Latest / CLEs You Actually Want to Hear / Depositions in the Age of AI: The Competitive Edge You Can't Afford to Ignore
Transcript
- 0:03Hello and welcome to depositions in the age of
- 0:05AI, the competitive edge you can't afford to
- 0:08ignore. I'm delighted to introduce our speaker
- 0:10today, Carl Seelbach. Carl, the floor is yours.
- 0:14All right. Well, thank you very much for that
- 0:16brief introduction. Let me get right to it. In
- 0:19the interest of time, let me get my screen share
- 0:21set up. And if you can't see that for any reason,
- 0:25please interject and let me know. Otherwise,
- 0:27I'm going to dive right in. All right, we're
- 0:29going to be talking about a topic that I am very
- 0:30passionate about, which are how to leverage AI
- 0:34to get ready for depositions in real time during
- 0:37the deposition itself and then after the deposition.
- 0:41AI is really changing how many of us take depositions
- 0:44and hopefully by the end of this course, you'll
- 0:46have some ideas of things you could actually
- 0:48put into practice at your firm tomorrow. So let's
- 0:52dive in. Any good litigator knows that you've
- 0:54got to give your audience a roadmap. So here's
- 0:56kind of an overview of where we're going to go
- 0:57today. I'm going to give you a brief introduction
- 0:59of myself and my background as well as my law
- 1:02firm inscribed talk a little bit about non -stenographic
- 1:05depositions, digital reporting, ethics rules,
- 1:09how to use AI responsibly. And then we'll really
- 1:12talk about it in three separate phases, pre -deposition,
- 1:16live during the deposition, and then post -deposition.
- 1:19And we'll go through examples of how to leverage
- 1:21some of the latest technology at every step,
- 1:23really to make you a better advocate so you can
- 1:25take a better deposition in your cases. And then
- 1:28we'll wrap it up with a few other points around
- 1:30responsible AI use, how to implement it in your
- 1:33law firm. And then, of course, if I haven't answered
- 1:35some of the questions along the way, I can do
- 1:37it at the end. Speaking of questions, you're
- 1:39certainly welcome to drop questions in the Q
- 1:42&A box. I will do my best to monitor it. I can't
- 1:46promise I will see all of them. But I will try
- 1:49so I have the Q and a panel up on my. computer,
- 1:53and I'll try to keep an eye on it. So, if you
- 1:55have questions, you're certainly welcome to ask
- 1:56them. So, a little bit about me. I think this
- 1:59is always kind of fun. I try to keep it under
- 2:0030 seconds. I was born in Houston, Texas. I am
- 2:03a dual citizen of Norway. I actually grew up
- 2:05there for a few years before moving back to Texas.
- 2:08I grew up in Northeast Texas, went to Stephen
- 2:10F. Austin for undergraduate, where I met my lovely
- 2:12bride Adrienne. I did mock trial in Moot Court
- 2:15at South Texas before joining Winstead, which
- 2:18is a very large Texas firm. And in Roughly 2008,
- 2:23moved to Austin, where my wife and I had two
- 2:25little kiddos, Annabelle and Millie. Started
- 2:28my own firm in 2015. Started a legal tech company
- 2:31just a few years ago, Scribe. I'm a member of
- 2:34the DRI. I'm actually on the National AI Task
- 2:36Force for the DRI. And I speak regularly on kind
- 2:40of the topic of leveraging AI use and litigation.
- 2:43Most of my time, money, and effort, if it's not
- 2:45dedicated to my legal practice or to legal tech,
- 2:48it's going to club volleyball. So I've learned,
- 2:50like many of you who have kids in club sports,
- 2:53that it is all -consuming and kind of like a
- 2:55part -time job. So enough about me. A few disclaimers.
- 2:58I'm not your attorney, OK? Please don't cite
- 3:01this presentation in some type of motion or brief
- 3:04or anything like that. I'm hoping it's going
- 3:06to be entertaining and educational. That's kind
- 3:09of my goal. That's the way I like to present.
