone. The webcast is about to begin. Please take it away, Jeff. Thank you, Cooper, and hello, everyone. My name is Jeffrey Metz, and I’ll be your moderator for today’s webcast, titled Actionable AI that Adjusters Actually Adopt, Strategies to Realize 200% Plus ROI. This event is brought to you by Property Casualty 360 and sponsored by Voltaire. Before we begin, let’s go over some basic housekeeping items about the webcast console. This event is completely interactive and features many customizable functions. Every window you currently see, from the slide and speaker windows to the Q&A panel, can either be enlarged or collapsed. So if you’d like to change the look and feel of your console, go right ahead. If you have a question for one of our speakers, please enter in the Q&A panel on your console. We’ll ask as many questions as possible during the live Q&A after each segment, so please ask away. If we don’t get to your question, we do send these to our speakers after the event, so you may receive an e-mail response afterwards. There are also a few polls during today’s event, so when one comes up, please respond as best matches your experience. And we’ll take a look at those results in real time. And also a few resources, including white paper and one pager, as well as a link to Voltaire, have been made available in the resources tab, so you can click through and review them at your leisure. And now let’s go ahead and introduce you to today’s speakers. First, we have Cy Watson, or Silas Watson, who is president, head of voice of the customer for Voltaire. Cy spends nearly every day speaking with carrier claims leaders and translating their needs into product requirements. and his claims policy knowledge rivals most senior adjusters. Sai is a serial entrepreneur in cutting-edge technologies, had encountered both a consulting firm and two hedge funds. He is widely recognized for his expertise in yield farming and his ability to capitalize on emerging market trends. Next, we have Jo Ferris, who’s president of TSI Adjusters, Jo founded TSI nearly 20 years ago and had successfully weathered dozens of storms for carriers across the Southeast, transforming TSI into one of the leading independent adjusting firms in the region. Most recently, her firm implemented and embraced AI, leveraging the results of the early adoption to of their early adoption to impress and transform operations for carrier partners. Jo, Sai, it’s great to have you both. And did I forget anything?
Cylus Watson
No, that was great. Thanks, Jeff.
Geoffery Metz
Fantastic, fantastic. So before we get started with today’s topic, we wanted to 1st get to know a little about you, the audience. This is our first poll. So what’s your organization’s current stage with AI adoption and claims? Is it exploring researching AI solutions? Piloting AI tools in limited capacity, actively using AI in some areas. AI is integrated across multiple functions. No current AI initiatives, or not sure, not involved in these decisions. If you just take a few minutes here to respond, or a few seconds here to respond, we’ll go ahead and push those results and then get to our topic in today’s discussion. And again, what’s your organization’s current stage with AI adoption in claims? Piloting AI tools in a limited capacity. actively using AI in some areas. AI is integrated across multiple functions. No current AI initiatives. Not sure, not involved in these decisions. Feel free to select the one that best matches your current experience. Okay, I’m going to give it about another maybe 10 seconds for people to get their responses in and then we’ll push those results. Okay, so let’s go ahead and take a quick look. And it looks like it’s kind of all across the board here with the audience. The plurality is not sure, not involved in these decisions. We have some people piloting, some actively using, and of course, no current initiatives. Is this about what we thought we’d see with the audience today?
Cylus Watson
So I think a big part of this conversation we’re going to have today is about being able to use AI solutions actively within your organization. So it doesn’t surprise me that a lot of the participants are researching or piloting or have no current AI initiatives, but they’re on this call. So obviously it’s interesting to them. They’re hearing about it at conferences. And I think a big part of what I’m going to get into today is how our tool can allow you to actively start using AI in your organization. So I think that this is a great group of participants here that are interested in AI, but maybe are looking at how to use it.
Geoffery Metz
Yep, and I think we’ll have a lot to learn today. So let’s go ahead and get to today’s topic and get a little background on what we’re talking about today. So claim letters commonly take insurance desk adjusters hours to complete. And workarounds to shorten that time beyond rigid templates can be costly. The typical cost of a litigated claim versus a non-litigated claim can be 4x. Many carriers have sought AI solutions to this problem, but homegrown solutions produce many AI-induced hallucinations. to require adjusters to review deeply and correct anyway, mitigating any potential time savings. For this webcast, we are covering the story of how the cycle can be broken using some actionable AI strategies that we’ll share today and that have driven some adjusters to a measurable return on investment, the 200% that we were talking about in the title. And now, Sai, I think you wanted to get us started with a demonstration on how AI can drive these gains. Feel free to take it over the floor.
Cylus Watson
Yeah, thanks, Jeff. So I should be screen sharing now. Can you see my screen?
Geoffery Metz
Yes, we can.
