Name: Going Beyond Manual Review with Predictive Coding
Uploaded: 2026-03-31T18:16:13.767Z
Duration: 56 min 48 s
Description: Going Beyond Manual Review with Predictive Coding

Transcript for "Going Beyond Manual Review with Predictive Coding": Hello, thank you everybody for joining us. We're going to get started shortly. If you don't know me, my name is Zach Souza. I am a senior customer marketing manager at Everlaw, and I am excited to be your host and kick off the AI discovery series with part one today, which is AI and going beyond manual review with predictive coding. Our team is gonna focus on how to use an existing predictive coding model to help you find those hot documents even faster, and we're also gonna get a chance to have you practice within a practice project in Everlaw itself. So we're gonna have some activities to go along with that. Our user education team will be going through instructions shortly if you, need any help. But the whole goal here is to help you experiment and get exposed to predictive coding in a low stakes training environment, really to help you feel prepared and confident to use predictive coding on a real world matter hopefully very shortly. Before we get into all of that, I just want to get into some housekeeping items for everybody. So the first thing I want to mention is today's session will be recorded. If you want to rewatch the session at any time in the future, all you do is come back to this event link, and in just a few hours to later today, once the event wraps up, you'll actually be able to access the full recording to watch on demand. If at any point you run into audio or video issues during the session, you're gonna wanna click and toggle on light mode. There's a button below the main stage where you can toggle that on and off. It should resolve most audio and video issues. As you can see on the right hand side of my screen, there's a chat for the session. Say hi to us. Tell us how you're doing. Say hi to each other. It's a good place just to chat and talk about what's happening with other members of the audience here. We may also pin, important resources throughout the session in the session chat as well. Now if you have a user for the user for sorry. If you have a question for the user education team, I do want to draw attention to the q and a tab. So the q and a tab is gonna be where you can submit questions about anything that relates to today's content on the far right hand side of the chat tab, and one of our team members will follow-up to assist. Please do your best to ask any questions in the q and a tab and not the session chat. We don't want questions to get lost. We wanna make sure we, can follow-up with everybody and, see all the questions that come through. So please do it in the q and a tab, and we're going to do our best to answer all the questions. Lastly, our docs tab right there between your chat and q and a. We've got a number of resources to help you today, including supporting resources about predictive coding, as well as a link to sign up for parts two and three. Just as a reminder, this is part one of this AI webinar series, so you can check the docs tab for a number of supporting resources to help you. And then with that, we are actually going to kick off today's session with a poll question. So to get us started, we just want to get a sense for how the audience feels about leveraging creative coding. So you should see the first poll question up now. How confident do you feel leveraging creative coding for? We're going to keep that up for a couple of minutes, so feel free to take a look. Tell us how you feel about preventive coding. And from here, I'm going to turn things over to our user education team, Margo and Carrie. -Hi, everyone. Thanks for joining us today for our first webinar in our AI series, where we will be covering predictive coding. My name is Carrie, and I'm joined by Margo, and we're both members of Everlaw's user education team. So, this session is going to be interactive, and you'll have some opportunities to practice what you learn in a special Everlaw training project. You should have already received an email invitation to join the training project. There are project access instructions listed in the Docs tab here in Goldcast, So go ahead and open that document up if you need support logging in. And if you're having any trouble getting access, just go ahead and send over your email address in the Q and A tab, and we can help you get set up. One note about today's project. You may have noticed that you won't have model creation permissions, but don't worry, you will have access to review and leverage the models for all activities that we will be working through today. So, as we mentioned, today's focus will be on predictive coding. First, we'll take a moment to provide an overview of what predictive coding actually is, how it works, and what requirements are there needed for setup. We won't spend too long on that piece because we want to give you a chance to actually practice using predictive coding. But with that being said, if you would like to set up a new predictive coding model in your own case, or learn about analyzing and improving its accuracy, go ahead and check check out that resource in the Docs tab titled Learn More About PC. After that, the bulk of our time together will be focused on practical ways that you or anyone in your team can incorporate the predictive coding model into your day to day work. Specifically, we're going to focus on searching for potentially relevant documents, also filtering and sorting by the results table, and then finally, analyze documents using data visualizer. Finally, we'll take a moment to discuss some additional ways that predictive coding can be leveraged to guide review and how to increase the value of a predictive coding model. As we go, please drop any questions in the Q and A and our team will respond to you in real time. For the rest of today's session, we are gonna go ahead and turn our cameras off so everything can be full screen on your end. All right. Thanks, Carrie. Let's start with our overview. So you've heard us say predictive coding, predictive