- 3:11So hopefully this will be engaging and you'll
- 3:14get some really fun ideas out of it as well.
- 3:16So with that said. A few other intros. So what
- 3:22is Scribe? I'm going to talk about that a little
- 3:23bit. It's going to be weaved into this presentation.
- 3:26I've taken hundreds of depositions over my career,
- 3:29taken and defended, and I always thought there's
- 3:32got to be a better way to do this. It would be
- 3:34nice if we could leverage technology, layer it
- 3:37into every step of the process, being able to
- 3:40get a real -time transcript for no extra cost,
- 3:43being able to have video synced to the text to
- 3:46where if I want to create clips, I can do it
- 3:48really really easy without extra software without
- 3:50having to spend a ton of money and time. That's
- 3:54what my co -founder and I have built at Scribe.
- 3:56It's a new way to take depositions non -stenographically.
- 3:59We'll talk about that in a minute. But I'm going
- 4:01to go over that. I'm going to go over all sorts
- 4:03of other tools as well. Again, depositions are
- 4:06something I've just taken a lot of them. So I
- 4:08really enjoy the technology around it. And I
- 4:11think you're going to have a lot of fun along
- 4:12this presentation, hopefully. So what to expect
- 4:15kind of in a non -stenographic deposition, whether
- 4:18it's with Scribe or some other companies, digital
- 4:20reporting is becoming more and more common. kind
- 4:23of tell you what to expect demystified a little
- 4:25bit. During a non -synographic deposition, you're
- 4:28going to have professionals in the loop. There's
- 4:30going to be a human being, a digital reporter,
- 4:33there as a certified notary to soar in the witness,
- 4:36to administer the oath, to monitor the quality
- 4:39of the audio and the video recording and the
- 4:41signals. to monitor the quality of the real -time
- 4:43transcription. And then you may also have access,
- 4:46depending on which platform you utilize for this
- 4:48type of tool, to generative AI features things.
- 4:53And we're going to go through some examples of
- 4:54that later, but being able to chat with that
- 4:56transcript, being able to analyze it in real
- 4:58time, analyze documents against it. There's a
- 5:01lot of really powerful things you can do with
- 5:03some of the latest large language models as you
- 5:06take depositions. And we're going to get into
- 5:07that today. With a non -synographic deposition,
- 5:11aka digital reporting, you get the benefit of
- 5:13a same -day record. And what I mean by that is,
- 5:16as soon as the deposition's over, you have a
- 5:18high -quality rough transcript synced to video
- 5:21that you can leverage. So again, litigators on
- 5:24the video call, you can draft discovery, you
- 5:28can send your settlement demand or counter, you
- 5:30can... report to your client. You can summarize
- 5:33it all same day. So it kind of takes the waiting
- 5:36game out of it. I know when I took a ton of depositions,
- 5:39I'd always have to ask my assistant or my paralegal,
- 5:42you know, where's the transcript? Oh, it's been
- 5:44three weeks. We still don't have the transcript.
- 5:46Those types of days are over. You know, if you
- 5:49modernize your system, you can get things really,
- 5:51really fast. And then within a week, you get
- 5:54a certified final transcript. That includes a
- 5:58verbatim transcript, word for word, line by line,
- 6:01legally formatted, and the companies like Scribe
- 6:04handle the errata process end to end. So that's
- 6:07just a brief introduction. Let's talk about not
- 6:09Scribe specifically. Let's just talk about the
- 6:11law on it. So if you practice in federal court,
- 6:15the rules around not stenographic. depositions
- 6:17changed in the early 1990s, probably before some
- 6:20of you were even practicing law. And that was
- 6:22before I started practicing law. And I've been
- 6:24practicing law for over 20 years at this point.
