Filevine Unveils AI MedCron for Medical Record Timelines (Full Transcript)

[00:00:00] Speaker 1: We're very excited to have you all with us today. Thank you for joining us for the latest announcement of Filevine's AI-powered Medical Chronologies and Summaries. A little bit about the individuals that are going to be presenting today. You'll see some familiar faces and then some new ones as well. We're going to have Ryan Anderson with us today, Jonas Kratosvill and Madison Doyle, who will be doing a demonstration of the product. Going through the agenda that we're going to cover today, we want to give a really deep dive overview into how we created this product and how we came to the conclusion that this will be a beneficial product for our customers. We will do a live demonstration of the product so that you can actually see how this is going to affect you. We will also cover the enablement, the big question of how do I get this product enabled for me and actually use it? Then we have some time carved out for Q&A. Now, during this whole presentation, we are also watching that Q&A. We have additional staff members who are going to be answering questions throughout. Please feel free to chat in, ask questions during, and then last, we'll touch on some future developments that we are focused on. With that, I will hand it over to Ryan Anderson.

[00:01:13] Speaker 2: Good afternoon to most of you except those on the West Coast. Great to talk to everybody today. I am absolutely thrilled to announce this new product. We think you are going to love it. But before we do, I want to give a little bit of the reasoning why we felt this product was so important to build and how we built it and understand a little bit of the background for where we think this fits into the work that you-all do on a daily basis. The vast majority, if not everybody on this call, is a legal professional and it is a legal professional's job to spot issues. Those of you who went to law school know that your job is to spot issues. You were trained on it, you're good at it. But you can't spot an issue if you don't see the issue. It is absolutely critical that you see the issues, otherwise you can't handle them. No matter how good of a lawyer you are, no matter how good of a paralegal or a case manager or a law firm owner you are, if you don't know what's going on in the case, you cannot affect the outcome. Whether you're an insurance company, and we have today many insurance defense customers, or whether you handle plaintiff's cases, in both situations, it is likely the issues that you never see that most affect the outcomes to a case. The reality is that personal injury cases require legal professionals to sort through thousands of data points and documents. We built Filevine because we believe that Filevine has the best tools to categorize and organize and collect data about your cases, and then reflect it back to you when you need it most. But through all those data points and documents, oftentimes critical information is missed. Because obviously, evidence has to be found, not just collected. A stack of documents in a binder or in a box or in a folder in Filevine is only as good as the person who has time to look at it, understand it, and spot the issues. We all know what happens if in a case, you might have an undetected brain injury. You could be doing your client a massive disservice if that wasn't surfaced by a good medical chronology in your case. For those of you on the defense side, you might've missed inconsistent pain reporting and signs of malingering. There might be gaps in treatment that could affect the client's credibility and could ultimately affect how much money gets paid out in the case. You might miss a potential opioid addiction. There might be a client who allegedly received overtreatment. And there's the classic, an accident that happened years before the one that you're handling that maybe your client didn't say to some doctors but did to others. Those are just examples of the kinds of potentially devastating or potentially critical pieces of information that might be found in documents and evidence sitting in Filevine today. And you as a lawyer, you as a legal professional are required to understand every single one of these facts and to surface them. Failure to do so can be devastating and yet the opportunity is also massive. So for us, building a tool like this, we believe has real world implications. The last thing we want your client to be is vulnerable. And a client who doesn't have a lawyer who knows all the facts is vulnerable. Vulnerable to misinterpretation, vulnerable to being hurt by poor representation and vulnerable to having their words and actions twisted. Unseen evidence and facts can completely change an outcome for better or worse. Our job, the job today at Filevine is to help you have a quality MedCron that can answer the unseen evidence problem and create the foundational document for your client's story. But medical chronologies are time consuming, expensive and labor intensive. We gave an example of our MedCron to a customer just yesterday. And the customer said to me, we don't do MedCrons on every case. In fact, we only do them on the very biggest cases because MedCrons are simply too challenging to build. We don't have time. I get it. Busy lawyers, busy legal professionals can't produce MedCrons on every case or it may be too hard. And that results in them only being created for high value cases. But what a miss. It's your job to know everything that goes on in a case, but also there may be missed opportunities on the downside and on the upside to understand what a case is like, no matter how big you might think the case might be in the beginning. And yet, as we've discussed, the process is a challenge. It's not like a MedCron is something that everybody has time to do. I don't talk to any lawyers ever. This has never happened. Filevine has well over 5,000 customers today. And I have never spoken to a lawyer or a legal professional who says, you know, my paralegals just aren't very busy. I never hear that. And yet, when a medical chronology has to be done, the lawyer is going to turn to that paralegal and ask him or her to build a MedCron. Now, you might use an outside company, but to even get to that point, you have to make sure to have all the medical records in your file. Importantly, those aggregated records might be jumbled up. They might be jumbled up with invoices from a medical record collection company. They might be jumbled up with records from other medical providers. They might have a police report intermixed in them. For any number of reasons, the records don't come to you clean. Now you have to get them in the right place. You might be slicing and dicing them to get them in a clean, organized fashion for whoever is going to produce the MedCron. You can't even begin the work because just the organization to give the work to somebody else is so challenging. Even if you do all that, you're then asking a paralegal or a legal nurse consultant or somebody else in your firm to review the records and draft the chronology by hand. It is my firm belief as somebody who believes that AI is critical to the go-forward world of legal that humans are essential, but humans can make mistakes too. So it is of the utmost importance that we use every tool at our disposal to produce an accurate MedCron. But even if you've done that, the best case scenario is that you have a word document returned to the lawyer that looks something like this. This isn't bad. It's far better than nothing. It's way better than not having done a medical chronology at all. But if this is all you're getting back, you're getting a static document that can't be configured the way you want it to look. You only have one views. What if you want to sort by date, provider? What if you want to sort, what if you want information about the client? What if you want to zero in on a particular set of injuries or a particular course of care? None of that can be done easily with a static MedCron built-in word. Also, this document is unintegrated. It's not sitting in your case management system. There is so much rich detail that comes in a MedCron and yet none of it is making its way into your case management system. Of course, the medical chronology itself, while critically important, isn't actually the point. The point is to do something with it. And yet what you're getting today is a flat document, not in your case management system, not searchable, not indexable, non-standardized format. And critically, it can be out of date the minute you receive it if new medical records come in. It's also inflexible, again, only one view. So for us, we think in the world of AI, we can do much better. So I am thrilled to announce MedCron by FileLine to uncover case-cracking evidence. We're going to show you the tool here in a minute, but I want to go over a couple last things. First of all, we are thrilled that you as a lawyer have the opportunity to use a tool that is now searchable, up-to-date, configurable, and integrated. Let me talk about a couple of these. First of all, the MedCron that you will see today is extraordinarily searchable. I think you'll have a ton of fun playing around with these MedCrons and looking at all the ways. You can not only hyperlink to the exact source document to where that treatment was found, but also zeroing in and finding exactly the kind of care that you want to look at. But number two, up-to-date. Today, we are announcing for the first time a new FileLine feature that is sort of sitting in the background and operates as a foundational technology for this product to work. FileLine now has a feature called document classification. All documents that come into FileLine for FileLine AI customers, and many of you we know are FileLine AI customers on this call, your documents will automatically be classified by type. We're starting with medical documents, not all documents, not all types as of today, but as of today, as you all get added and implemented on this platform, any document that you upload to a FileLine project will be automatically classified as a medical record if we believe it to be a medical record. Our testing shows well over 95% accuracy, approaching 97% accuracy on recognizing a medical record as a medical record. Why is that important? Because so many of you have said, hey, my files, while we want them to be perfectly organized in FileLine, sometimes a medical record comes in via email and sits in my FileLine, but it didn't get uploaded to the correct folder or the correct section. Others of you have said, my client texted me a picture of her medication or my client texted me a picture from her doctor visit. We will pick up all those, no matter where the document sits in your FileLine project. It doesn't matter the field, it doesn't matter the type, email, mail, fax, it does not matter in your document section, in a collection section, in a static section. We will recognize it automatically and build your MedCron as it goes. Of course, it's configurable. We can't wait to show you how it's configurable and integrated right inside of FileLine. This is a big leap forward for medical chronology technology, but what is so cool is that it's actually incredibly accurate as well. We believe that your MedCron should be simple, fast, complete, and accurate. Again, the process is that FileLine's AI doc classification automatically identifies any medical record uploaded to a project as a medical record. Really exciting technology. Then the MedCron product will summarize thousands of pages, often in minutes. We're not always able to do it in minutes, but often it's that quick. And then you can see the full case timeline and expand to individual summaries right inside of FileLine. You never have to leave the FileLine pane of glass. We want this to be a completely embedded experience for you. We are so excited to show it to you. Real quickly, I want to touch on a larger story. Those of you who follow FileLine have probably been able to tell our view is that you need a product and a platform that cuts across multiple stages of the life of a case. We started with intake with LeadDocket. Now with MedCron, you can investigate and collect and analyze data, especially medical record information. And of course, with document assembly, we have document drafting. We use this technology for Demands AI. Those three tools really speak well to those three stages, but those are predominantly pre-litigation stages. This year, we are much more focused on the litigation process. Pre-litigation with Depot Copilot, litigation with depositions by FileLine, and all the way to resolution with chat with your case. MedCron will support every single step of this case. And I just want to mention, Depot Copilot will give you real answers and can actually review what's in the MedCron as you're giving your deposition or as your client is being deposed or as your client is being prepared to be deposed. You could actually run an entire deposition with Depot Copilot to see how your client will do with the MedCron as reference. You now have the tools at your disposal to defend your client in ways you've never had before. And for those on the insurance defense side, you have ways to zero in on vulnerabilities in the plaintiff's case in ways you've never had before. But FileLine can actually now host the deposition. That's right. We can actually conduct the deposition for you with depositions by FileLine. And of course, if you haven't tried chat with your case, please do. Project AI customers already have access to chat with your case. It is our most popular feature we have ever built. We're super proud of it. To give you a sense of just how cool this is, FileLine can now, with MedCron, help you prep for a deposition, help you host a deposition, and then of course, after the deposition, analyze it. A MedCron is only valuable if you use it. And with MedCron by FileLine, you can leverage insights throughout the life cycle of the case. We've had customers large and small already using this product for a long time, including insurance companies. Travelers is a MedCron customer, and we can't wait to show it to you. But first, we want to explain how we made it. The team that made MedCron is incredibly talented, passionate, and skilled. And I want to introduce you to the leader of that team, Jonas Jovill, who will walk you through how we built the tool. Jonas, I'm so excited for you to show it to everybody. I will let him take it from there.

