Video: Discover your Data Faster with Everlaw AI Assistant | Duration: 5368s | Summary: Discover your Data Faster with Everlaw AI Assistant

Transcript for "Discover your Data Faster with Everlaw AI Assistant": Alright. Hello, and thank you for joining us today for our webinar, Discover Your Data Faster with Everlaw AI Assistant. My name is Corinne Hodges, and I lead product marketing here at Everlaw. And I'll be your host for today's webinar. I'm joined today by Jason Cassidy, solutions architect, who will show you some practical ways to use AI assistant in review, in evidence analysis, and in building narratives. In just a few minutes, I will set some context around generative AI, our our philosophy around building tools using this technology, share what we learned from our beta, and give you an overview of Everlaw AI assistant suite of tools. Right after that, I'm gonna hand it over to Jason who's gonna show you those tools in action. But before we get there, I wanted to just, share a couple of housekeeping items to introduce you to our webinar platform. So first note, we do wanna confirm that today's session will be recorded. So to rewatch today's session, all you need to do is revisit the event by clicking on the same event link that you used to join us live today. And then in a couple of hours after we end the live feed, the recording will be available here to watch on demand. So now let me go over a couple of features and resources in our webinar dashboard. On the chat features that you see here, on the right, the session chat, that's where you can say hi, have fun chatting with other members of the audience. While you're there, don't be afraid to start off by telling us where you're joining from today. And then if you have any questions, you can please submit them to the q and a tab on the right here. Our team will do our best to answer all of the questions that are submitted before the end of today's webinar. Note that the session chat is not monitored throughout the question throughout, for questions. So if you wanna ask a question, please make sure to post it in that q and a space on that q and a tab. Next is the docs tab where we have several resources and links to some of the related content that, we'll cover today in the webinar. So that's all it is around the sort of logistics around the platform. I know you're all excited to see the product in action. So with that, let's just go ahead and get started. So I wanted to just to set a little bit of context and start with one of the biggest pain points, our customers in this industry are facing, which is the growing data size. So legal professionals need to make sense of 100, thousands, even millions of documents before a deadline to start building cases, and the challenge of around this is only going to continue to rise. Based on IDC forecasts that you see here, discoverable data is growing at an incredible speed with an estimated 181 zettabytes of data being created by 2025. So connecting and processing that amount of data is so difficult, especially when legal teams are expected to manage this data with little growth or in some cases reduced human and financial resources. Generative AI technology is a potential solution because one of its strengths is the ability to synthesize data at incredible speeds. But one of the questions is whether legal teams are ready for this kind of innovation. Is that time now? Well, we think yes. In a recent survey we conducted with ASEDS, we saw that generative AI adoption is happening roughly 5 times faster than the Cloud. Just a year and a half after the launch of ChatGPT, more than a third of respondents are currently using generative AI in their practice. And that breakdown is 17% are actively using it in production on Live Matters, and 16% are using it in some form of beta testing. This rapid uptake contrast sharply with adoption of cloud based ediscovery software, which took nearly a decade to reach a similar level of use. So why do we see adoption in generative AI increasing faster than adoption of cloud based ediscovery software? Well, what we're learning is that legal teams see the potential of getting a leg up as a result of using these tools. Earlier this month, we released another study with the Association of Corporate Counsel. And there, nearly half of respondents say generative AI will significantly increase legal teams' effectiveness. Another 11 percent anticipate a fundamental shift in the nature of their work. All of this is to say that generative AI is gonna change litigation. It's not going to replace litigators, but it's going to significantly alter the way they engage with data, the way they collaborate with each other, and approach their work. There will be a massive competitive advantage to being able to find better needles in the haystack faster and enabling case leads to tell more powerful stories leading to better outcomes for clients. So that's just a little bit of context. Now let's turn over to see how Everlaw has been building generative AI into our solution. So first with Everlaw's philosophy, we build for the long term. Our approach to software develop development has always been to prioritize taking time to being thoughtful, asking the right questions in order to create the highest quality tools for litigation and investigation management. So our goal with Everlaw AI assistant is to create a set of features that drive real practical value for our customers. And that is guided by our AI principles, which is to ensure that these tools give the privacy and security and control, which are table stakes, and the confidence that our customers come to expect from Everlaw. To instill confidence, we designed our experiences focused on generative AI strengths and the four corners of your documents. We embedded it into Everlaw's workflows with precise use cases and context at hand and made outputs easily verifiable. We also deliberately planned for a year long beta, which saw participants across a 174 different organizations and 455 individual users so that we could better understand