- 6:26And then the federal rules, it's federal rule
- 6:2830B3. Most states have similar rules. I am well
- 6:32aware that there is We'll call it a little bit
- 6:35of controversy and lobbying going on in the state
- 6:37of California around digital reporting. I'm also
- 6:41aware that there's a shortage of stenographers,
- 6:43even at the court level. But I think this is
- 6:47where all the states are headed. Most states
- 6:48are already there. Hopefully, if any of you are
- 6:51involved in California's legislative process
- 6:54or lobbying, you will try to get them to modernize.
- 6:59Georgia just did. Georgia just in the last year
- 7:01or so is updated there. around it to allow other
- 7:05forms of depositions, including non -synographic
- 7:08options, because it's become a proven process.
- 7:11You always have the video, you have a human in
- 7:14the loop monitoring everything, and you get a
- 7:16final verbatim transcript. So it's very proven.
- 7:19One of the things when we talk about technology,
- 7:20and we're going to get into some generative AI
- 7:22tools in a minute, but when we talk about tech,
- 7:25I think it's important to kind of try to... demystified
- 7:28a little bit and separate some of the fear -mongering
- 7:31from kind of the reality. And we'll talk about
- 7:33that in the context of data privacy. I want to
- 7:36briefly talk about it in the context of using
- 7:39software to help create a transcript. So what
- 7:44you see on the screen is kind of a dichotomy
- 7:47of sorts. When companies like Scribe and others
- 7:50are creating final transcripts with the help
- 7:53of software, they're not using ChadGPT. They're
- 7:57not using so -called generative AI. What they're
- 8:00using is a proven form of software called automated
- 8:03speech recognition. It's been around for over
- 8:06a decade, probably 20 plus years. It's become
- 8:09a lot better in the last five or so, but it essentially
- 8:13takes that spoken audio, that high quality video
- 8:16and audio recording, and it transduces the phonetic
- 8:21sounds into the words that you and I can read
- 8:24on a page. It's actually, ironically, very similar
- 8:27to what a stenographer does. They type phonetic
- 8:30notes. This is software taking phonetic notes
- 8:33essentially and converting it into something
- 8:35you and I can read. And then once that first
- 8:38draft is created, which is the rough draft, then
- 8:42a professional transcriptionist goes line by
- 8:44line, word by word, and checks it against the
- 8:47source audio and video file. So nothing's ever
- 8:50added. Nothing's invented. Nothing shows up on
- 8:52the page that wasn't actually set. Whereas with
- 8:55generative AI, for those of us who work with,
- 8:58whether it's legal tools or some of the more
- 9:00common, chat GPT, Gemini, Claude, all the different
- 9:05tools that are popular nowadays, it's a completely
- 9:08different type of system, right? You're prompting
- 9:11it or giving it a task. A lot of us now are maybe
- 9:14you're working with AI agents and it is trying
- 9:17to predict the right answer. I co -presented
- 9:20with someone who said it's like a word calculator.
- 9:23It's basically trying to predict through some
- 9:27very sophisticated algorithms what the answer
- 9:30should be. So it is generating new text. It is
- 9:33predicting the next word. So those two concepts
- 9:36are very, very different. I see we have some
- 9:38questions. Some of these are going to be answered
- 9:40as we go. And so I'll answer a couple of them
- 9:46now. So in most jurisdictions where non -synographic
- 9:50depositions are permitted, the video record is
- 9:53the official record and the transcript becomes
- 9:56essentially a demonstrative, an aid. And so what
- 10:00I mean by that is you still file the transcript
- 10:02and the rules still permit you to file the transcript.
- 10:05But if anything in the transcript is ever questioned,
- 10:09you have the source audio and video as the official
- 10:11record to check it again. So both are certified.
- 10:14The video is certified. The transcript has a
- 10:17declaration of verbatim transcription accompanying
- 10:19it. So that's a very great question. And then
- 10:22there's another question about recording methodologies
- 10:25for non -stenographic depositions and which jurisdictions
- 10:27that's allowed in. You're welcome to check that
- 10:31out on Scribe's website. There's a page dedicated
- 10:33to that. It's 40 plus states that permit non
- 10:36-synographic depositions, many of them by write,
- 10:40some by stipulation. California, at this point,
- 10:44it requires stipulation, but hopefully that will
- 10:47change in the future. The recording method is...