[00:15:21] Speaker 3: Good afternoon, everyone. Thank you for the introduction. I would like to briefly share my background so that you know what expertise I'm bringing to the MedCron tool you're seeing today. I was born and raised in Prague, Czech Republic, where I currently reside. I came to LegalTech through an interesting path. I first completed my undergraduate degree in economics and finance, but I've always been fascinated by finding patterns in data and discovering new insights hidden within them. This passion naturally drew me to computer science and AI during my graduate studies, where I could build systems that actually understand and interpret these patterns. This combination turned out to be very useful for me for understanding both the product side as well as the technical side of building AI products. I've spent my career working on applications where accuracy really matters. And I learned that in these high-stakes environments, precision is critical. I first worked at the U.S. Embassy, helping American companies establish and expand their businesses in Czech Republic. Then I was one of the founding members of startup developing speech translation system for the European Parliament that enabled translation of parliamentary hearings in real time. For the past six years, I've been part of a LegalTech startup, Perot, which has recently been acquired by Filevine. In my role, I focused entirely on leading development of AI for legal applications. I've been core part of building our state-of-the-art automatics transcription system for depositions, document analysis tools, systems for automation of internal processes, and more recently, the MadKron tool you're seeing today. MadKron reflects what we've learned throughout the years about building AI that attorneys actually want to use. One of the most important things we focused on is making sure you can verify everything. When MadKron tells you something, it shows you exactly where that information came from, the specific page and the exact location within it, so that you're never left wondering if the AI got it right and can always verify its outputs. We've also learned that the best way to build legal tools is with lawyers involved from the very beginning. We had over 50 attorneys testing early versions of MadKron and telling us what worked and what didn't. This feedback was instrumental in a way how the product currently works, and we plan to utilize the feedback from all of you who will be using the tool for improving it even further. Now, don't worry, I'm not going to bore you with complex code examples or math equations, but I do want to help you understand the core concepts behind what makes MadKron work. At its core, the system tackles what might seem like relatively simple task, but it's actually quite complex. We break down medical records into individual documents. Then for each document, we identify what we call Trinity of MadKron. It's the main date of the document, medical facility, and the providers. Each document is then carefully summarized while maintaining links back to the original sources. All of this complexity gets distilled into a simple, intuitive, yet powerful interface that's designed to save you time while giving you the control and verification you need. We've seen this product being successfully used by our clients, ranging from small firms all the way to large organizations and insurance companies, and we are confident that the success will continue with final users. I will now pass the work to Madison, who will walk you through the MadKron tool itself.