from our customers how they use the product, how we could improve it overall. And over the course of the last year, the feedback we've received had been integral to its development. We iterated on each feature many times, and we really refined the user experience to provide better value and avoid challenges and pitfalls. So for instance, here's our AI writing assistant. Initially, it was very open ended, the results were inconsistent, and our users struggled to learn how to use it. Now it's more structured, it has preset templates and output formats, and question types. Users can immediately get value from it. So this is just an example of one of many ways where we've evolved during the beta, and we've made similar improvements to our review and batch tools to just simplify them for our users and improve their usefulness. So Everlaw AI assistant. It isn't just one generative AI feature, but it is a suite of tools that you see here to help from review to case building. On the review side, you can find important evidence that may never have been uncovered or take many more days and dollars to find. And then on the case building side, you get a leg up in telling more powerful stories in crafting arguments and other legal documents. With review assistant, it is a set of tools that support faster and more comprehensive document review with the help of generative AI. It includes the ability to summarize documents, extract topics, and that's to pull out key concepts or topics from the documents, and then embedded in those topics are entity extractions and sentiment analysis too. You can perform both of these tasks with a single document or even in batch. And then you can even ask your own questions on the document to get answers back in real time. With coding suggestions, this helps you to find and prioritize relevant documents faster with the help of generative AI. So using a set of user provided instructions, the AI suggests whether codes should be applied to documents or not. Explanations for its suggestions is available and also a link to relevant snippets in the document. Like summary and document, like summarization and topic extractions, coding suggestions can also be performed at a single doc level and across the entire document set with just a few simple clicks, and Jason will show you in just a minute here. And then finally, writing assistant. This is embedded on our case narrative workflow story builder. So with writing assistant, you can create, evidence centered research by synthesizing and organizing organizing insights across documents, and you can also explore different ways to frame evidence or argue based on evidence by rewriting a section of legal writing and selected pieces of evidence with your own set of instructions. You can also insert AI generated outputs into draft documents as you see fit so you have more control and visibility on AI generated and human written content. So with that, I'm gonna hand the floor over to Jason so that you can get an up close look at practical applications of Everlaw AI assistant. Alright. Thanks so much, Corinne. I'm gonna go ahead and share my screen here, and we'll get started. So alright. Getting that kicked off. We are in a dataset that I'm sure many of you are very familiar with. This is the Enron dataset. And so you are likely familiar with a lot of the context of this matter from kind of the early 2000, late nineties. And I'm gonna walk you through how some of this generative AI based technology can really help you speed through the review of these documents as well as sort of build your list of deposition questions and collaborative documents throughout. So the screen that we're currently on, I will always point this as a view. And so I have, you know, my list of metadata that's important to me. You'll see some of these have this sparkle right next to it, and that is denoted as our AI features and functionality. So you're seeing some coding suggestions created. You're seeing quick descriptions of these documents as well as topic based breakdowns and sentiments along with this. This was all done in a batch action. And so without me having to go individually into every one of these documents and say, hey, I'd love to see what you think about what codes I should apply, you know, what is this document talking about, all I did was pop in here and go to generate coding suggestions, topics, extractions, etcetera. And this is where I got this breakdown. You can see we've got some sentiment analysis on this document, for example. So Sharon Watkins being, you know, the courage and vulnerability to speak out in the face of corporate scrutiny. Good. Very positive. That is what we'd expect to see there. And then on the same very document, unraveling Enron's Raptor and Condor accounting scandals. Bad. Sad. Sad sentiment in there. So pretty easy to search through and find those things. One of generative AI on the whole's big downfalls is that it loves to be positive to the point where you get things like hallucinations. You know, it's telling you yes when the real answer is no. And so I always like to call out, you'll see this throughout the platform. But as I scroll down kinda towards the bottom here, there are some documents that it will say no, you know, in in cases in this particular case for responsiveness, I don't believe that you should be applying that code to this document. And again, without me even clicking into these documents to see what they are, you can kinda get a a snippet based on the description and topic. These are fantasy football based documents. So without even having to click in, I'm I'm fairly confident that that is correct, but I can always click in and kind of, you know, ensure that that is actually the case as I go through. Quick aside here that I wanted to show, I'm doing all of this from a a particular binder that I clicked into. You can see I'm in my binder for generative AI. This is, you know, more or less the full Enron datasets. We're dealing with 1,600,000 documents. And if I were to run a much larger batch action looking for