- 10:50There's multiple methods, so it's like a redundant
- 10:52method, and the real -time transcript is occurring
- 10:54as the deposition is unfolding. So most companies
- 10:58that do it, do it in a way that the recording
- 11:01doesn't fail. It's really not possible for it
- 11:03to fail. It would be caught immediately through
- 11:06the quality control that's taking place with
- 11:08that digital reporter. So that's a great question.
- 11:10Hopefully I took a few of those questions. Chain
- 11:13of custody is managed by your digital reporting
- 11:16firm. So whoever you use for that, whether it's
- 11:18a scribe or someone else, they handle chain of
- 11:20custody. They gather exhibits. They certify the
- 11:23authenticity of the video and audio recording,
- 11:25all those things. So getting some really great
- 11:27questions. We've got an inquisitive bunch, which
- 11:29I really, really like. What's on the screen is,
- 11:32I want you to think of this as kind of table
- 11:34stakes for any legal tech solution you're using.
- 11:38Okay, none of this is specific to Scribe. You
- 11:41need to be working with closed systems. And so
- 11:43what I mean by that is your data has a very secure
- 11:46pipeline and it's never exposed to other law
- 11:50firms. So for example, for Scribe's digital reporting,
- 11:54that doesn't include any generative AI features.
- 11:58feed through a large language model. For the
- 12:00record itself, that's all siloed, private. Scribe
- 12:06has optional, like many companies. generative
- 12:10AI features. Scribes are powered by Anthropic
- 12:13Claude's Frontier Model. You may hear Frontier
- 12:16Model thrown around lately. That basically means
- 12:18the really big ones, right? The companies that
- 12:21are powering most of these experiences. It's
- 12:23OpenAI, it's Anthropic, it's Google. Recently,
- 12:26it's XAI with their latest Grokbot that just
- 12:29launched a week or two ago. But the point of
- 12:32that is you want to make sure that the companies
- 12:34you're working with understand the importance
- 12:36of what we do as attorneys, right? That it needs
- 12:40to be secure, it needs to be siloed, the company
- 12:44needs to be willing to sign a BAA, that it's
- 12:47HIPAA compliant, that it's SOC 2 compliant. All
- 12:50those things are extremely important. And the
- 12:52second item, maybe the most important, which
- 12:55is what you are putting into the system. If you
- 12:58use the generative AI features, that it's not
- 13:01being used to train the model in a way that your
- 13:04information could leak out to someone else. And
- 13:08so all those things are really important as you're
- 13:10vetting any tools for your firm. Ethics rules
- 13:14kind of overlap with some of the things I just
- 13:16talked about. There's a lot of rules around this
- 13:18being discussed in a lot of different states.
- 13:20I am not licensed in California. Disclaimer,
- 13:23maybe I should have said that at the very beginning.
- 13:25I am licensed in Texas. I practice in state court,
- 13:28federal court, and in arbitrations with the AAA
- 13:30and Jams. So I'm most familiar with federal rules
- 13:34and Texas rules. There has been some guidance,
- 13:38right? There has been some guidance promulgated
- 13:40by the American Bar Association as well as many
- 13:43of the state bar associations. I've read a lot
- 13:46of it. And really what it seems to boil down
- 13:48to are three core concepts. Every time I read,
- 13:53whether it's a new case or a new opinion or a
- 13:55new rule or something, they all seem to revolve
- 13:58around three core concepts. And those are confidentiality,
- 14:03right? Protecting the confidentiality of your
- 14:06client in whatever tool you use. Two is competence.
- 14:11And it's kind of the duty for us as lawyers,
- 14:13and it's probably one of the reasons you're taking
- 14:15this in addition to getting CLE credit, is to
- 14:18remain competent and trained and aware of what
- 14:21the latest tools are that we can put in our tool
- 14:23belt as attorneys. And then third is client communication.