Cylus Watson
Okay, great. I’ll pause for a moment. You might want to make the screen bigger on your console. I know that it can be hard to see because I think it’s by default pretty small. So just kind of drag that window larger so you can see what I’m doing here. While you’re doing that, I’ll give a brief explanation of what our tool does. So essentially, as Jeff said, some of these letters that you write to your customers, especially the complex ones that are citing specific policy provisions to justify a denial or any sort of complex letter that you’re writing to your customer describing your decision on a claim, can take adjusters 45 minutes an hour. And we also know that even if they spend 45 minutes or an hour, they will also, they will still contain mistakes. So what our tool does, it allows adjusters to write a letter quickly while maintaining that accuracy that is needed within your organization to avoid things like litigation costs and settlements that we know can be expensive. And this is particularly valuable in a catastrophe scenario where you need to handle a lot of claims quickly. And Jo is going to talk a little bit about that later. I think the easiest way to understand the tool, though, is just to see it in action. So I’ll show you how it works. This is what the adjuster would see when they sign into our tool. What you’re looking at is the web application, but we also do integrations into claims management systems. We do offer this first, so adjusters can see how they can start using this tool, see if it fits for them within their organization as an actionable AI product that they can actually use. So we make this very easy to get started. What they would do is put in the claim number they’re working on. So I’m just going to put in test. They would select the letter type that they want to do. So here for this demo, we have reservation of rights or settlement, but we can add any additional letter types that you want to include that you send out to your customers. So we charge by the claim, not by the letter. So if you have additional letter types, and you send out multiple letters per claim. If you tell us what those letter types are, we add it to your account and it’s no additional cost because we just charge for every claim that you use our product for. Here they’re going to add the state that they’re operating in to select the state filters and then the template that they want to use. So our templates are dynamic templates that work across a variety of denial reasons or claim types. Here I’ll just show you a partial denial so you can see the population of the financial table as well as the policy language and see the full functionality of the tool. So in order to use this, we need to give the AI some context about the claim that we’re working on. So we’re going to upload 3 documents. One is the general loss report. or the field adjuster’s report. And this is going, the AI is going to summarize this and present the facts of the claim in the letter. So the field adjuster came to your house on this date, this is what they observed, et cetera. We’re then going to upload the deck sheet. So the deck sheet has a section called forms and endorsements, and that tells us what specific claims are in scope for this, what specific policies are in scope for this claim that we’re working on. This ensures that the AI never has access to any policy language that isn’t part of this customer’s claim, which is important because if you’ve worked in claims, especially catastrophe, you know that adjusters will cut corners sometimes and copy and paste from an old letter or do things to write a letter quickly and end up sometimes citing policy language that isn’t part of that customer’s claim, which can cause a big headache down the line. And then we’re going to upload the estimate. And the estimate has all that financial information we need if we’re making a payment. This next page is where the adjuster is telling the AI how it’s deciding to handle a claim. So we don’t make a determination for the adjuster. The AI is simply allowing them to write the letter faster and more accurately after they’ve made a determination. So let’s say we’ve looked at the facts of this claim and we’re going to deny the ceiling. And it’s a fairly complex denial. We’re denying it for wear and tear, water damage, faulty inadequate or defective maintenance. And let’s say they also had evidence of existing damage. So these are our common denial reasons. These are things that the carriers that we work with are often denying for. These are fully customizable though. So if there are things in here that you don’t want to include in your common reasons that your adjusters see, we can remove them. If you have additional common reasons that you want to add, we can add a checkbox for those. We also have like seasonal or regional checkboxes. So let’s say that a hurricane hits and so we will, when a hurricane hits, add several checkboxes like tree removal, off-premises power outage, flooding, things like that, are specifically valuable in a catastrophe scenario. So these are fully customizable checkboxes. And also, if you had an account to do like automobile letters with us, you would see different checkboxes for the common reasons for that. If the adjuster doesn’t see anything that fits their common reasons, they can always write stuff in. So they’re always welcome to add other reasons here and write them in, and the AI will find the best possible language to cite for the reason that they wrote in. We always like them to use the common reasons and for you to inform us about the common reasons, so we can make sure that it’s just an easier user experience for them instead of making them figure out how to write in what they’re describing. But for this one, I’ll show you an example of both. So we’re going to deny this for wear and tear, water damage, faulty inadequate or defective, and existing damage.
Geoffery Metz
Now, sorry, real quick, we did get a quick question here. And I actually want to mention to the audience, please go ahead and submit questions. During this demo, we’ll actually have a little Q&A session also at the end here. But I did get one in that I just wanted to ask you. So how are your customers confident the AI is citing the correct policy language?
Cylus Watson
Yeah, great question. So before we would make this live with your organization, we would run thousands and thousands of examples through the AI with your policies, right? So we would load all of your policies into our system. We would find all of the combinations that people could hold, so base policies, special provisions, endorsements. And then we would run our AI across, thousands of examples of those and run it through something we call critic, which analyzes the quality of the policy language that was chosen and the letter in general. So what we would do is we would run those tests, we would run it through our critic, and then we would fine tune the AI to improve the responses. And we would do that through your, with your feedback too. So Different carriers have different perspectives about what should be cited for different denial reasons, and let’s say how much policy language should be cited. And so what we would do is we would run it through Critic, we would show you the results, and we would show you the actual policy language that we’re citing. And you might say, hey, actually, you know, for water damage, we usually only cite numbers one and two, but not one, two, three, and four, or whatever your preference is. And then we would tweak the AI, run the test again, and then show it to you. So before it ever goes live, we do a lot of testing to make sure that the policy language is both accurate and