where should I apply certain codes, you know, give me quick descriptions of these documents, You can also search on these things. So if I just wanna see documents that might have a positive or negative sentiment, it's extremely easy to run a quick search like that. Let me pop over into my project search, scroll down very slightly, and you have a section for each of these things. If I click into the topics, yes, I could pick a particular topic that I might be looking for, but also, I could just grab any sentiment that I might be looking for. You know, any document that has positive sentiment, I wanna give that a look. Here's 437 of them, and you can preview what all of the positive or negative, you know, if I were looking for negative sentiment on each of those is. So you're not, you know, stuck to purely looking at the the list of documents as it is. You can always run searches on these generative AI based topics as well. So with all that said, I'm gonna go ahead and click into one of these documents so you can see it in action. Up until now, we've been doing batch actions. What I'm gonna do is click right here which pops open a new tab. This is a fairly large doc. If you didn't see it on the metadata, this is a 72 pager. It's a bit long, but it had some negative sentiment in it. And I would love to dig into this further and kinda see what's being discussed. But, you know, I'm short on time. I don't wanna read through 72 pages of a document, so I would much rather have it lead me in the right direction. If you're familiar with EverLaw, we have what we call our context panel on the left hand side over here, and that includes any document that is related to the one you're currently viewing. Viewing. So think near duplicates, think email thread, produced versions, that sort of a thing. You've also got that same, you know, sparkle based icon, which is your review assistant. I'm gonna start in the overview section just for the sake of the demo. But, again, we have that description that was generated on the view itself. So it's giving me a quick breakdown here. You know, this text reveals that Enron employees are shredding documents, submitted investigations. It implicates this particular individual, there's Kenley in here, all these names that you've likely heard before if you've been in eDiscovery for a little while. This is great, but it is still a 72 page document. And so I would love to see maybe more specifically where these things are being discussed. Couple of different ways to do that. I have show summary by section, and so we don't limit you to just here's, you know, a paragraph based on the 72 page document that could certainly leave some things out. Over here, you can see, you know, just in the first 11 pages, we're talking about shredding documents at the Houston offices and former executive Maureen Castana, Castaneda rather, stating that the shredding continued through last week, blah blah blah. Like, this is pretty, you know, right off the bat stuff that I might wanna take a look at. And if there was anything that jumped out at you, you can click on those pages which will jump you right to that section of the document. As I scroll slightly further, we've also got this topic based breakdown. If you are looking for a particular topic, maybe you're a bit later in the case and you know that you're looking for manipulation that's being discussed or some particular type of issue, you could search for that. But separately, this is where our generative AI tends to stand out as opposed to many, in the space in in all spaces sort of, throughout tech currently, is as it gives you these answers, you are not, you know, required to just take it at its word. If you were, it could do things like going to the Internet and get, you know, false information that it would give you based on these documents. At Everlaw, we require it to stick to the 4 corners of the document that it is viewing. And in order for you to kinda give the final sign off on that, it's always required to show the relevant area for what it's looking for. So if I go to one more, you know, further down here, we're discussing Ken Lay, the CEO of Enron. You know, he utilized a company provided line of credit, this whole big, you know, scandal. You can see we're popping out the entities as well as the individuals associated with this particular topic. And the big differentiator is that you can show the relevant area. You're going to highlight on that document where this is being discussed so that you can see with your own eyes as the individual making that final decision. You know, here it's mentioned. Throughout the year, Ken was using using a company provided line of credit. You don't have to take it at its word. It's not getting that from some, you know, random source on the Internet. It's not grabbing it from some guy on Reddit. It is within your actual evidence. So very nice, for if nothing else, you know, the the ease of mind that this is pulling directly from your document. Another thing you can do, I won't spend too much time on this one, but again, it's it's very hard for systems that utilize generative AI to give you a negative, to tell you no. That thing you're looking for is not here. So we really prompted it in the background quite aggressively in order to ensure that it is when it's talking about these things. So I asked 2 very basic questions of this document. Does it contain any mention of Enron? Should be a slam dunk. It absolutely mentions Enron all over the place. You know, I could run a search and it would probably pull up about a 100 times or so. And so as I asked that question, you know, yes, this document contains numerous mentions of Enron. It discusses various aspects of the scandal, blah blah blah. But when I throw my own name in the mix, hopefully, you know, I was 4 or 5, at the time of the the Enron scandal, and so hopefully, I'm not being mentioned. And it does not, you know, it does not provide sufficient text to answer that question. I am not, in this particular Enron document or likely any of them, and so good. It