[00:19:33] Speaker 4: Jonas, thank you so much for explaining how the MadKron process has been built out and the care and consideration as well that went into the development. I'm really fortunate that I have the pleasure of showing the tool live. So we're going to dive right in to how this is going to work within your Filevine. So I've launched us right into an existing case, our Catherine Grace case, here inside of Filevine. And you'll notice there's a new section here, our MadKron by Filevine. And so this is a new built-in section that will allow you to access the medical chronology. Now, I want to reiterate a few things that have already been brought up because there's been a lot of questions already.

[00:20:16] Speaker 1: So let me be clear.

[00:20:17] Speaker 4: We are automatically identifying what is a medical record from the documents added to your Filevine case. I'm trying to be really careful and say added to your Filevine case because I don't want to suggest that it's only documents uploaded. Because like Ryan said, it's faxes and emails and direct uploads. And so, yes, even if that medical record is integrated amongst medical bills and non-medical documents themselves, the tool will extract out the pieces that are the medical records to process through the MadKron that I'm about to show. Now, a few other things that I just kind of want to cover before it is that I show the MadKron. This is but one tool in the entire Projects AI umbrella, ecosystem maybe I should say. And so we are focusing right now today on a tool that is trained to extract out medical treatment and the like, not quite medical bills. So just know that what I'm about to show is focused on the medical treatment, but that we absolutely have other solutions in place for handling questions around medical bills and extracting out charges. So those tools include some of the AI data mapping we have, our ability to chat with your matter. So again, keep the questions coming, but know that what I'm about to show is really focused on the medical treatment itself. So when you have this, it's a built-in section. And again, when it's enabled, you're going to be able to see that the documents are processed automatically because of our document classification. That is a big change from how we have covered and used our AI tools in the past, right? So instead documents come in, no further action really needed on your part, but once they are ready, you will be able to launch into the MadKron here. So let's take a look at that MadKron. So here's our medical chronology of that same Catherine Grace case. All of the medical records have been consolidated together into a kind of single packet. The records themselves obviously were classified automatically by the tool. And so you do not need to sort them for the tool ahead of time. Instead, what the tool will do is sort them for you so that you can get at a true chronology. But this is more than just a chronology, right? These are also summaries. And so across the board, you will find for each case that your records are, the way I've been thinking of it is, you know, analyzed or looked at through four different lenses. And I'll talk about each of these four, but I think the first two and the last two kind of go together. So in the case overview, as the name suggests, you're going to get a bit of a holistic view of the contents of the medical records themselves. And the tool also generates a kind of a title, if you will. And so we have a multi-trauma motor vehicle collision. That might already be a piece of information that I didn't know coming into the records, that this is a multi-trauma motor vehicle collision. So it only makes sense that we probably want to drill down and figure out why is it that this has been identified as a multi-trauma. And one of the sub topics under the main topic here is that there were multiple motor vehicle collisions that took place actually within a 24 hour period in this case. You do not have to read each topic one by one, of course, because we do have the outline over here. And so I think that this is a really, really helpful way of at a glance, knowing what is contained in these medical records. So I think it's fairly logical that you're going to receive documents that you yourself probably will not have had a chance to review before the tool picks up on. Especially again, if you're using faxing or other integrations of medical record companies and you're getting the documents in, or maybe it's legal assistants, case managers, paralegals who are bringing those documents in. And so by the time you or someone else at the firm or organization looks, they've already been processed. So the outline is really helpful in helping you orient as to what you are going to find within the records. So for each case, there will be a unique outline created with topics and sub topics as well. To be abundantly clear, this outline is uniquely generated. So an outline might vary from case to case, just as the facts of treatment and the background and patient history and such change. This makes it really easy for you to navigate within these different topics. So I could come in for a deeper review of the initial assessment. And then if it makes sense to drill down further into some of the neurological findings, I'd be able to do so. So this is a really easy way to navigate the overall holistic view of the documents. But what if you want to check the actual document itself? I know that a lot of people are asking questions too about, well, what about medical imaging and how do I know that it's grabbed the right information? Did it duplicate this information? So check the source and it's only one click away. So you are going to be able to come in here on any of these chips and click the number where the number represents the page in the source materials over here. And what's more is in many cases, the tool will also highlight to the page where you should go to review this information and verify its accuracy. So here we can see that there had been, it looks like three motor vehicle collisions within the same 24 hour period of perhaps some of the original or principal crash we'll say. Well, we could come in here and get more details and maybe I'll take this in and zoom in just a little bit more so people can see. So this highlighting is part of the MedCron by Filevine technology. Again, to be clear, this record was not highlighted originally and now we're just showing you. Instead, the tool is directing me and really my eye as to where to go to confirm the accuracy of this essentially subtopic of the record. And so we could see, it was her third motor vehicle crash within 24 hours. And so maybe you didn't even know that coming into the records, right? Maybe the only information you had was about one of the principal crashes. Well, it's very easy then for you to come in here and understand the order of events of the three crashes that are all essentially going to be interconnected here for some of the medical treatment that we're finding. So it's only ever a click away to get at the source document, but I promised to show you some of these other lenses as well. And you'll notice that there's a lot of the same principles across the four tabs. But when I come into the patient profile, the outline is a little bit different. So these are different topics than what we were seeing just a moment ago. Instead, this outline is going to reveal information about the demographics and patient history, past medical treatment, et cetera. And so again, I think this could be really helpful in uncovering some of that unseen evidence, as Ryan called it, where it is that perhaps you didn't know something in the social history that does make an impact on the case. It's certainly been the case that I've worked with some firms in helping to evaluate these tools, and they might give us records to try the AI against, and they'll say, here are the facts, and no, this is not true of the patient or the plaintiff. And then it doesn't take very long sometimes for us to say, oh, well, we see this in the history. Did you know about the substance abuse disorder here? And they understandably are concerned maybe that they didn't know that information going in. So it's that ability for us to detect the unseen, those things that you couldn't have forecast that I think does add a lot of power to this tool. And then your own ability to come in here and just verify the information. So in this regard, the case overview and the patient profile I think do really go hand in hand at