what you want for your voice of your carrier. It’s a great question. Yep. So here, if you were adding a primary counsel to the letter, in addition to the recipient, you would add that information here and it would add it to the letter. So I’m just making up a law firm here. Same with here for carbon copies. So let’s say that you’re sending this to the mortgage company as well or any other recipient that you want this included on this letter. And then this is a page they can use to provide any additional financial information that wasn’t on the estimate. So let’s say that there was a prior payment or something like that. Instead of editing the payment table in the final letter, they can add it here and it would automatically include that. And so you can see here, now that we’ve made it to the final step, the policy language has already populated. So let me kind of walk you through the different sections here. So we have this intro language that once again is fully customizable for your liking. So this one says, your policy of insurance provide coverage for sudden and accidental direct physical loss to your property. Although damage to your ceiling was observed, no coverage was found. And then it says, your policy includes coverage from loss caused by, and then the different things that the adjuster put in that excludes them. We also tell them to direct their attention to the base policy. as amended by the special provisions of Florida and the existing damage exclusion endorsement. So you can see here that the AI identified the base policy, the amended language from the special provision, and an endorsement that included some information about existing damage that it pulled in. And I’ll show you kind of down here how we did that. So you can see here, just in the perils insured against section, we do not insure, however, for loss caused by any of the following wear and tear. Then we move down to the exclusion section. And you can see here that water damage once he is deleted and replaced by the following. So the AI saw that there was a definition of water damage in the base policy. It then went and looked at the special provision and to see if there was any instructions about how water damage is amended from that document. And it did find an amendment and cited the correct version of water damage. If this customer also had a water damage exclusion endorsement, it would then also find that exclusion endorsement, which may further modify the definition of water damage. So you can see in a catastrophe where water damage is a common policy provision that we’re citing, and maybe a new adjuster that isn’t super familiar with the policies is writing a letter, they might not know to check the base policy, then the special provision, then the endorsement, and which of those versions of water damage to cite. The AI does all of that for them. So all they need to know is that they’re denying for water damage. They don’t need to know exactly where to go through the process of finding that that definition of water damage, which is one of those things, like I said, that makes it so that it reduces the total amount of time that they spend writing the letters while still maintaining or improving the level of accuracy. And then going further down, everything in order here, we have the faulty, inadequate, or defective language. And then you can see here once again, the water damage exclusion endorsement was identified and the following exclusion is added to this section, which is the definition of existing damage. So you can see here that the AI identified all of the relevant documents and then put all of them in the right order to cite to the customer. Just this alone might take the adjuster 20 minutes to do. And I think that would be a short amount of time. You can also see here all the building blocks that went in. to creating this language. So I see this as similar to like when you take a math test and they force you to show your work, it’s not just that you got the final answer right, it’s that you show how you got there. So for this here, you can see that we identified the definition of water damage from the base policy. We also identified the definition of water damage from the special provision. And so the adjuster, that whole process that I just talked you through, the adjuster can see that happened. They don’t just trust that the AI got the definition of water damage right. They can actually see the logical steps that took place to get to that final definition of water damage that is cited on the final letter. So this is useful for the adjuster writing the letter. It’s also useful for team leaders or QA or whoever wants to see why the AI chose the language it did. And then you can also see the AI reasoning. So we’re fully transparent about the reasoning that the AI used to make its decision about what policy language to cite. So down here with water damage, the user’s request for a denied ceiling due to water damage, the ceiling is considered part of the dwelling coverage A, so identifies the right coverage to apply. It located the water damage exclusion in the base policies. Additionally, the special provisions of Florida document explicitly deletes and replaces this section with updated language. So, and then it says it provides both the original and the replaced language in part of the building block. So you can see exactly why the AI made the decision it did to cite that language. So once again, valuable to the adjuster, but also valuable to anyone on your team that wants to make sure that these letters were written accurately.
Geoffery Metz
So let me just ask real quick. Can the carrier customize the voice the AI writes in? And I think you touched a little bit about customizing the language, but can you customize the voice?
Cylus Watson
Yeah, they can customize the voice. So the policy language, as most people will understand, is very, you’re just stating the policy language, right? You don’t have much of a voice there, but down here is where you can see where you can modify the voice, right? So the claim overview is just like a basic explanation of the claim. saying what cover, whether coverage was found or not. Then you have the claim summary here, which is that summarization of the field adjusters report. So here’s where you can really customize your voice, right? So this claim summary can be written in a variety of ways. We get a lot of different nuance, you know, asks about how this should be phrased, whether it’s, more, just stating the facts or if it’s a little more empathetic or things like that. And then you can also tell us what things to include or not include in this. So if you always want to include the date that the inspection occurred, we can do that. If you never want to include certain things, like I’ve had carriers say, never state in this summary that there was evidence of mold. Always say that we, you know, we’ve looked for evidence of mold. They’re just like certain, you know, certain things about your voice that may not just be about the voice of your carrier, but about protecting you legally by not getting into too many of the specific facts of the claim that might be sharing too much information that can be kind of latched onto later by an attorney that wants to fight the claim. So if the adjuster says, okay, great, this is what I want to include in my letter, they’re going to hit this submit button. And it’s going to download to their computer as a docx file. So I’ll just go ahead and open that in my web browser. So it’s easy to see it. And you can see the final letter now, right? I’m going to assume yes. Okay, am I still there? Sorry, can you, are you looking at the final letter?
Geoffery Metz
Yeah, we can see you. Sorry.