will tell you no. You can always give feedback on these as well. You know, does this seem accurate? Does it not Provide feedback to the team and we can make adjustments on this sort of a workflow. So it's a bit about the overview of the document. You know, understanding the document on its face without having to scroll through all 72 pages in this instance can be extremely helpful for saving you time on these docs. But to be, you know, completely frank and candid, one of the coolest things that comes along with this sort of technology is something like coding suggestions. As I pop in here, I'll I'll let you kinda look at it as I walk you through what's going on. But if anyone has used TAR, used tool assisted review in the past, as a solution architect, I demo it all the time and having to explain to someone that may not be the most technical that there are patterns and words that are, you know, basing itself on how you code documents. And as you do that 100 and thousands of times, it gets more accurate. It's very hard to to sell someone on that. And, yes, it is extremely accurate tech. But what would be much more easy is if everything was in plain English. I am going to tell you what a responsive document is in the case of Enron. I'll even give you information about the case itself. And what I want you to do in return is tell me if you think I should apply that code and why. And that and why is again kind of what sets Everlaw apart and what really cuts down on the hallucinations that tend to come with generative AI. So, yes, it's just telling me the answer here. But if I pop into view configuration, this is not generative AI. This is something that my broader team, maybe myself, maybe a team of partners, maybe some senior level associates has gone in and said, we are looking for responsiveness. This is what responsiveness is. This is what a responsive document is in the enron dataset. And what this helps with is I'm not gonna say it's perfect every single time right off the bat, but it is much, much more accurate than, you know, having no seed set and attempting to do a tar based workflow. You're telling it in plain English, this is what I'm looking for, these special group of entities. Every single time that something like that is mentioned or brought up, it is going to be a responsive document, pretty cut and dry. If it were to get it wrong, let's say this said no, but it is actually a responsive document, it's gonna tell you why it thought it was. And so right here in plain English, I can then go in and edit my description of responsiveness and make sure that it gets it right. So it makes QA and QC of these docs much more accurate too. But, you know, in this case, it did nail it. You know, I showed you a quick description of what it was looking for, those special purpose entities, and it takes it even one step further. Here's the names of the entities. Extremely helpful. Now I can run searches on these things and find very key documents using these couple of names that it pulled out. But again, at its core, it is not applying this code on your behalf. It's saying, I think you should. If you want to, you know, go ahead and do it. Here's why I think that. And then the most most important thing that a lot of these tools tend to miss is where does it say that? You know, I need this information in the document. It cannot be pulling this from some false source. And so here is, you know, one of the instances where that is mentioned directly here. Also, I'd just like to call out this, this page is or this entire document is formatted very poorly, so this makes the use of generative AI not much more useful as opposed to trying to read sort of these blocks of text on the stock. So as I'm kinda talking through this, Corinne, I know there were some examples of, you know, specific, you know, workflows going back and forth, at Everly. Did you wanna speak to a couple of those? Yeah. Yeah. So we did conduct an ex an experiment putting coding suggestions against first level human reviewers across 4 discovery datasets, and these were all part of an active, civil litigation matters involving a broad range of different formats, you know, emails, spreadsheets, PDFs, calendar files, etcetera. And in this test, based on interviews with customers and provided review protocols, Everlaw provided initial context for coding suggestions for each of those cases. And what we saw, is that coding suggestions performed, as well as initial human reviewer. And in one case in one, test, 36% better. So what this shows us is that these tests, so that legal professionals can reasonably be reasonably rely on coding suggestions to be able to help prioritize and classify documents and reduce the workload associated with some of these manual processes. However, coding suggestions performance is affected by a variety of different factors, particularly the kinds of codes being evaluated and the quality of the context that's provided, by the team. But based on our testing here, we were able to see some, really positive results actually. Awesome. Great to hear that. Good to hear real world examples kind of coming in on this side of the house. So kinda putting a a bow in things here and and transitioning over to more of the writing assistant side of things. All of this so far has been on the review assistant side of Everlaw's generative AI. And so we were looking to help you make review decisions in the most efficient way possible. One other thing, surprisingly not connected to generative AI, but it kinda loosely is, is our story builder feature. So as I clicked into this document, I have maybe eventually made my coding decisions. And if I determine that this is a key document, one that I might wanna utilize in a timeline, one I might want referenced in my list of deposition questions or in a draft based document, instead of using separate software or having to, you know, download this out and upload it to a different part of the, you know, system that you may be using for ediscovery today, you have this button right here, right next to where you