- 14:28And that's really making sure that we are communicating
- 14:31to our clients. the types of tools that we're
- 14:34using to advance their cause. And I will say
- 14:38the way that I do it at my firm is I've weaved
- 14:41it in to two different places. One. my engagement
- 14:46letter. I'd be happy to share it with you if
- 14:48you're interested. I'll put my email on the screen
- 14:50in a couple of slides. But I have a standing
- 14:53clause in all my engagement letters now that
- 14:55reference my firm is using software including
- 14:58but not limited to generative AI and kind of
- 15:02explains the policy that my firm uses to the
- 15:04client. I also have started putting it in my
- 15:07firm's rate increase letters, which go out every
- 15:11year or two, depending on the client, to let
- 15:13them know that we are using generative AI and
- 15:16other software tools. And partially, it's to
- 15:19try to save them money. So make sure you are
- 15:22communicating to your client. The ABA has even
- 15:24gone so far as to say in some of its guidance
- 15:27that if using generative AI would... save your
- 15:32client money. You really need to let them know
- 15:34if you're choosing not to use it. So I'm not
- 15:38saying I necessarily agree with that guidance,
- 15:40but they have said that, which is interesting.
- 15:43Texas, I won't spend a lot of time on this, but
- 15:45Texas has looked at this. The Supreme Court in
- 15:47Texas actually looked at some proposed rules
- 15:49and ultimately decided that for most of the proposed
- 15:53rules, they really weren't going to make a lot
- 15:54of changes. Most of this is really the core idea
- 15:59of We have a duty as attorneys and as advocates
- 16:03and as litigants to make sure that what we file
- 16:07with the court has been verified, that the cases
- 16:09are real, that the statutes exist, that we aren't
- 16:12exaggerating or fabricating what they stand for,
- 16:15and that the facts, as far as we know, are accurately
- 16:20represented. So that's become easier to mess
- 16:24up in the world of generative AI. And we're going
- 16:26to talk about some ways to try to avoid that.
- 16:29But that duty at its core has always existed.
- 16:33It's just now we have, you know, attorneys using
- 16:35tools that maybe they're not familiar enough
- 16:37with and seem to be making mistakes. Someone
- 16:40seems to pop into the news every month or so
- 16:42having filed something with the court that's
- 16:44hallucinated. So we're going to talk about some
- 16:46ways to avoid that as well. We know generative
- 16:49AI use is accelerating more and more firms. more
- 16:53and more attorneys are using it. So we've got
- 16:55to figure out a way to do it responsibly. So
- 16:59let's talk about kind of risky versus safer uses
- 17:03of generative AI. I built this slide a couple
- 17:07months ago, actually. And I'm going to talk about
- 17:11the top bullet on the left first, which is clients
- 17:15putting case materials into free chat GPT or
- 17:18Gemini. And I'm going to say this because this
- 17:21article just came out, I think it was yesterday
- 17:23or day before yesterday, an expert witness in
- 17:26a case involving 3M was exposed for putting into
- 17:31chat GPT, he put in a lot of stuff, but one of
- 17:35the things that came out was the headline was,
- 17:38by the way, this was a $61 million lawsuit over
- 17:42an explosion that resulted in, I think, three
- 17:45fatalities. And the expert in preparing the expert's
- 17:49report actually put in, show how 3M is 0 % at
- 17:55fault for the explosion at Watson Grinding. So
- 18:00that and some of the other prompts this expert
- 18:04used to build the expert's report were pretty
- 18:07telling. He was really just trying to get the
- 18:10system to help him build with the conclusion
- 18:14in mind. right, without necessarily doing the
- 18:17steps and analysis. Just, hey, here's the conclusion
- 18:20I want. Help me build an argument and build a
- 18:23report that supports the conclusion. It's interesting
- 18:27because I could see attorneys asking those same
- 18:30types of questions and that being considered
- 18:32attorney work product. And I think what continues
- 18:35to get some of these people in trouble, whether
- 18:39it's a client putting case materials into free
- 18:41chat GPT, or whether it's an expert witness,
- 18:44is a lot of them are not on paid enterprise or
- 18:48business use case licenses of these products.