Cylus Watson
Okay, great. So you can see here, this is just an insurance company that we created for this demo. Up here on the top right corner is all of the information of the adjuster that created the letter. So that’s my information right here with some fake information. Down here, you can see where we populated just kind of like the basic info about the claim, which saves the adjusters a bunch of time and ensures accuracy for things like the policy number, the date of loss, the property location. All of these things that are not groundbreaking, but they just save the adjuster time and it ensures they don’t copy and paste something over. Or let’s say they’re copying, they copied and pasted from an old letter to write this new letter and they accidentally forget to change the property location. Just a silly mistake like that could cost you a bunch of money down the line that you don’t need to deal with. Everything that is in blue was generated by AI. So we do that to let the adjusters know exactly which parts of the letter they need to carefully review. And when they review it, they can always edit it too, right? So if they don’t like this last sentence, they can delete it. They can add more things here, right? So they can always add or remove things from this letter. It’s not like we generated with AI. and then they have to send it out exactly how it’s generated, they can always edit this letter. And we encourage them to, if there are specific facts about the claim that the AI didn’t capture in this summary or things like that. Everything that’s not in blue is a template that you can customize with us. So you can see down here in Florida, they have this notice of right to mediate that we need to include on every letter. If there’s any brochures down here that you want to include, anything, on this letter that you want to include in every single letter, we put into this template that you fully customize and we work with you to customize. So you can see here we generated the… In summary, the field adjuster conducted an inspection on this date. You advised that you experienced roof leaks and interior damage following a wind event on September 26, 2024. The field adjuster inspected the roof, pool cage, baseboards, drywall, et cetera. So you can see that this was all summarized with proper spelling, proper grammar, all of the facts of the claim. You don’t need to worry about any of the things that might be common with adjusters when they’re writing a letter quickly that need to be revised for this summary. We’ve also populated the entire financial table, so they don’t need to do that. This has all of the information about the claim financially that we’re paying out. And then you can see down here, this is all of the policy language that is used to justify our decision to deny the ceiling. So we’re denying for wear and tear, water damage, et cetera, all the stuff I showed you on the previous page. So it’s that simple to write a letter. This is a fairly complex letter that might take an adjuster, like we talked about earlier, 45 minutes or an hour to do without our system. We just did it here in a few minutes. It took a little bit longer because I was walking you through everything. But you can see why this would be valuable for any adjuster writing a letter for any claim, but particularly in catastrophe scenarios where you need to handle a lot of claims quickly, this can be really valuable, especially getting those newer adjusters up to speed, and Jo’s going to talk about this a little bit, as the president of an IA firm, when you’re bringing on these adjusters and they need to handle a lot of claims quickly, and they’re still getting up to speed on which carriers they’re looking at the policy language for, it can be really complex. And we allow them to write these letters quickly while maintaining accuracy. So I will stop there. I don’t know if there’s any more questions or things that people were confused about or wanted me to dig into a little further with our app, but that’s a basic walkthrough.
Geoffery Metz
We actually did get a couple questions high, so why don’t I go ahead and get to the first one? I did mention, though, we are going to have a Q&A after each segment here. So this is the demo segment. So please get them in and we’ll try to get them as close to the demo as possible while it’s fresh for everybody. First question I have, so what happens to the claim data after I use the tool? Is it stored, shared, used for training? Furthermore, will there be options to opt in or opt out of how much data can be collected?
Cylus Watson
Yeah, exactly. The last part is exactly how we do it. So you can opt in or opt out for if we’re storing your data. So by default, what we do is we get rid of the data as soon as it’s used to generate the letter. So we won’t even we wouldn’t even store a copy of the letter that we generate, because we don’t want to hold any PII from your customers, and you don’t want us to hold it either, like when we first engage with each other. If you’re using the tool and you like the tool, we do store the letters for several of the carriers that we work with in order to improve how those letters in order to improve your letters over time, right? So one of the best pieces of data that we can get is look at the letter that we generated and then look at the letter that you sent out to the customer, the final letter, and then look at the differences. And we can use that to train the AI to get closer to the the final version of the letter that you send out. So we will store a copy of the letter for 30 days if you allow us to and analyze it to make those improvements. But if you don’t want to do that, then we don’t even store anything by default. We use the data to generate the letter. We generate the letter, it downloads to the user’s computer, and then we don’t store the letter or any of the PII used to generate it. That’s a great question.
Geoffery Metz
Great. And then another question I have here, is there a way to monitor the claims my adjusters are working on?
Cylus Watson
Yeah, another great question. So we take a very high touch approach to customer success. So with any AI tool, there is a learning curve and we want to make sure your adjusters are getting the most out of the tool that they possibly can. So we have a great customer success team and they monitor this page and you also have the ability to monitor this page. So anybody that wants to see a claim that’s being worked on can come here and you can see everything that happened. You can see the time, the user, the claim number, the item being denied, the reason it was denied, all of the policy documents that were used to generate that response. You can see here any documents that were skipped or ignored. We often use this column to request documents from carriers. So this means that the AI recognized that a document existed. but it couldn’t find it in our database. So then we’ll reach out to you and say, hey, could you send us that new special provision or the new endorsement you just came out with? We don’t have a copy of that yet. We also look at the generation for the policy language, the building blocks, and the AI reasoning. So you have access to all of this. We also use this to get your adjusters up to speed on how to get the most out of the tools. So like I said, we hope that they always use the checkboxes and you tell us what checkboxes to include. But we all know claims are nuanced and sometimes there are things that they want to deny for that aren’t part of their checkboxes. So in that case, they would need to write in the reason. And we monitor this page for users that are writing in their own reason. And if the reason isn’t going to get the most out of the AI, we’ll reach out to them and set up a 5, 10 minute call to give them some pointers. on how to use the tool. So for example, if I had put in the item as ceiling, and then I put the reason in, and I just said the phrase not covered, like the ceiling is not covered, the AI wouldn’t really know what to do with that information, because yes, it’s not covered, but that could be for several different reasons. So it forces the AI to just provide a really wide variety of reasons for the ceiling not being covered, when in reality, you want to be more specific about what you type into here. So we’re always monitoring this page. We’re always reaching out to your adjusters, with your permission, of course, to give them better information about how to get the most out of the tool.