coding several times now, but what actually is predictive coding and how does it really work? So predictive coding is Everlaw's version of technology assisted review, AKA TAR. It studies codes and review work that you and your team are applying and uses that to recommend other documents that are likely to be relevant. This can be a huge help if you need a simple way to refine a search or filter that is bringing in too many results, or maybe you're not trying to filter out documents because you plan to review all of the documents in your case or matter, and instead, you can use predictive coding to prioritize documents that are more likely to be relevant earlier in your review cycle. And once review is underway or nearly complete, predictive coding can also be used to quality check your team's review work by flagging any documents that haven't been coded at all or were inconsistently coded. Now, for some folks, predictive coding may still feel like a strange concept. However, a program that can learn what you want based on what you do is actually really common. So social media platforms, targeted ads, and streaming platforms all function in a similar way. For example, if you've ever used Netflix, your homepage has a top pick section with videos and shows that ideally reflect your interests. As you watch various content and give your thumbs up, thumbs down, Netflix is analyzing those shows and movies, seeing what they have in common, and from there recommending other content you're likely to enjoy. Now I said ideally earlier because sometimes your recommendations can get a little mixed up, and that's especially true if multiple people, like your kids, use the same account and they all like different things. With Netflix, creating separate profiles allows each person to get their own tailored recommendations. And in Everlaw, creating separate predictive coding models to look for documents that are responsive, privileged, hot, etcetera, will prevent predictive coding from getting confused about what makes a document relevant to you and your team. Best of all, you can have as many predictive coding models as you need to find relevant documents at no additional cost. Speaking of costs, what does it cost to use a predictive coding model, who can use it, and how can we set up our first model? Luckily, the answer to that first question is nothing. Predictive coding is included with every Everlaw subscription and is available straight out of the box. But not all users will see it or have access to it right away. By default, only a project admin will have access to the predictive coding setup page, but they can give others permission to view existing models and create new ones. So, as far as setup goes, you'll need to provide Everlaw with a few things. I'm going to go ahead and pop over to our platform, and I am going to move over to our predictive coding page. So, some of the things we need. Everlaw is going to need to know how do we know if your documents have been reviewed? How do we know if that document is relevant or what you are looking for? And finally, you'll need to review some documents in the model so it can learn from your directions. So, I'm going to click new model first. So, we have our reviewed and our relevant. So, for example, if I were creating a predictive coding model to look for responsive documents, any document that has been coded as respond under that category responsiveness, whether it's actually responsive or not, has already been reviewed by my team. So, for that, I would select responsiveness. If I want my predictive coding model to find more documents that are likely to be responsive, these would be my relevant documents. So, I'd select responsive. For this, I'm going to exclude produced documents. And then my final step is going to be to name my model. And now that my model is all set up, my team and I will need to review at least 200 documents before my predictive coding model can actually start giving me suggestions. Of those 200, only 50 of them need to be responsive, and the rest could be not responsive. Best of all, you can set up a predictive coding model before review even begins, and your model will start generating results once analysis on your case documents is complete. I know we just use responsiveness as our example, but predictive coding models can be used to find a wide range of things, such as privilege status, confidentiality, key issues, ratings, and more. It just depends on what you're looking for. If you'd like to learn more information about creating a predictive coding model and reading your model, those details are in the resource that we shared earlier, that learning about more PC, which is also linked in the docs tab. Likewise, if you'd like more information about any feature we shared today, check out the feature page on our training center for more information. All right, now that we've explained a little bit about predictive coding models at a high level, we're going to put you all to some work. So let's pretend Margot and I are lead attorneys on a massive case against some group called Enron, which maybe you've heard of. Anyway, we're dealing with a massive dataset, so we'll have to do several rounds of production. Unfortunately, somebody dropped the ball with scheduling, so we have less than an hour to find as many relevant documents as possible and share them with our review team before our production deadline. Luckily, we had the forethought to set up two predictive coding models at the start of the case, one looking for potentially responsive documents and the other looking for potentially privileged documents. And since we are all new to the team, I'm going to ask you to watch Margo and I while we hunt for responsive documents, and then we'll hand it off to you to find privileged documents. I'm gonna go ahead and switch my screen share so that way Margo can take lead now. Thanks, Carrie. Let me get that going. All right, so let's jump into the Everlaw platform and we'll start with searching. So when I'm on the Everlaw homepage, can get to the search page by clicking the magnifying glass in the toolbar at the top. And from here, I