- 18:50They're using the free version. And when you
- 18:52use the free version, and you don't configure
- 18:54your privacy settings properly, you are potentially
- 18:57exposing it to the public, essentially, because
- 19:01it can be the models could be trained on what
- 19:03you're putting in. Hopefully, hopefully you all
- 19:06have heard that enough by now that any tool you
- 19:09are using is a paid tool preferably a legal tech
- 19:12specific tool and you have looked at the privacy
- 19:15policies and you've opted out of training if
- 19:18that is even an option at all. Many of the good
- 19:21legal tech companies don't even have that as
- 19:23an option. So just know that with your clients
- 19:26and with your experts we have to start having
- 19:29some of these conversations. So let's get into
- 19:33what I promised with your depositions. And I
- 19:37know I have questions coming in. I'm going to
- 19:39do the best I can to tackle some of those questions
- 19:43as we go. And so I'm going to mark a few of them
- 19:47now as answered. I got a really interesting question,
- 19:51actually. I am going to address this one because
- 19:53I find it fascinating. So how do you make sure
- 19:55the testifying person is real and not AI generated?
- 20:01I'm assuming that probably means we're at a virtual
- 20:03deposition, which a lot of depositions are remote
- 20:08now. Scribe utilizes Zoom, also does in -person
- 20:11depositions. But I know many of us do Zoom depositions.
- 20:15I think the best way to answer that is The digital
- 20:20reporter or if you're signing onto a zoom deposition
- 20:22with a traditional reporter stenographer are
- 20:25usually checking the person's identity by either
- 20:29verifying the last four of their social or the
- 20:32driver's license or whatever the identity verification
- 20:34process is that your reporter is using pursuant
- 20:37to the notary rules before they sort that witness
- 20:40in. At least that is how scribe does it. Could
- 20:44you fake even that? I don't know. I think no
- 20:47one's immune necessarily to that. Think about
- 20:50your average deposition experience with any court
- 20:53reporting firm. Again, most of these are just
- 20:55on Zoom or Teams or something else like that.
- 20:59Would it be possible to show up with an AI avatar
- 21:01and fake the entire thing? That's kind of a scary
- 21:05thought, but it is a very interesting question.
- 21:07So, and I see another question, does that mean
- 21:10no human notary? No, you definitely want professionals
- 21:13in the loop. Okay, I think that is extremely
- 21:15important that we have certified notaries, qualified
- 21:19digital reporters, human beings in the loop,
- 21:22end to end, to make sure that testimony is captured,
- 21:26redundancy is working, real time is working,
- 21:29and that you get a record that you can. rely
- 21:31on. The experts that I mentioned in that 3M case,
- 21:35again, I'm taking a few questions before we dive
- 21:37into the next section. That came to light in
- 21:40a deposition. The expert was questioned about
- 21:42it. I do not recall off the top of my head if
- 21:45they had already sent discovery requests targeting
- 21:47it prior to the deposition, but that's my understanding.
- 21:51So let me dive in here preparing for depots with
- 21:54AI. I'm going to tell just a quick story. Hopefully,
- 21:58you see the road on the screen. This is technically
- 22:02an AI -generated version, because I didn't want
- 22:04to take pictures as I was driving. But I had
- 22:07about eight depositions a couple of years ago,
- 22:09back to back to back, in Midland, Odessa. For
- 22:12those of you who've watched Friday Night Lights,
- 22:15or the latest show would be Landman. I was driving
- 22:18from Austin all the way to Midland, Odessa, which
- 22:21is a pretty long drive. And I had all these depositions
- 22:24that I needed to, some of them I was taking,
- 22:26some of them I was defending, and I really wasn't
- 22:30completely prepared. I was driving down the day
- 22:32before. I was going to have, you know, I know
- 22:35none of you have ever done that, staying up late
- 22:37the night before to get ready for all these depositions.