providing background information as well as a holistic view of the case. But as I well know, many of you are looking to create a date-by-date summary of the exact treatments and history so that perhaps you could better identify things like gaps in treatment or over-treatment, and that's where this chronology is really going to be helpful. So let's dive in. The chronology and the summary tabs, they do absolutely go hand in hand, and so you'll see that here as I demonstrate a little bit further. Inside of this chronology, we have what many of you I think would expect, of course, which is a date here with summaries of treatment within. And what's really nice is that we have the ability to identify the facility as well as the providers of each facility. And so I think then that this could be incredibly helpful if you're looking to, let's, maybe you're reviewing disclosures or putting disclosures together, or as Ryan indicated, perhaps you're preparing for a deposition. And if one of these providers is someone who is, you know, testifying about treatment, let's say, it only makes sense then that I would want the ability to drill down into that provider's actual treatment and filter out the rest, filter out the noise. And so we do have the ability to interact quite dynamically with this chronology through the use of these filters over on the right-hand side. But we still maintain the ability to click on the source document and see that in a side panel, but now we have more dynamic filtering. So if it is the case that I want to just review a particular facility here, then I could identify that facility, make a selection, or also narrow this down by individual provider. And so it could be the case that I just wanna see this one, doctor's treatment, and now it's hard for me to prove this to you, but I'm trying to scroll. There's not really more to scroll on because this particular provider was only treating for these two dates. These are kind of known filters as I've been thinking of them. It's really easy to think about a chronology in terms of the date range, and then perhaps who provided the treatment. But I was actually speaking with a medical provider about this tool just the other day. And for that medical provider, what he told me is he actually doesn't really care when he's reviewing records about which provider said what. He's much more interested in the findings. And so that's where this bottom topics portion of the filter panel is going to be really helpful. What you'll find there is that in all of these topics, there are actual, again, kind of subtopics identified under each. It's fairly standard that every one of your chronologies will have these topics identified. But of course, when you expand further to see the injuries themselves, those are going to be dependent on what is found in the records themselves. And so this too would allow you to really drill down into a particular injury, a particular issue, or a particular assessment or finding. And so there's different diagnostic tests and assessments that have been conducted on this patient. So we could come down here and see what have various CT scans revealed. Clicking on these filters dynamically updates the results here, making it easy for me once more to identify some of the source material itself. You can view in context. So sometimes it's the case that you might want to see what was said right before and right after. Especially if things do get split up between pages, as is often the case. And so we're giving you that ability to see the source document alongside the AI summaries, alongside your ability to also filter down and find what's most relevant. Something that has, I think, intrigued a few folks was this pain scale when I've shown this on some records. Because I think that there's usually a question of things that have been maybe over-treatments, malingering, anything of the sort. And so the pain scale is going to help me identify any time pain was discussed, whether it was being improved, if it was not improving, if it was just staying the same. And so I imagine that this is information that folks would want to know. So if we can see that there'd been an indicator that pain had improved, I really want to know in what context here. And so you do have the ability to, once more, look at the source document and see that information. You're also going to be able to, again, evaluate this over time. Is the pain ongoing? Is it controlled? Has it been moderate? And these different filters, once more, update what you see over on the left-hand side. I personally, myself, do not have a medical background. I do have a little bit of a legal background. I was a paralegal working at a personal injury firm. And so I sometimes am not always familiar with the different assessments and codes and things, but what I really appreciated here is the ability to search. So again, someone was telling me, and I might still get it wrong, but I think it's the Glasgow Coma Scale. Someone, I hope, in the chat will know what I'm talking about here. And I didn't know if I should be looking for that under a diagnostic code or something else. And so what's really nice is, again, that I don't actually have to be that medical expert myself. Instead, I could come in here and start either looking for the abbreviation or perhaps spelling it out. And we could see that there have been references to findings from this GCS assessment, right? And so it's just helpful, again, if maybe you are a paralegal who's been tasked with doing this work. Maybe like me, you don't have a medical background. You're not entirely sure what you are looking for in the records. You're not sure what's important. This tool could nonetheless help you find that information. You know, it's not Grey's Anatomy and going to necessarily explain what a GCS is or does or why you might use it. But nonetheless, I think it could be helpful to over time also learn what's consistent, right? What assessments are applied in cases like this? Is anything that's contained here then relevant for determining if this plaintiff has had a traumatic brain injury, right? And even though, again, I myself am not a medical expert, nonetheless, here we could come in and drill down into some items that might be things that we do wanna drill down into further. And so in this case, there had also been some reports of things like I think loss of consciousness, kind of earlier too, there was a comment about like a poor historian. But again, everything is connected. There's context to everything. And so they're saying that the plaintiff is a poor historian of the facts here. But again, I think that in some of the records, we could also see that she was having, well, arguably, this is a non-technical finding, a very bad day, three car crashes within 24 hours. And so that understandably traumatic and might have presented an opportunity for like concussions or injuries to the head. And so we might now be on high alert for things like those loss of consciousness or other indicators of traumatic brain injuries. And so then if you can connect those dots, when you come into the chronology, you would be able to use those filters and assessments to drill down into what the records say about those topics. I have shown this to a few different firms. I've kind of given some a sneak preview. And it seems like for some, they're like, this is great, but maybe I want a little bit more. And perhaps that's because I'm looking for a more faithful representation of the data. And so I'll be really clear. If all you were looking for is a copy and paste of the medical records, this isn't the tool for you. But then again, I don't think that's why you'd be using AI on medical records to begin with. So what I will say though, if you are interested in more information, more details, know that under the summary tab, we do have the ability to click into that additional layer. And so I found that this usually satisfies some of the concerns here, again, that you might overlook something if you're using an AI tool, right? Does this mean then that you're gonna forget to review something, that it won't pick up something that's important? I don't think that's necessarily the case, but I do think that this helps that legal professional fulfill a duty here to make sure that they are familiar with the contents of the records, that they're verifying output, and just checking on the work of their staff too. And so you do have the ability to come into this detailed view, still have the ability to view the source document itself.