can build a search using the box in the middle of the page and the menu of options on the left. Now, our case is one that involves fraud, embezzlement, and downright scamming of investors by these folks at Enron. So if I'm interested in documents that mention fraud, I can type the word fraud into this contents box and hit enter on my keyboard. Once I do, I'll see a window pop up at the bottom of my screen that shows me the total number of documents that match my search results, and I can see a random sample of about 10 documents here. Now I've set things up so I can see prediction scores in my example results. That way I can have an idea of how relevant these documents are likely to be. If I scroll here, I can see that some of these documents have high prediction scores for responsiveness, which suggests that they're more likely to be relevant, but several others have pretty low scores. So if I'm curious about any of these sample documents, I can also click the eye icon over here on the far right to see the actual contents of the document itself. Now, having to review 2,074 documents wouldn't be an issue if most of them were likely to be relevant. But since there are several here that have low scores and I'm short on time, I wanna update my search to be more specific about what I'm looking for. So for example, I can edit my content search to add the text Ian liability, and this will look for documents that mention both of these words. When I hit enter, we'll see that the number of documents in my results decreased, but there are still quite a few here with low prediction scores. So to narrow down these results even further, one thing I can do is refine my keyword search and continue to add more terms here. And I do want to point out that when you click on this term and you can click on the eye icon, that will pull up a nice cheat sheet for how to construct advanced content searches like wildcard searches and proximity searches. Alternatively, I can minimize this preview panel by clicking on the bar here so I have a little more space, and I can use the menu on the left to add more criteria to my search that might help me find those documents of interest. And again, since I'm short on time, I want to focus on the documents that the predictive coding model thinks are most likely to be responsive. So under the review category, I will click show more terms and select the predicted term. I also want to point out you can use the find a term option at the top left or press F on your keyboard, and from here I'm going to type in predicted. So from this dropdown, I can choose the specific predictive coding model that I'm interested in and input the desired score range. So I'm looking for responsive documents, so I'm going to select the responsiveness model. And if I want to see documents with a score of 50 or higher, I would type 50 into the first box and leave the second box blank. I can also type in an end number if I want, like 80, if I do have a specific range in mind, but in this case, I just want to see everything with a score of 50 or higher, so I'll leave that blank. When I click out, we can see that this dramatically reduces my search results to just a couple 100 documents. I'll reopen the preview panel to check out the sample docs, and I can see many of them have a high prediction score for responsiveness, which is great. All right, so I just found the potentially responsive documents, and now I'd like for you all to find documents that are potentially privileged. Let me get back to my slides here. So from the search page, I want you to look for documents that contain the word attorney and that have a privilege prediction score of 90 or higher. Don't worry if you don't see all of the prediction scores in your preview window. We'll talk about how to set that up next. Your timer starts now. All right, we're about halfway through our time. When you're finished, go ahead and fill out the poll on the right to share how many documents you found. It should show up on the right under the poll tab. And if you finish the activity early, go ahead and try updating your search to look for documents that have the word attorney and the word privilege. All right, time's up. So let's quickly go over how to build this search together. Back in the Everlaw platform, I'm going to click the magnifying glass to go to the search page, if you missed that at the beginning. And we're looking for documents that include the word attorney. So we'll use the contents term here and I'll type in attorney. And then from here, I can use the find a term button on the top left and add in the predicted terms. And I'm looking for documents with a privilege score of 90 or above, so I'm going to select the privilege status model and then type 90 into the first box, And I'll leave that last box blank. And when I'm done, I should have about 1,600 documents here. So sixteen forty eight is the precise answer. And for those of you that tried the challenge to find documents that also have the word privilege, we go back to the contents term and type and privilege. I'll end up with five thirty documents. Now to actually see all of these documents, I'll click the search button. And here we are in what we call the results table. So for those who aren't familiar, this page works a bit like a spreadsheet. Each row is a different document, and then each column tells me different information about these documents. So in my results table, I've customized my view so that I can see the predicted responsiveness scores for my documents. And let me show you how I set that up. First, I'll click the view button at the top, and a view is what we call the layout for your results table. I'll click add or remove columns, and in this pop up window, I can either scroll through the options here or type in prediction to find my predictive coding models. So since I already have my responsiveness model selected, I'll check the box for the privileged status model and then hit save. And now I can see over on the far right, I have my privilege status column, and I can drag it over and resize these columns as needed just so it's super easy to see what they are talking