- 22:39And I thought to myself, you know, I know I have
- 22:42a lot of studying to do, but wouldn't it be nice
- 22:45if I could leverage the latest technology to
- 22:50just talk to me as I'm driving and basically
- 22:52tell me about all these witnesses. So that was
- 22:56the idea that popped into my head. And this was
- 22:58a couple of years ago. And the technology has
- 23:01gotten better since then. But I'm going to kind
- 23:03of show you what I came up with back then. And
- 23:05then we'll talk about it. And I'll show you kind
- 23:07of how you can try to build something similar.
- 23:11If for some reason you can't hear audio in a
- 23:14few seconds, if the host could just alert me,
- 23:18but hopefully it will work, it usually does.
- 23:20So here we go. All right, so we're diving into
- 23:22this personal injury case against Family Dollar
- 23:25today. You sent a ton of material over witness
- 23:28interviews, expert reports, legal filings. It's
- 23:31a lot. A real legal drama unfolding. Yeah, a
- 23:34lot of moving pieces for sure. Mission with this
- 23:37deep dive is to really break it all down. analyze
- 23:40those witness accounts, understand everybody's
- 23:42role and any potential biases, and then ultimately
- 23:46try to uncover the strengths and weaknesses of
- 23:48family dollar's case. Yeah, what's really striking
- 23:51is the number of people involved and then all
- 23:53the conflicting accounts of what actually went
- 23:56down. It's a real challenge to try to piece together
- 23:58the truth. That's what we're here for. But that's
- 24:00what we do. Exactly. To set the stage. All right,
- 24:04so I cut that off by design so we didn't start
- 24:07getting into all the ins and outs of my case.
- 24:09But what I will tell you is that that tool basically
- 24:14gave me an hour or more, can't remember exactly
- 24:17how long it was, of in -depth discussion between
- 24:20two virtual AI -generated podcast hosts that
- 24:26were talking about the ins and outs of my case.
- 24:29all the conflicting witness reports, a little
- 24:32bit about each witness. Really, a lot of the
- 24:36facts that I needed to study that night, it had
- 24:39already kind of digested and even came up with
- 24:42kind of themes or questions. And it was really
- 24:44fascinating. I actually thought not only could
- 24:47it be a great way to get ready for a deposition,
- 24:49but I thought it would be a really good tool
- 24:51for getting ready to go to trial. Because the
- 24:54way that these podcast hosts were talking about
- 24:57the issues really reminded me of kind of how
- 25:00I would expect a lay jury to talk about you know,
- 25:05issues in a case and witnesses, and it talked
- 25:07about a perceived love triangle, because again,
- 25:10there was all sorts of big pieces to this case,
- 25:13but there were some real work relationships that
- 25:16were interesting in that one. So let me explain
- 25:20how I did it. And so the tool that I use to create
- 25:24that is one you've probably heard it by now,
- 25:26which is Google Notebook LM. and Google Notebook
- 25:30LM's very inexpensive tool. Last time I checked,
- 25:32you can get the paid version for like $25 a month,
- 25:35maybe even less. And what's interesting is, though
- 25:39at least it's advertised, again, you can vet
- 25:42the privacy policy and kind of their terms and
- 25:44conditions yourself, but the way that it's advertised
- 25:47is it's only grounding its answers in what you
- 25:50put in the system. It's not trying to go out
- 25:54and necessarily find answers online or something
- 25:56like that. So you can create all sorts of things,
- 25:59study guides, you can create presentations, you
- 26:03can create deposition outlines, all types of
- 26:07things based on materials that you put into this
- 26:10type of tool. So we're going to go through and
- 26:11I'm going to show you kind of what does that
- 26:13even look like. And so we will use a case that
- 26:16I'm sure none of you have ever heard of. Now
- 26:18I'm sure many of you have. heard about it and
- 26:20maybe followed along, which was the Amber Heard
- 26:23and Johnny Depp dispute. And so here's what we're
- 26:27going to do. I've abbreviated this a little bit
- 26:29for this presentation, but we're going to take
- 26:32some of the transcripts. And I realize the text
- 26:34is tiny. Trust me, you don't need to be able
- 26:37to read the slide right now. I'm going to explain
- 26:39what's on there. But what we've done is we've
- 26:42taken all of the transcripts from that first
- 26:46case. There were a couple of cases in that dispute.