[00:37:58] Speaker 1: I've seen a few questions here that have come up around maybe what is it that I do with this data now?

[00:38:05] Speaker 4: And so I wanna talk about some of those export options. So of course, while you're here within the tool, you can interact with it as I've been doing, and apply filters, and navigate between the different panels. But maybe what you're looking for is something that you can use to help you understand what's going on in the data that you're looking at. But maybe what you're looking for is something that you could easily print off, whether that's because you're going to mark it up yourself, or maybe because you're going to attach it to a demand letter, or send it off to opposing counsel, or something of the sort. And if that's the case, then I think you'll really appreciate all of these export options that we've been mentioning. So on the surface, you have two options, really. You have the ability to take this out into more of a traditional PDF, or a Word document. If you select the PDF option, you also have the ability to export out the source medical records themselves. And so you can end up with a very, very nice packet of a dynamic kind of hyperlinked artifact or output, stealing Ryan's word there, kind of the artifact of the lawyer's work. And so that is an option available to you. Although of course, maybe you already have thousands of pages, and you don't want those included in the export, you could choose to exclude them. And I'll talk about both scenarios, what happens. But if you exclude the medical records, you could also take this out into a Word document. And so this doc X is going to be the format that would open up if you were using your regular Microsoft Word. And so that's going to be a really great option for you if you are looking to configure, or rather edit the outputs that you see here to add your own notes, your own observations. So let me switch gears and just show you what the PDF export with medical records looks like. So we have here our MedCron by Filevine, and I can zoom in a little bit more here too to make it easier. So this is a PDF document with hyperlinks built in. And so I could jump right into the chronology, and it's more of the same, right? It's the same information that we were just seeing within the web browser brought into the PDF with actual live hyperlinks as well. And so, like I mentioned, when it comes to the PDF, you have the ability to export with the medical records. And if you have those included in the export, when you click on the page, it's just going to take you right into the actual source material itself. While there are breadcrumbs up at the top, Jonas will have to tell me if that's a technical term or not, right? Breadcrumbs, but I have the ability here to just navigate back to the top of the chronology, or maybe I just want to go back to the top of the document and look at some of the patient background. If instead I had exported out via, I'll say first, if I had exported with the PDF without the medical records attached, then for me, as someone who does have a login to the MedCron by Filevine, when I clicked those links, it would open up back in my web browser. And so again, that's a particularly helpful feature, I think, if you don't want to sit and wait for Adobe to open up like a thousands long page document and instead want to use the hyperlinks. But again, you do have the option to export this out into a medical record. And so that's going to be your best option for editing the output, because I think it's very easy to imagine you might have had more information to share. And so our tool is only looking at the medical records themselves, but perhaps you've had a phone call or a deposition or something of the sort, and you've had a chance to speak with the patient, you might have more information to add. And so you could add those notes, observations, add, delete, whatever might be most relevant for you.

[00:42:04] Speaker 2: Madison, you were probably going to cover this, but I can't help myself because I think this is so neat. If you were to send this PDF to an expert or opposing counsel, or even an adjuster, the hyperlinks would all work. So you could click them and it would take you right to the source document in the medical record. So if you were to click one of those, it'll actually take you to the medical record, which is really nice.

[00:42:28] Speaker 4: Yes, exactly. And so I think that that's something else that just kind of administratively saves time. I know that for a lot of others, if you don't have a tool like this in place, what you would be doing instead is having to go through your folders, find the documents and combine them. And so this is doing that work for you. And again, even if you had combined them, what you then might do is make a bookmark in Adobe or otherwise create some sort of share link back to the records themselves. And so this solves all of that. If you export with the records, you're just going to build in those hyperlinks automatically. Great point, Ryan. And so if we come into some of the bookmarks here, you are able to see that this bookmark was created automatically by the system. So I did not, I promise, come in here ahead of time and build out a bookmark. Why would I? The tool can do it for me. But you do have the ability to still drill down into this, I suppose, to see the original records or navigate between some of the dates in the summary. There's always more to cover. And this is just one part of the otherwise AI platform that we offer at Filevine. And so this has been a look at the MedCron by Filevine, considering medical treatment in particular. I want to leave time for more live Q&A. I want to leave time for Kendall to talk about how you can get this enabled in your environment. And so with that, I think I will pass it back to you, Kendall, for some of those next steps.