might apply some of your codes that is this Quill icon. This is story builder. So I can add this document directly without having to download it to my chronology or, you know, to a a list of depositions or drafts like I have right in here. We've got some examples for these. And so it currently lives in these couple of them. And, again, sort of no download required. I'll show you what this looks like and how they can then be utilized to build, you know, your list of deposition questions or a a breakdown of, you know, the key events throughout your case. So gonna use that as sort of my transition object. We have been in this document the entire time. All you have to do to get back to that home screen is close out of the tab. This one was open the entire time. And then on the home screen, you've got that same Quill based icon for story builder. So two sides of the house here. We have drafts and we have depositions. I'm gonna start with the draft side of things. And so these are collaborative documents. These are living documents that can do something like a Google Doc where you can have multiple people in here at a singular time and make edits on these things, as you go through. So right in here, I'm gonna click on GenAI memo, and you'll see, you know, really nicely crafted factual basis and background of the Imran scandal. Looks awesome, you know, but I I could've gone in and edited this and cleaned it up and had a whole team, you know, make adjustments to this. This was what came out immediately, you know, upon creation of this task. And so I'll walk you through kind of start to finish how we got here. But you have, again, Sparkles icon right on the screen. Click into that, and you can see tasks that have already been run. So here is one that I ran, you know, on November 8th at this time. And what it's asking to do, if I click into it, is to compose a memo that analyzes the factual basis and background of the case using the evidence in this draft, and here is your output. What that looks like in real time is if I am to compose a new one, you tell it, you know, do I want all the evidence for my entire chronology in there? Do I just want what's in this draft that is likely more accurate? Or you could do something that has a particular label, you know, a label for manipulation or, you know, trade secrets or something like that. So as I do that, then you have the ability to structure this in many different ways, analyze or argue many different things, and we'll walk through a few different examples of this. But for the example today, I actually just used a template that someone else created. So you can riff on this with individual people at your firm, at your company. You don't have to be an expert in prompting is kind of what I'm getting at, and I'm not necessarily. So what I did was just grab this template for factual background, and it fills in all this information for me. Know, we're using a memo based template and then we're analyzing this exact thing that I was looking for. All I have to do is hit generate and it is going to kick off that functionality for me. And so what we have here kind of in the the full document over on the left is just the end result of what we got from this output. So you can see, you know, very similar based context over there. But while that completes, I can x out of this and show you. You know, this is an incredible breakdown of the background of the Enron scandal, and each of these little snippets that look sort of like, you know, white rectangles are real documents that have been referenced from your evidence over on the right hand side. And so incredibly useful way to start things off. This is a living document as well, and so individuals can go in here and make edits and really clean this up. It is a great tool for just kicking off the initial draft of whatever document that you're going to be creating. Thus, we call them drafts. But that's one example. I did wanna give one that is it was very, very sought after if I'm just being, super candid. You know? Incredible that we're able to do that, but I really love the structure of a table. So can you structure it in a table? We are able to do that as well. So here is a different example. Very similar set of documents as you can kinda see on the right hand side here. But what we did here, again, just using the evidence in this draft, was I want a cast of characters here. No. So I wanna see each entity identifying as many as possible with the given sources. I want columns for name, the type of entity, facts about them, relevant sources. You could add more. You could add as many of these as as you're looking for and be very specific about what they are. And as I do that, again, you know, we're gonna generate that breakdown and you'll see something that looks just like it does on this. We got the name that we're looking for. In this case, it's in a slightly different order, the type, you know, corporation, individual, and then some facts about them. So really easy to break down things like this in real world situations. Can be extremely helpful. And then an additional, I'll I'll say kind of quality of life feature that we have here is as I'm creating these things, you saw it finish in the background. Until I decide that this is an acceptable output, this is something that I like and I want to use, you can edit it constantly. This is not something that you would get charged for until you choose to insert it as sort of a a final object, in the tool. So you're not going to be wasting a bunch of money, you know, riffing on different versions of what you may be trying to come up with. It is very easy to go kind of back and forth on these things. And the last bit that I'll kinda throw in on that as well is you can see if I highlight something in particular, maybe something that I don't like the output for, but I like the majority of it, you can also rewrite it. You know, you can say I wanna rewrite this in a different format. I wanna include additional evidence for this because maybe it was a bit off. Anything like that. We make it