Geoffery Metz
So let me ask, how long does it take for an adjuster to learn how to use wheelchair?
Cylus Watson
So it’s very quick and we do that on purpose because we know during these catastrophe scenarios, you might need to onboard. we’ve onboarded hundreds and hundreds of users in a week or two weeks to get them set up with the tool because we know that’s what you need to do when you’re preparing for a catastrophe. So it’s a couple days. We provide some training sessions. We go over a letter in real time showing them how to use it. And then they And then they always have access to us through, we’d like to get integrated into whatever chat system you use or Microsoft Teams or things like that. So they can reach out to us or we can reach out to them about, helping them better understand the tool. We also have this feature down here where you can ask us anything, right? So I am having a hard time using the tool. And then if I hit enter here, it would go right to our customer success team and it would tell them what user is having the problems. We would reach out to them and make sure they’re getting the help that they need. So it’s a pretty simple tool to use once you get the hang of it. And if they are struggling with the tool, we provide different ways to make sure that they get up to speed quickly to get the most out of it.
Geoffery Metz
Great, great. And then so how much time does it actually take to more generally set up Voltaire?
Cylus Watson
So the biggest bottleneck for setting up Voltaire is collecting the documents that we need to get your organization set up to use the tool. So we generally like to try to generate the first letter with Voltaire within 10 days of engaging with a new carrier. But that requires the carrier to provide us with your policy documents and provide us with some examples of your general loss reports so we know where to get the information we need. So there’s that early back and forth between us and the carrier where we’re collecting a bunch of documentation, you know, fine-tuning their organization for their needs, and then and then getting them set up and generating the letter. So there is a bit of a back and forth early on. But once we have all of those documents, it’s pretty smooth sailing after that. And like I said, we shoot for 10 days after you engage with us to generate your first letter.
Geoffery Metz
Okay, and then I have one more question here. And just a reminder, if you didn’t get your question in during this portion of the Q&A, we also do plan on having a Q&A session after the general interview section towards the end. So we may be able to get to it there, or alternatively, we may send you an e-mail response afterwards. So Sai, is it possible to integrate Voltaire with claims management systems?
Cylus Watson
Yeah, so that’s always our goal is to do an integration with the claims management system because if you do that, then they don’t need to take this step of uploading these documents. We grab that information directly from the claims management system and they also would never need to leave the claims management system to a website that’s outside of that system to deal with the claim. So it just makes it a little bit easier for the adjuster to use the product if we can integrate into, let’s say, Guidewire or other claims management systems And we are currently doing that for our carriers. Like I said, this is, we want you to get started here because we pride ourselves on being really easy and cheap to get started. Like we think that once you use this tool to generate your letters, you’re going to see a ton of value in using this instead of the traditional way of generating letters. And once you get the hang of it with this web interface, then we can move on to a claims management system integration. But the people on this call probably know that these IT departments in these carriers are often backed up and won’t necessarily have time to do that. So we don’t want to wait until you’re ready to do an integration to allow you to get started. We want you to get started here. And if you like it, then we can work together to figure out when is a good time to integrate it into your current claims management system.
Geoffery Metz
Sounds great. Thank you, Sai. Thank you. And now, before we get to our actual interview portion of this, let’s go ahead and get to another poll. So what’s your biggest concern about implementing AI for claim correspondence? accuracy and E&O exposure, adjuster resistance adoption, integration with existing systems, cost, ROI uncertainty, regulatory or compliance issues, or loss of the human touch. And again, we will just take about maybe 30 seconds to get those results in, so please answer as best matches your experience, and we will go ahead and push those results live and get to the rest of our event today. So what’s your biggest concern again about implementing AI for claims correspondence? accuracy and E&O exposure, adjuster resistance, adoption, integration with existing systems, cost, ROI, uncertainty, regulatory or compliance issues, or loss of the human touch. I’m going to give this about 15 more seconds, so if you haven’t answered, please go ahead or respond. Please go ahead and get that in, and we’ll push those results live.
Cylus Watson
Okay.
Geoffery Metz
We’re getting pretty close to the end here. Okay, so let’s go ahead and push these results. And it looks like the big one, from what I can see here, although it’s really widespread, is accuracy and e-mail exposure. And that’s, Jo and Sai, that’s about what you’ve seen from your experience. Was that kind of what you saw?
Cylus Watson
Yeah, I think that any time you’re thinking about AI systems, especially when you’re talking about something as important as citing the right policy language, accuracy is going to be your biggest concern. And that’s why when we engage with new carriers, we make sure that they are comfortable with the accuracy before we ever make it live with the adjusters to use on real world claims. Because we know that that’s what gets you in trouble with letters. Like we built this product because we know that letters that are inaccurate and contain the wrong policy language, it is what costs you money in litigation costs and settlements or things like that. And Our product improves on the accuracy that human, like we know that we’re more accurate than the average adjuster writing a letter, but you don’t know that yet, right? So it’s not surprising to me that people are concerned about accuracy. I think that once they start using our system, they will see that we actually increase accuracy within their organization. It’s not something they need to worry about with us.
Geoffery Metz
Okay, great, great. Thank you so much, Cy. And now let’s go ahead and get to our interview portion of today’s event for some actionable insights. First, we welcome in Jo Veras from TSI Adjusters. Jo, it’s great to have you.
Jo Farris
Thank you, Jeff.