[00:44:09] Speaker 1: Awesome. Thank you so much. Seeing a lot of really great questions come into the Q&A right now. We are going to get to that in just a moment here. Let me just address one of the most common ones I'm seeing so far, which is how do I get this enabled for my environment? How do I get this for my organization? The good news is the actual setup and implementation process of this is very fast. So in just a few slides here, you guys are going to see a QR code that you can scan and then register and get connected with our team to get this enabled for your organization. I think we're also going to drop that into the chat if you're having any trouble scanning it, and then it'll be available one last time before we wrap today. So quite a few opportunities to make sure that you get connected with our team there. As soon as you have been enabled and it is on for your environment, you're going to receive an email with some step-by-step information, how to get this turned on and start using it. The really, really great news, again, it is pretty plug and play. So a quick look is if you're familiar with this screen, you're probably an admin of your firm, but within Customs Editor, there will be a new section available as soon as the feature flag is turned on for you, and you will add on this MedCron by file line section into the project where you'd like to then go ahead and use it, and then you will then see it on the front end. One thing to kind of keep in note, here on my screenshot, you can see it says Run MedCron. That means, hey, go look at this project, go start running all of our previous medical documents to get them indexed and ready to go, and then as soon it is ready to actually be used in that kind of demonstration that Madison showed us, that will change to Launched MedCron. In that email that you receive when it is enabled, it will include all of these instructions as well as support articles for use going forward. Here's that QR code, and I wanna get everybody just a moment to go ahead and scan that, and then we're gonna jump into the Q&A portion of our presentation here. All right, so a few of these questions I've seen come through. I'm just gonna go ahead and read off, and then I just invite my panelists to go ahead and answer live with me. So the first one is, will MedCron be able to identify duplicate records that may have already been uploaded and might have a different experience, excuse me? Maybe Jonas or Madison, maybe if you guys wanna jump in and take that one.

[00:46:48] Speaker 3: Yeah, I can take this one. So the duplication currently happens on level of one document. So if you have document that contains the duplicates, we would deduplicate it for you, and you would see one chronology item for the duplicate. If the duplicate happens in separate documents, currently we don't support this functionality of the duplication, but that's something that's on our radar and we plan to add it soon.

[00:47:19] Speaker 1: Right, and then this one might be kind of an add-on to that question as well. Does the Trinity extraction feature prevent duplicate entries of the same documents, visits, notes, or images to be entered into the MedCron?

[00:47:38] Speaker 3: It shouldn't prevent it from adding it. The Trinity is extracted for every document individually. So the duplicates would end up with the same Trinity, and we would basically merge them to one chronology item that way.

[00:47:58] Speaker 1: Thank you, great. Yeah, there's a lot of questions in here just diving a little bit deeper into kind of these duplicates or these multiple versions. So a similar question as well is if multiple versions of the same medical record give or take extra pages that are uploaded into the Filevine project, will MedCron AI recognize that, or would there ever be duplicate entries in that situation?

[00:48:24] Speaker 3: So this would be the situation I mentioned when you have two documents and between those there are duplicates. Currently, we don't support the duplication functionality across the documents, but again, this is something we are aware of and will integrate shortly.

[00:48:41] Speaker 1: Thank you. Other questions just a little bit more about searching. So are there already tools today that make MedCron searchable with filters, or how might this change?

[00:48:54] Speaker 4: Perhaps I could add just a little bit. I don't think I really covered it. There is the ability to do a keyword search that I didn't quite cover in addition to searching the topics, which I did. So you do have the ability to perform a keyword search looking at the contents of the document or rather some of the output, but keep in mind too that this is still situated within your Filevine. And so if you are looking to perform other searches across these documents and more, your regular OCR keyword search is still available. You have the chat with your case built into the sidebar tool where you could also ask questions. And so the search exists within MedCron, but again, you still have all of the same searching available within your Filevine core platform too.

[00:49:41] Speaker 3: Maybe to add to Madison, I will mention this in the future roadmap session, but we plan to integrate the chat interface directly inside the tool itself. So then you can also perform your question answering, which is a form of searching and retrieving the information you need.

[00:50:05] Speaker 1: Great, thank you. Next question on here, can the AI prioritize or flag items of significance? I think-

[00:50:15] Speaker 2: I can maybe take that question and then Madison and Fiona's too. Probably not the current MedCron you just saw might not flag like problematic areas, maybe gaps in treatment or maybe issues with your case. However, chat with your case, which all of you have if you're AI, Project AI customers already, can actually focus in on that specific document. This is current functionality. It's not a new document. This is current functionality. It's not on the roadmap. I mean, it's not on the roadmap because it's already built. You can simply upload the parent MedCron to chat with your case and ask it that very question. So you could say, what are the red flags in this case based on this MedCron? What are gaps in treatment that you perceive based on this MedCron? And chat with your case should be able to handle all that. There will be continuing functionality and UI improvements, but you could ask for red flags right now and chat with your case.

[00:51:12] Speaker 4: Yeah, I may just kind of add my two cents to that. How I've thought about the MedCron by Filevine right now is that it's something that can happen really just automated. So I'm not interacting. I don't have to tell it what I think that the key issues are. Excuse me. Because the idea is I might not yet know what they are. I just want this on an ongoing basis, providing insights into the treatment. For those times that you do know what might be an issue in this case, or you want some help consulting on what those might be, that Filevine AI Assistant is going to be a really great resource for you. The ability to go through documents with a fine tooth comb via our AI Doc Review, those are also great options. So we've shown just one tool within the kind of tool belt that really is available.

[00:52:02] Speaker 2: We're getting quite a few questions about folks who already have our sort of, call it V1 of our MedCron feature. That actually does some things differently, and it's probably still quite useful for you in different ways. So you keep that. This one doesn't displace it. It just adds to it. The nice thing is if you've already got a section called MedCron, you can actually make the section any name you want. For those of you familiar with the Customs Editor, you already know that. But this is a totally different MedCron. So it's a different tool, and will sit within your sections.

[00:52:37] Speaker 1: Another good question on here is, is there a way to limit what is shared or available to outside counsel with a downloaded document? So for example, there might be unrelated or sensitive information that they might not want to share.