really easy to riff and, you know, sort of do different generations of these sorts of documents. So we've been talking a bit about the draft side. I wanna pivot over into depositions. This can be extremely useful for deposition based workflows. And so over here, I've got my GenAI based deposition. You can see here I've got a transcript. Those who are astute will notice this is not actually a Jeff Skilling deposition. It's for Frank Johnson. We'll get into that, and why that is. But clicking over into my deposition tab, what I did was had it come up with an initial list of deposition questions for me, and then it throws in the sources right along with that too. So same general idea here as I click into assistant, you now have an additional tab. And so I'm gonna talk about the transcript analysis second, but you can see in here that we do have, you know, a breakdown of this. You know, the outline that analyzes manipulative tactics used in the Enron case, each section being phrased in the form of a deposition question. That is effectively what I wanted created and this is the output. And as you choose to insert it, you get something that looks just like this. You know, here's your main points, here's the different sections, and it gives you a breakdown of all of these things. I have not edited anything that you've seen today. It is purely straight out of the writing assistant from Everlaw's side. And all I have done is chosen to paste it right in. You also get a breakdown of which documents have been used and which haven't based on this color coding sort of workflow. So anything that's white has been used. Anything that was gray, it didn't. And so, you know, you added this as a, a physical kind of human being to the deposition. It's very likely that you should have a question for, you know, let's call it this email in particular. If I wanted to add that, you absolutely could. 1, you could re prompt it. You know, tell it to add something in for that particular doc. But also, you can just click in, view the document itself, and say, oh, yeah. You know, I remember this one. I'm just gonna throw in that question myself. I I had a good idea for what that was going to be. So very collaborative, I'll call out. This is not a tool where it just does all the work for you and there's no human input to be, you know, checked or give that final sign off. It's extremely helpful that it is, you know, so back and forth in these workflows. So that's your list of deposition questions. The last one that I wanted to kinda throw out there, and I I did actually throw it in the summary here just to save time, was a transcript analysis. And so as I load my transcript and the actual deposition is completed, we've loaded a transcript. If I click on analyze, you have a few default options here that are quite useful. You know, a deponent background, topic based summary, exhibit and witness summary, or inconsistency and discrepancy analysis. You can also do a custom task. So anything that you might be looking for of that transcript, just tell it what you want, give it the task instructions, and it's gonna run with it. But in our example, we did the inconsistency and discrepancy analysis. And all I did was paste that into the summary just so we could have them both on the screen. It did such a good job. Like, certainly catching things that I may not have. I'm, you know, not necessarily an expert at these things, but looking at, you know, how much in-depth this thing goes, it you know, the deponent did not fully understand the legal implications of an AppWell employment when signing the offer letter, and here's the exact source of where that happened. So you're 1310 to 14. You can jump right over into, you know, your your actual transcript and highlight out any snippets of testimony that you may be looking for there in addition to that. So I know maybe a lot to to take in, but that is sort of my my overview of the writing assistant. Couple other very small things to call out before I hand it back over to Corinne for kinda some some final, thoughts and and wrapping up things. An additional thing that happens here is also, audit history. And so if anything, you know, doesn't turn out the way that you'd expect it to or if you have too many people working in here, you got, you know, unfortunately, a new paralegal that deletes half of your draft that was, you know, really incredible to start with. We have safety checks for those built in and things like history. So if any new content is added, you can always go back to maybe an initial version or anything that, you know, change the document originally. Just some small things that I like to call out along with that. But with that, I'll I'll pause here. I'll I'll kinda stop sharing my screen. I I think we've had some questions coming in on the q and a as well. But, Corinne, I'm gonna hand it back over to you. Sounds good. Thank you so much, Jason. Appreciate you showing us how Everlaw AI assistant is applied in, common workflows. Before we wrap, I just wanted to share, some of what our customers are saying. These are just some of the earliest adopters of generative AI, and they have seen positive experiences and results in their practical application of the tools. And finally, you know, thank you for joining us today. I hope you've taken away a few things from today's webinar that, will help you see how you can use Everly AI assistant, to benefit your firm or organization. As a reminder, you can download some of the content, on that content tab of some of the things we've reviewed today, the, ACC, report, the ASEDS innovation report, the coding suggestions, performance document, all in the content tab. And then finally, if you are curious about how AI assistant can be applied to the work you're doing at your for firm or organization, you can feel free to reach out to our team at everlaw.com to get a tailored demo, and our teams would be happy to partner with you on that. You can also find additional resources there as well. And thank you so much, and I hope you have a great rest of your day.