Geoffery Metz
And so now let’s go ahead and get to get started here. Can you take us back to the beginning? How did you first discover Voltare and what was the business problem you were trying to solve? Maybe walk us through how you made the decision to pilot it with your team and then how you approached introducing it to your carrier clients?
Jo Farris
Absolutely. So I was introduced to Voltare through an industrial relationship and for all of us in this industry, whether you’re on the carrier side or you’re on my side as an independent firm that does TPA work, the desk adjuster piece is such a critical piece. And when there’s large events and you’re bringing in a bunch of desk adjusters, the letters were always a challenge, right? We’re working for numerous carriers. Desk adjusters haven’t worked for these carriers before. Everybody has a different format, a different way they want to do things. Some templates are great, some are not so great. So there’s just always that challenge. And then on top of it, getting a good, accurate letter, like you guys were talking about a minute ago with Cy, just even down to the grammar, spelling. I mean, I know you guys know, if you see letters, sometimes you would be shocked at the result you’re getting. So one of the things that we had the challenge with was just the time factor. You know, we’d send a letter in for approval and it would get kicked back for citing the wrong policy language or having grammatical errors, whatever it may be. And so when I was introduced to Voltaire, I was very curious and very interested in the product and had a lot of the same questions that have been asked already on this webinar. And so there was some hesitancy there, but I wanted to see how it worked and what they did. And so when I saw the demo, I was highly impressed. But then again, with AI being such a hot topic in our industry right now, there’s still so much hesitancy of does it really work? Has it been tested? Has it been tried? Has it been proven? You know? And so that was what I saw quickly with Voltaire. And so what they gave us the opportunity to do is test it with just our immediate group of desk adjusters. let us run some trial and do some letters across the board through denials, the partial denials, payments, whatever it may be, so that we could see how it worked on our end before I even presented it to one of our carriers. Once we were confident in that product working, we took it to one of our carriers and said, hey, we want to try this with you guys because we have a pretty much a permanent desk spot with this carrier. And so we really wanted to show them how well it was working. And then quickly after that happened, one of the storms hit Helene, right? And so then we’re onboarding 60 adjusters plus for this carrier. And so when they saw the accuracy of that, then it was, hey, we want TSI to use this product with all of their desk adjusters that they’re bringing in because they were seeing how much time was being saved. I know from prior storms when we would do stuff for this carrier, the time factor involved, and I don’t even know how often insurance companies really realize how much a letter gets kicked back and forth to fix and make corrections. And so it was just a huge asset for us to utilize this product and the carrier between Helene and Milton last year, the thousands of letters that we created and the accuracy that was done with Voltaire and the time that it saved was just a huge asset for the carrier and for us of being able to utilize that product with them. So we were grateful for that opportunity to team up with Voltaire on those past storms and then continue to use it now in our day-to-day with a couple of our clients.
Geoffery Metz
That’s great. That’s great. And I know a big question we have often is just about adoption and building a culture of adoption. So what were your biggest lessons learned in getting adjusters to actually use this tool. Were there specific incentives, training approaches, or communication strategies that moved you from initial skepticism to near universal adoption?
Jo Farris
So yes, some of our adjusters, first and foremost, the carrier, again, with the hesitancy with some of your all’s questions already, just really how often does this work? And then somebody had brought up the E&O issues and things like that. So I think the biggest thing what we had to teach our team and our desk adjusters and even our clients was AI is not replacing the desk adjuster’s job, right? The desk adjuster still has to review the letter, has to approve the letter and make sure that it’s done right. AI is just there to enhance the experience and the product. And that is where that accuracy came in. So when we were able to show and prove that, helped them look at it in a different light. The desk adjusters, however, especially you guys know some of these old school ones that have been doing this for years, they’re set in their ways, right? So getting them to make that change, there was a challenge, if I’m being very honest. They, if they, especially if they had already worked for the carrier prior, or they’ve been doing, you know, that same job for a while, Nobody likes change. And so when it was introduced to them, they were like, this, that, whatever it may be, that was their challenge. But at the end of the day, when they were told they had to use it, made such a difference because they saw what a great result they were getting. And then of course, when new people were being brought on, there was less of a training process for them, right? To be able to teach some of these new adjusters coming on because a lot of that heavy load was lifted for them. And so when we were onboarding a bunch of adjusters, like Cy said, there was a less of a training challenge there because there wasn’t, you had to go to this and grab this template, you had to do this, do that. There was already a set program in place that it was just basically choose your options and let the letter be created for you and then make your enhancements or adjustments as needed. So it had its challenges, but proved to be successful in the end.
Cylus Watson
Yeah, cool part about that story too, Jo, is after you used our tool to handle the storms for that carrier, the next January of the next year after those storm claims have been handled, we flew down to their office and they wanted to onboard all of their daily adjusters to use our tool as well. So we kind of like cut our teeth in those storms and proved out that this thing works. And they were like, okay, great, like let’s use this for every letter that we write within our carrier, which I think speaks to the improvements that these they saw over the traditional way they’re writing letters.
Jo Farris
Oh, yeah.
Geoffery Metz
Fantastic. Fantastic story there. So now, so now the next question is, what’s happened with your relationship with the carrier since Jo?