about. And as I can scroll through here, I can see that these results do indeed have a privileged prediction score of 90 or higher, just like I hope. So to help me prioritize which documents to review even further, I can also sort my documents by their prediction score so the highest scoring documents are first in the list. To do that, I'll click on the small arrow icons here and then choose the single column sort descending, so that way I'm starting with the documents that are most likely to be privileged. In addition to sorting, I can also filter documents by their prediction scores. So for example, in the responsiveness column, I can see there are quite a few documents here with low predicted responsiveness scores, and since we have this tight deadline and I can't afford for reviewers to worry about checking documents for privilege if they aren't even responsive in the first place and won't be produced. So I will click the funnel icon here and I'm going to adjust the range so I only see documents with a responsiveness score of 50 or higher. I can type that in or drag it here. Let me do this. And then I'll click add. And now we can see that we've narrowed down our results to just over 100 documents. Now I need to share this list of documents with our review team, and the simplest way for me to do that would be to click the share button up here and then choose a person or group. I'll pick Carrie. I'm going to send them a message. Then click share. So that's one way to do it. Alternatively, I could click the batch button and that gives me a few different options. I could click modify and then put these documents into a shared binder for us. Or if you're a project administrator, you can click assign and follow these steps here to create an assignment for these documents. And assignments allow you to evenly distribute these documents among your reviewers and have Everlaw automatically track their work so you can see everyone's progress in real time. Best of all, if you sorted the documents before assigning them like we just did, that sort order will be reflected in the individual reviewers' batches. So they're gonna see the documents that are most likely to be relevant first. Now, one last thing to note about the results table is that you can have more than one view or layout depending on what information you want to see. So to save this layout so I can use it later, I can click the view button and then select save view, and then I can click new view and give my layout a name. Fun fact, if you are a project administrator, you can also create a default view for the entire project, so everyone will see this view automatically and you can make sure that they are seeing prediction scores. I won't do that today, but just wanted to point that out. When I'm ready, I'll hit save. All right, it is your turn to create your own view now, so let's go to our next slide here. And since we already know your documents have a high privilege score, I want you all to try adding the prediction model responsiveness column to your results table. If you finish early, try using filtering to show documents with a responsiveness prediction score of 80 or higher. I'm going to give you all three minutes, starting now. Alright, that is our time. So, I'll quickly run through these steps so everyone can see and we can all be on the same page. So let's go back to the Overlap platform and I'm gonna remove my filters here and then I'm gonna click into the view and make sure I'm using the same view as you all. So let me switch here. Cool. So to adjust the view like you all just did, I'm going to click view again, then add or remove columns. And then I will add in the prediction model responsiveness column. Make sure that box is checked and then hit save. And then again, that's going to show up on the right, but I can drag it over and resize it as needed. Make sure I can see both here. And so for anyone who did the bonus task, we can click this funnel icon to filter our results to just see documents with a responsiveness score of 80 or higher. All right, so now you know how to use predictive coding to refine your searches and then prioritize your results to find those that have higher prediction scores. I'll be honest, I am a little nervous that we might have missed something that's potentially relevant by just focusing on those keyword searches. So Carrie, I'm wondering, do you have any tips on other ways that we can check for documents of interest? Definitely do, Margo. So I'm gonna go ahead now and share my own screen. So that way I can walk through some of those additional tips. Alright, so if I want to see the bigger picture of all of our documents and the different prediction scores, I can use Everlaw Data Visualizer tool to get that really high level view. I can access this tool by coming up to our document analytics tab and then selecting data visualizer. Here, I'm going to see a bunch of different graphs or visualizations that give me a breakdown of the number of documents with certain qualities. This dashboard has a few pinned already down here. And then there are a ton of additional options on the left hand side. Currently, these are the visualizations that reflect all of the documents in our case, and we can see that from this number up here. I can also then select certain visualizations to get down to a smaller subset of documents. Now, I know we've been talking about finding responsive documents, but as we mentioned earlier, we also need to find documents that are likely to contain privileged information. For our case, we're particularly interested in reviewing documents from a specific custodian, S. Shackleton. So, to narrow down my results for just documents from S. Shackleton from that custodian, gonna come over to the Peoples tab here. And then I'm gonna select Custodian. This is gonna show me all of the documents from different custodians across my case. Here, I can see S. Shackleton. This is one of my most frequent custodians. And if I want to hone in on Shackleton's documents, I can click this bar. And then I'm going to go ahead and on the right hand side, hit the button that says add filter. Now, you can