[00:52:51] Speaker 2: Let me take a crack at answering this, Madison and Jonas. And you tell me, I may not have it right. I think I'm pretty familiar with the product at this point, but I may not have it right. I believe you can save views, particularize views. Particularize views. So you could exclude a certain provider, maybe even a treatment range, Jonas. So yeah, you'd exclude providers. You could exclude, for instance, prior treatment, prior to the accident. You could probably even exclude based on maybe some diagnostics. I'm not 100% sure. Jonas is nodding his head. So you can essentially save a view of the MedCron for that particular client, and then you can download that view. I believe that's all correct. Is that right, Jonas?

[00:53:33] Speaker 3: Yes, I can confirm this. Basically using these filters, Madison was demonstrating that you can kind of distill down just the view you're interested in that might be through your date filters or these topic filters to really drill it down only to the essentials you then want to share with your opposing counsel through this PDF export or the Word document export.

[00:54:00] Speaker 2: Pretty cool, Jonas.

[00:54:03] Speaker 1: I think another question coming up is, once we have this product turned on in our environment, if MedCron is automatically updating the chronologies when medical records come in, does that interrupt any additional items going on within Filevine? Is that something that just happens as the document is being uploaded in OCR'd or how does that process work?

[00:54:27] Speaker 3: Yes, so basically everything should happen automatically after you upload. We run this processing, the document classification that was mentioned. We run it through MedCron on a background. The MedCron just automatically gets updated with the new information.

[00:54:48] Speaker 1: Perfect, thank you. How does the AI handle handwritten records or notes?

[00:54:56] Speaker 3: So this is a good question. As I guess, even for humans, sometimes the handwritten decoding can be tricky. So we do our best effort in decoding those, but we are aware that this is something that the AI can struggle. And this is where we always would recommend double checking, verifying against the source. And we do have this editing capability that if the date, for example, is incorrect, you can always fix it yourself.

[00:55:31] Speaker 4: What I would just add kind of anecdotally, so I've been processing different record sets that do have handwriting. And it's, I think, kind of the joke, but also maybe a little bit true. Handwriting is difficult. And then of course, we can make jokes about doctor handwriting too. I've seen in two different cases, some fairly difficult handwritten notes getting picked up by the tool. In the first, it was dental records. And I think that I see those often are more handwritten or they sometimes are more prone to handwriting than I've seen in other documents. And so there had been handwriting about, and again, forgive me, I don't remember the exact example here, but it had been about three to four millimeters of some sort of probe that they'd done in the gums. And so it was on lined paper, the handwriting, and it was a bit of a mix of like cursive and print. And the tool picked up on that. And so I was quite impressed with that. But anecdotally, I'll share, Jonas was right. There's sometimes instances where it's not perfect. And so I myself read what had been a date as happening in the year 28. It was abbreviated, right? And so I really thought, I was like, this doesn't make sense. It's not 2028 yet, as far as I'm aware. So I'm not sure. But in fact, the person had written a 22, but kind of looped it in an interesting way. And so I think that it's certainly the case that you're going to want to double check and just verify some of the handwriting. That's where it was really helpful for the group that I was showing this to. They were like, how does it handle the handwriting? And I said, where in the document is their handwriting so that I can drill down into it? And they said, it'll be in the dentist records. And so we drilled down to that provider, pulled me into the document, and we did a side-by-side. So that's also, again, the vision of the tool is that you have the source material right there so you can corroborate what the AI has pulled.

[00:57:26] Speaker 1: Thanks, Madison. I think another common question that I'm seeing is around Bates numbers. Do you think you could dive into that for us as well?

[00:57:35] Speaker 4: Yes. So one of the items that I failed to neglect in my live demonstration is that the PDF export already does the Bates stamping for you. And so I should have shown that, my apologies. Sign up for a demo and I'll show it to you then. But at the bottom of the page, it will do the Bates stamping for you as well. So you are able to, someone can correct me if I'm wrong, but when we are talking to that page, let's say it was page 94, it'll take you to Bates stamped document page number 94. It might've been page 103 in what's, well, maybe that doesn't make sense. Maybe page 42 in what had been in the original document set, but in our export, it's the page 94.

[00:58:21] Speaker 1: Perfect. Thank you. And then Jonas, I think this question may be a little better for you exactly. What is the accuracy of the AI and does it require any final human review?

[00:58:33] Speaker 3: Oh, it is always difficult to say like overall accuracy. There are multiple steps as I demonstrated throughout the pipeline. So you would evaluate each of those separately. We always, of course, recommend the human review. It is AI, it can make mistakes. That's why we focus so much on this verifiability step and trying to make it as easy for you to verify against the source materials. But based on our evaluations by our human annotation team that we have in-house, we are reaching very high accuracies. It might be around 95% for these trinities. So the date extractions and on those 5%, yeah, it's those maybe the handwritten dates or these very hard to read information on the maybe scanned documents.

[00:59:34] Speaker 2: I'll take a crack at it. Thanks, Jonas. You know, I think Madison could say in the samples that we've done for customers who demoed this product, the accuracies come out very high. We actually had a customer put us to the test just yesterday. Again, somebody who got a sneak peek of this and they actually gave us 10 demands. Some had been written internally by their own internal lawyers. And I think about half of them, maybe a little bit more had actually been written by a competitive product. I'm sure most of you would have known the name of the competitor. And that competitor has some AI technology, but also human review. And we compared our MedCron literally line by line, document by documents. We did this with two different people at FileVine, both of them have extensive experience in legal, one of whom is a lawyer himself, the other one a paralegal. And we found no substantive errors across 10 different demands for FileVine's MedCron. So we were super excited about it. Not a single provider was missed across 10 different demands. In a few cases, very few, there was transposition of dates. So maybe November 20th didn't come out as maybe it was the second of, maybe it was January 11th or excuse me, February 11th. A few cases like that, but it was very few in number. I think we're talking less than 10 across probably thousands of dates of service. And again, in any one of those cases, you can go and click the document and find out the correct date of service. So 100% no, but extraordinarily accurate. And by the way, when we did that compare, the only substantive mistake was not on our side, but our competitor had made a fairly significant substantive mistake where they attributed treatment from one provider that had actually been given by a totally different provider.