Jo Farris
So, well, it enhanced it greatly, you know, Several of our carriers grade us on our quality and the work product that we are providing for them. So we were able to improve our QA scores across the board because we are graded on those letters that are going out and the accuracy and how many times it’s getting kicked back and whatnot. So it improved our scores greatly to where we ended up being #1. with this particular client. And of course, enhanced our working relationship and almost a trust factor too, right? Like, hey, TSI was willing to try something new with us and showed us a product that worked. And so it’s also opened up, you know, great conversation pieces with newer clients and potential clients for us. just to be able to have that conversation of, again, a product that we’ve tested, tried it, and now proven that it is actually working and enhancing the desk adjusters experience across the board for productivity and accuracy.
Geoffery Metz
That’s awesome. That’s awesome, Jo. And now, Cy, let’s turn the spotlight on over to you. So first question I have for you is, There’s a lot of AI hype in insurance right now. What specifically makes Voltaire actionable AI versus other AI solutions that claims departments might be evaluating? To put it another way, what’s the difference between AI that sits on the shelf and AI that adjusters actually want to use every day?
Cylus Watson
Yeah, great question. And I notice that when I go to these conferences and every booth says that they’re using AI. It’s all over every conference. It’s all over the talks. Everyone is claiming that they use AI and it’s going to make your organization better. What we did with this product is the first thing we did is we identified an area for improvement within the insurance industry that actually has a deliverable that you send out to your customer. So this isn’t, an AI that claims to provide insights or a summary of certain things or things like that, your organization may or may not use, that you don’t know whether you’re getting efficiency gains from the AI, we provide a complete deliverable that you actually see the results in your hand that you can send out to your customers. We help you not get behind on your deadlines, have better, more accurate letters, and you can see the results in real time as it’s being used. And the other difference about our product is that we’re hyper-focused on letter writing. We’re not claiming to do 8 different things pretty good. What we claim is that we are a company that is just strictly focused on using AI to write the best possible communications that you need that are compliant, that are consistent, that are accurate, and we’re just hyper-focused on that one thing. And we know from talking to carriers and from independent research reports that come out, The AI products that people use in their organizations that actually work are these products that are focused on a single thing and just approving the efficiency on that one thing. So I talk to, I guess the last thing that we do that makes this actionable is that we don’t just convince the executive team that this is going to make your organization more efficient. We built a product that works for both the decision makers and the end users that are actually using it. Because at the end of the day, if the users aren’t using AI, you’re not going to see anything from it. So all those pain points that Jo was just talking about of getting these old school adjusters to try this new method of writing letters, once they see the time it’s saving them, I mean, I talk to adjusters every day. And they tell me, this saves me two hours a day. Or one that resonated with me is that an adjuster had been waiting to write a letter for three days because the letter was so complicated. It had like 5 different reasons they were denying it, all these different documents to cite. And she wrote the letter in 15 minutes with our tool, including all the time it took to review the, you know, the reasons that she was denying it. And then she opens the tool, checks 5 check boxes, and then boom. boom, there’s the final letter to send out. So they don’t procrastinate writing these long, complex denials that maybe they were before. They have no anxiety about their, you know, writing the letter quickly and accurately. They just get it done. And so it’s not about convincing a decision maker that the tool is going to work. It’s about building a tool that the end user actually uses and ends up being more efficient because of it.
Geoffery Metz
That’s great. That’s great. I can really see the time savings there. So the next question I have. So, sorry, excuse me one second. This webcast title promises strategies to realize about 200% plus ROI. Can you break down where that return actually comes from?
Cylus Watson
Yeah, so if you’re curious about the Digging into the math behind that, we do have a business case that we’ve created in collaboration with the claims leaders and the organizations we work with. And we’d be happy to share that. That’s available on our website or I believe in the documents provided through this webcast. So I’m happy to dig into the math with you on that. But just kind of on a high level, the ROI comes from time and labor savings and reduced litigation. So what we do is we enhance your existing workforce with AI. So similar to what I was just saying, we don’t replace desk adjusters. We enhance your existing workforce and make them more efficient. We relieve tension across the entire organization, not just the adjusters, right? So if you know claims, you know that these bad letters, they’re more time for your QA department, they’re more time for your team leaders, they’re more time for your legal department that has to defend these letters that were written poorly. When you remove bad letters from that equation, everything runs smoother. So the entire organization gets to feel that relief of tension. So fewer claims are going to litigation, fewer customers are calling and asking for clarification about their claim. There’s just all these things that are downhill of letters being good that ends up saving a bunch of time. We also help your organization stay ahead of the curve when it comes to adopting AI. So we’re not the only ones using tools like this for letters, right? We know that public adjusters and plaintiff attorneys, whoever these people might be, are using AI to identify bad letters and then causing you problems when they identify them. So the, let’s say you’re writing a bunch of bad letters right now and your letters aren’t as good as they could be. A fraction of those are getting caught and you’re being punished for writing those bad letters, but not every single one. Like a lot of them slip through the cracks. The reality is that’s going to start happening less and less. As people are using AI to identify when these letters are not high quality, and then fighting back against them, it’s going to become more and more important that your letters are good just right off the bat. So this is the way to get ahead of the curve and make sure your letters are good now and increase quality using AI. So when the other side is using AI to identify when letters are bad, it’s not a problem for you. Your letters are already good enough that they don’t find anything because you just are always sending out good letters.
Geoffery Metz
Okay, great, great. And I think you started touching on this, but let’s go ahead and ask the big question. So how does Voltaire specifically address the accuracy and defensibility issues that keep them, everyone in the audience up at night?