see I've narrowed down to just a little over 15,000 documents, but we definitely don't have time to review them all. And this is where that predictive coding can help us prioritize which documents to review. I'm going to come back over to the left hand side, and this time I'm going to type in the word prediction. And you'll see my prediction models that I already have created. For this one, I'm going to select Privilege Status. And now we're going to see a graph showing all of the documents from our custodian at Shackleton organized by our prediction score. You'll notice that there are a lot of documents over on the left hand side of our graph that have a low prediction score because the model thinks that they're likely to be they're unlikely to be considered privileged. However, there's still several documents on the right hand side that our model thinks are very likely to be coded as privileged. We definitely want to prioritize these documents over here on the right hand side. So I'm going to click and drag, So, that way I can see my documents with a prediction score of 80 or higher. And now, with that, I can see that there are nine twenty five documents with the prediction score in this range. To actually start reviewing these documents for privilege and see if the model is right, I can then hit add filter. So, I'm combining those two filters, that custodian and that privilege status. And then I can go ahead and hit apply filters to search. Now I'm seeing a results table with just those nine twenty five documents instead of that original 15,110. So now it's your chance to try this out for yourself. On your own, go ahead and open Data Visualizer, filter for the documents with the custodian T. Jones, and then filter for documents with the prediction score for the responsiveness model of 95 or higher. Bonus task, you can sort your results table by the prediction model responsiveness. End of this activity, make sure to answer the poll to let us know how many documents you found. You got about two minutes left. All right, time is up. The correct answer is 2,624 documents. Let's walk through how we got that number. So, starting out on our homepage, step one is open Data Visualizer. So, it's that Document Analytics button and then Data Visualizer. Our first stop is going to be over on the People's tab, and then we're going to select Custodian. And again, we're looking for the custodian T. Jones. If you stopped right there, that's how you might have ended up with that answer 15651. However, we want to find emails that are predictive to be responsive. So we need to add this filter, and then we need to go ahead and find our prediction model and select responsiveness. We want to filter specifically for a prediction score of 95 or higher, so we'll go over to 95. And then we're going to add filter and then apply filter to search. And that's how we get that answer 2,624, which you can see in the top left corner. If you got the answer that was a little over 100,000, you probably just filtered on the responsiveness model, but not the documents from the custodian T. Jones. If your answer was three eighty, you probably filtered to the right custodian, T. Jones, but filtered on the privilege status model instead. For the bonus task, remember that you can sort results table by the prediction model responsiveness column, so that way you can review the most relevant documents first, which we have here. All right, so as we just saw, Data Visualizer offers a very powerful way to get a quick bird eye view of your documents and their different qualities, like their prediction scores. Data Visualizer is also a great option for building searches visually without having to input specific values like you do on the Everlaw search page. Fun fact, you can also access Data Visualizer for a specific set of documents directly from the results table. So, if I go back to my results table here, I then can select where it says visualize and then have these filters already applied. Great job, team. We have put our findings now in a binder to share with our review team, and now they can sort and filter these results as needed. With the help of our search tools and predictive coding, we managed to narrow down an initial review set of over 300,000 documents to fewer than 4,000, and it only took a matter of minutes. Now that our part is done, I want to share some other ways that our team can use predictive coding, including a sneak peek into how our review team is going to get through those documents so quickly. If you use our search term report to find documents that hit on one or more key terms or search criteria, you can then edit the searchable set to narrow down the documents in the results by the document prediction score. Or if you're using clustering to see which people or terms are most common across an unfamiliar set of documents, you can actually color your groups by their prediction score to see hotspots of potentially relevant documents. Thanks for mentioning all that, Carrie. And it's great that there are so many ways to find documents of interest depending on what you're comfortable with, and any of those tools can be used in combination with predictive coding. So, so far we've talked about how predictive coding can be used to take your searches to the next level. Now I want to talk about how you can take predictive coding to the next level using a tool called coding suggestions. So predictive coding on its own can point us to the documents that are predicted to be relevant, but it won't tell us why a document was predicted relevant. Someone would need to review the document itself, define those keywords, people, or dates, and then determine and confirm if a document is indeed relevant or not. However, if your organization has signed up for the Everlaw AI assistant suite of tools, then you can take advantage of coding suggestions. This tool can suggest what code you should apply to a document and it can also direct you to the specific passage in the document that the suggestion was based upon. So I will take over the screen share here and show you a little bit about coding suggestions. All right, so back in our project, I