[01:01:31] Speaker 1: Thank you, great. I think another question on here is maybe if we disagree with what the output actually came out as, or if I wanna make an edit to what is actually produced, is that possible?

[01:01:44] Speaker 3: Yeah, so basically currently as Madison explained, the way of editing is through this workbook export. As again, I will mention in the future roadmap plans, we want to build the editing functionality directly inside the UI so that you can do all your edits and changes to the summary or to the dates or whatever you would like directly inside the tool itself, then export already this edited version.

[01:02:19] Speaker 1: One question on here as well is, would we do a separate more in-depth training? I would be happy to answer this one, we absolutely will. We plan to do additional webinars as well as as customers are getting enabled on this, we'll have support resources and we'd be happy to connect with you and walk you through that as well. Let's see, how about MedCron and medical bills? Do they get organized chronologically with the treatment records?

[01:02:49] Speaker 4: Good question, so my understanding right is that this MedCron is focused on the medical treatment. So we are using the document classification to identify the medical records, not necessarily the medical bills. And so that's what's being ordered here in the tool for us to produce the chronology. It's some of these other AI tools inside of Filevine that could be used today to extract out charges from medical bills and allow you to query that information more deeply.

[01:03:22] Speaker 1: Thank you, great. I'm seeing questions kind of start to slow down here. Please continue to send them in if you have any additionals. Let's see, will it flag certain records that it has difficulty reading and alert us to potential issues?

[01:03:38] Speaker 3: So this is a very good question and something we care deeply about is basically the confidences of the AI outputs. And this is something we are bringing or will be bringing to the UI directly to basically highlight these low confidence regions as we call them, that exactly the AI might be uncertain so that you can then proactively go and verify maybe those or edit those against the source materials. This is not in the current version that we're launching, but it's something that we are very aware of and working towards.

[01:04:20] Speaker 1: Perfect, thank you. All right, I think we are getting close to being done with this Q&A portion. We would love to jump into a little bit more of our future state and what we would like to look at moving forward. So Jonas, I'm gonna actually hand it back to you to go ahead and cover this portion for us.

[01:04:45] Speaker 3: Yeah, so I just wanted to briefly tell you more about what we plan, what's ahead for the Matrons. One of the first things we want to add is this editing functionality directly in the UI. As I mentioned, this goes hand with hand with these confidence regions. So we want to be able to show you where maybe our AI is not confident and you can give you the tooling to resolve these low confidence regions because we are aware that AI is not 100%, probably will never be. We want to give you the tooling and the locations of the places where you can fix those mistakes. As I already also mentioned, we plan to add the chat interface, which should be very similar in nature to what you're used to, for example, with the chat with your case. But we want to do it in this kind of signature way of this side-by-side view. So as you're receiving your answer from the chatbot, you can verify the answer. The hyperlinks will take you to the source documents and show you the bounding boxes with the highlights so that you can, again, very easily verify these answers. We also plan to add a visual timelines component. So this would enable you to compare, for example, the treatments and expenditures. You can see them on the timeline. You can set your date that you can then use as a reference point and see basically how these categories resolve around this reference date. We will then enable you to zoom into the particle data points. So, for example, there is a treatment. So we will see the treatment, so you can zoom into particle treatment to get more information. We also want to evolve the MedCron basically around your entire case. Now, it starts with the medical records, but we will start adding specialized processing for documents such as bills and gradually expand to other case-critical documents like police reports, depositions, expert testimonies, and others. And this comprehensive approach will enrich the case timeline and really help you tell the complete story of every case you're working on. We also will be adding this section that we call Insights. This is like your analytical command center that will automatically surface things such as missing data or contradicting information between the input documents or some abnormal values or patterns that we found. Plus, you can export the key findings from your chat interface to have it on one place. So you will no longer have to do manual searching for critical inconsistencies or gaps yourself. So these insights will then become your strategic advantage and will create this natural bridge between the case analysis into the deposition preparation. You will be able to export these findings into your deposition copilot, enabling you to construct data-driven deposition goals based on these concrete evidence, identified case gaps, and suspicious patterns. All this in combination with the FileWire New Deposition Hosting offering will create very powerful end-to-end experience for you. So we believe that the future roadmap of MedCron isn't just about reading the medical data, but it's about transforming how you build winning cases.

[01:08:50] Speaker 1: Awesome, thank you so much. We definitely have a lot that we are, you know, continuing to work on all the time. It's kind of our motto here where the work is really never done. Just to wrap us up, just, Ryan Anderson, I don't know if there's any kind of final words you want to leave us with.

[01:09:04] Speaker 2: I'll be super brief. Thank you all for hanging with us for an hour. This has been one of the most successful webinars in the history of the company. For those of you who are Project AI customers, I hope you're starting to get the pattern. You've seen chat with your case come out. Now you're seeing MedCron come out. There will be other features, some really big ones that we'll announce at Lex that will also be part of Project AI. So becoming part of Project AI is a very good investment. If new products come out and they are AI products, our goal every single time is to make it part of the Project AI suite. I can't promise that every single product forever in the history of Filevine will always be, will never be any extra money, but that is our goal. Our goal is to every single time we can, and we think we can in almost every case, that will become part of the Project AI offering. And that offering that you paid for will become more and more valuable over time. So we are so thankful to Jonas and his amazing team for building this fantastic product. We can't wait for the future. Filevine is heavily invested in AI and in this MedCron technology in particular, but of course, depositions, which Jonas has also played a major role on. Thank you, Madison. Thank you, Kendall. Thank you, Alex McLaughlin. And we did not get a chance to thank the Dock Classification team, whose work was unheralded a bit in this demonstration, but it was critical for making a MedCron that happens automatically, that you can just have work when you need it. We're thrilled for all the many people at Filevine who've made this come to pass. Thank you all. And we thank you most, our customers, for being loyal and giving us the opportunity to serve you.

[01:10:44] Speaker 1: Thank you.