Cylus Watson
Yeah, so the quote that comes to mind for this, I was talking to a claims executive at a large carrier and they said, what keeps me up at night is my worst adjuster. We are only as defensible as our worst adjuster. And you know, that resonated with me because we built Voltaire with this in mind. We try to provide guardrails that increase the floor of all of your adjusters. So things like the AI only has access to the policy language for this customer’s policy that we’re working on today. Another feature that the carriers love is that if an adjuster tries to deny a claim for a reason that isn’t in that customer’s policy, they get a message that says no relevant policy language found. So as a quick example, let’s say it’s a catastrophe scenario. One policy I work with denies unattached fences. The other one has coverage for unattached fences. If the adjuster goes into our tool and tries to deny, checks the checkbox for unattached fence for a policy where that’s actually covered, they won’t get policy language back to justify that denial. It’ll simply say no relevant policy language found. So that guardrail is huge compared to the alternative of, let’s say, copy and pasting from a letter they wrote last week for denying an unattached fence. They get another claim this week where somebody’s trying to claim an unattached fence and they go, okay, great, I already have a letter for that. Let me just change the name, the address, some of these things and send out that letter, not knowing that now they’re dealing with a different policy where those unattached fences are covered. So we don’t make the determination for the adjuster, but we stop them from making bad decisions and citing incorrect policy language when it isn’t part of that customer’s policy, which saves you a ton of time and headache in the future because we know that’s a problem if that letter got sent out. So that’s one of the things we do. We also, as I alluded to earlier, we run all of these letters through a tool we call Critic. We do thousands of examples of these letters before we even get started with the carrier. So that ensures that we are a very accurate letter writing tool before the first letter is even generated.
Geoffery Metz
That’s great. That’s great. Thank you, Sai. And that 1000 letters thing actually really stuck with me from the demo, too. So now we’re going to go to audience questions. And as I mentioned at the beginning, if you have a question for either Sai or Jo, please get it in the Q&A panel on your console. We’ll try to get to as many as we can. We’re getting close to the end, but I think we do have time for one or two more questions. And again, though, if we don’t get to your questions, we will send these over to Jo and Cy afterwards. So you may receive an email response afterwards. First question I have here, is Voltaire designed solely for first-party claims, or can it also be applied to third-party liability claims?
Cylus Watson
Yeah, so this same technology can be used across a variety of types of claims, insurance lines, so we can do, you know, We do homeowners insurance, we do auto, we’ll do commercial property. It’s the core of our technology is the ability to pull the right policy language out of a long, complex legal document where let’s say page 87 modifies page 22 and there’s an endorsement on page 96, which then further modifies language. Like that’s really hard for a person to read that whole policy or be familiar with it enough that they cite the right policy language. So we can tailor this system for any type of letter that you want to write where you’re citing these complex legal language, as long as we know, as long as you provide source documents that we can use to generate the parts of the letters that need to be generated, like the summary or the policy language, we can tailor this system to write whatever types of letters you write. In fact, last week I was talking to a carrier that was describing a new letter type to me. It was really complicated. And they said it took their adjusters 4 to 6 hours for a seasoned adjuster to write that letter, which is just like, We know we can improve on that. It might not be the thing where just right out of the box, we can generate that entire letter. We would need to work with you to figure out exactly the nuances of this specific new letter type. But there’s always ways that we can improve on the manual process of letter writing, especially when it comes to pulling the right policy language.
Geoffery Metz
Great. Thank you, Sai. Thank you. And actually, we are nearing the end of our presentation. We did get a lot of great questions and we’ll go ahead and send these over to Sai and Jo. So you may receive any more response afterwards. But I wanted to get one last kind of temperature check on everyone in the audience now that you’ve seen everything here. If you could achieve one outcome with better claim letters, what would it be? Reduce litigation frequency and severity, free up adjuster time for high value work, improve compliance and reduce regulatory risk, more claims with existing staff, improve adjuster satisfaction and retention, or reduce claims leakage. And we’re going to go ahead and just give you a couple seconds here before we close out. In the meantime, I did want to thank everybody in the audience for joining us today, and we hope this was valuable to you. Big special thanks to Jo and Cy for their time and insights, and Volterra for their support. Remember, if you missed any part of this presentation, this webcast will be archived at propertycasualty360.com slash So you can view it again or refer to someone else. And you can also explore other upcoming and on-demand presentations while you’re there. Now let’s go ahead and take a look at the results and just see what we’re looking at here. It looks like we got a lot of people free up adjuster time for a higher value work. And it seems like that really is just such a valuable, valuable takeaway from AI.
Cylus Watson
Yeah, exactly. And this is what we tell our carriers that we work with is, we’re not trying to replace your workforce with AI. We’re trying to free… Nobody likes… writing these letters. Like it’s not like this is like their favorite part of the job. They know, they get, they figure out how they’re going to handle a claim, and then they’re like, oh gosh, now I need to go actually write the letter describing my decision. And so we’re removing the most annoying part of their job that they don’t like, and we’re freeing up their time to go, you know, have more time to examine the facts of the claim, to look at the pictures involved, to make sure they’re handling it right, to talk to the customer in a friendlier way. because they’re not stressed and they’re not staying late at work to get these letters done. It really is about improving the efficiency of your workforce, not replacing them and freeing up their time for not just higher value work, but work that they enjoy more. Like this is not, this is their least favorite part of the job and we’re replacing that for them, which is going to make them, you know, a higher efficiency employee for you.
Geoffery Metz
Fantastic. Thank you so much, Cy. Thank you so much, Jo, for your time. Thank you for everybody again in the audience and have a great afternoon, everyone.