already have a review set binder of potentially responsive documents. And I'm going to sort this results table by number of pages for each document. And we can see that there are several documents here with hundreds of pages. And I'll open up this first one. Now, in some cases you might not have time to read through every page, especially if there are hundreds of them, and that's where coding suggestions can really save the day. By clicking the AI Sparkle icon in the top of the context panel on the left, under coding suggestions, you can click generate to have Everlaw review the current document and provide suggestions about which codes might apply to this document and why. So for your own documents, you might be checking for responsiveness or privilege, and for external documents, you might be looking for key issues. Either way, once the suggestions are generated, you'll see the name of the coding category being considered, in this case responsiveness. You'll see the specific code that's suggested, and in this case we have a soft yes suggestion for the code responsive, and then a quick blurb about why Everlaw suggested this code, and a link to the relevant area of the document that was used to generate the suggestion. And this allows you to verify the AI suggestion and make an informed decision yourself about which codes to apply. Okay, so I'm going go back to our slides here and that was just a sneak peek at coding suggestions. If you're interested in learning more about this or any other AI features, definitely reach out to your Eberlaw representative for more information. We'll also cover some additional AI features in parts two and three of this webinar series. So during those sessions, we'll go over best practices for AI prompting and how to leverage a deep dive throughout the life cycle of your case. I do want to point out that like predictive coding, many of the Everlaw AI tools are now included in your contract at no additional cost. All right, we've covered a lot in this session. As you can see, there are so many ways to leverage predictive coding once you have the model set up, but it does require setup at least the first time. If you have a predictive coding model that is consistently and accurately locating documents of interest, for example, privilege, multi matter models allow you to copy the learnings of a predictive coding model from one case to another. If you want to learn more about multi matter models, go ahead and reach out to your Everlaw representative. All right, for folks on the call, I'm curious to know after learning a bit about predictive coding and getting hands on, how comfortable do you feel with leveraging predictive coding for review? I'll give everybody some time to answer that, and while folks are answering, I want to take a moment to recap what we've covered today. We've covered how to leverage predictive coding models when searching for documents, filtering and sorting the results table, analyzing documents using data visualizer, and managing review. Again, this can be super valuable if you need to find relevant documents faster, if you're running quality checks for your team's review work, or you want to further refine or filter a search that is bringing in too many results. If you'd like to learn more about predictive coding, be sure to check out the coding resources linked in the Docs tab, the predictive coding resources linked in the Docs tab, which includes information about how to improve the accuracy of your models. In your Docs tab, you're also going to find a link to all of our feature pages. If there's a particular tool that you found interesting, check those feature pages out for more information. And with that, I will pass it off to Zach to close out. Thank you, Margo and Carrie. Great overview of predictive coding. I also want to call out all of our support staff answering questions. Give them all some thanks in the chat. They really work hard to make sure everything runs smoothly and give you a great experience for this. So, as we get to a close, I have a couple of announcements. First off, we want to share the feedback survey for today's session. This goes straight to the user education team and it just takes a minute. Feel free to fill this out. You should see a link across your screen now. This helps make all of our future events that much more applicable to what you're interested in and want to learn from Everlaw. So please take a second to give us your feedback there. And if you're interested to learn more from your peers, so other members in the audience or other Everlaw users out there, don't forget to join the e discovery community. So we've got a button here at the top, join the Everlaw community. If you're not already a member, click there. You can go to the community, post and ask questions about workflow tips, see how others are getting the most out of, getting the most out of predictive coding, and even give feedback to Everlaw on new and soon to be released features. So we've got a lot of cool things going on there. Check it out. Click on join the community today. And then the last thing I'm going to share just as a one final reminder, we have two more sessions in this series coming up. One at the April. This session is going to be our next session focused on AI prompting. So we're going to be using coding suggestions as our tool to practice AI prompting here. But really what you're going to learn in this next session can be applied to any AI tool that uses natural language prompts. Right? So from simple chat GPT to coding suggestions, the tips and tricks and and really strategies that you're gonna be learning here is to help make the most out of your prompt for your legal work. We're also gonna have part three coming up later in May. This is a brand new series. We're gonna go into Everlaw AI deep dive, which we get tons of questions about from our users on how best to use it and when. So if you're interested in learning about Everlight AI deep dive, this is our first big training session about it. I highly recommend coming out for that one. Great. So with that, I will end today's session. Thank you all. Have a great day.