- 26:50And we have fed them in to Google Notebook LM.
- 26:53And we have asked it to prepare us to get ready
- 26:57to take some testimony ourselves based on the
- 27:00content of all of these other transcripts. So
- 27:03you've got Input on the far left, the files that
- 27:07we uploaded into Google Notebook LM. You just
- 27:10go to notebooklm .google .com and create an account.
- 27:13We have some of the text that it generated in
- 27:15the middle. And then we've got all the things
- 27:17it can do over here on the right. So if I zoom
- 27:20in a little bit, let me see if I can get it to
- 27:21zoom in. These are some of the things that it
- 27:25can do potentially for you. Create an audio overview.
- 27:29ports, infographics, mind mapping, all kinds
- 27:32of different really cool things that it can do
- 27:34in just the click of a button. So we're going
- 27:36to do the audio overview. So let's see what it
- 27:39sounds like based on what we put in. If you're
- 27:47looking at your news feed lately and feeling
- 27:50completely overwhelmed by just the constant barrage
- 27:53of media spin the tabloid headlines right the
- 27:56hot takes the think pieces if you're feeling
- 27:58that you are definitely not alone it's exhausting
- 28:01it really is it seems like every high profile
- 28:03event is instantly filtered through i don't know
- 28:06a dozen different biases before it ever even
- 28:08reaches us so today we are setting all of that
- 28:11secondary noise completely aside going straight
- 28:14to the source straight to the source We have
- 28:17a highly specific mission for this deep dive.
- 28:20We're looking directly at the raw primary source
- 28:24documents from one of the most intensely publicized
- 28:27and scrutinized legal battles in recent history.
- 28:31The litigation between. All right, so that gives
- 28:34you kind of a glimpse of these two again, virtual
- 28:37AI created podcast hosts about to talk about
- 28:39the ins and outs of this case. And if we skip
- 28:42ahead. I'm going to show you kind of the next
- 28:45segment of this. What I thought was really interesting
- 28:47is it began to compare and contrast the varying
- 28:50accounts of what happened. So let me fast forward
- 28:53a little bit and show you what I mean by that.
- 29:11or possibly jammed okay according to his account
- 29:15she then kicked or pushed the door open slamming
- 29:18it into his head he claims that he then stood
- 29:21up and she clocked him in the jaw that is the
- 29:24sequential narrative he argues on the tape then
- 29:27we have her narrative she claims she was actually
- 29:30trying to close the door and in the process it
- 29:33accidentally scraped his toes according to her
- 29:36account he then pushed or shoved the door back
- 29:40All right, so two conflicting accounts of what
- 29:43happened. Let's continue to kind of go through
- 29:46that. We talked a little bit about some of the
- 29:48other things it can do. So I've already mentioned
- 29:49these. I won't spend too much time on this. But
- 29:52one that fascinated me when I put in this Johnny
- 29:55Depp Amber Heard content was it actually could
- 29:58build slides and infographics. Those are the
- 30:00ones that really jumped out to me as pretty clever.
- 30:03And so here's an example. I'm not going to spend
- 30:06too much time. I just want to give you a flavor
- 30:07of what it can create. And you're probably like,
- 30:12well, would anyone really use this in a legal
- 30:14practice? Yes, they will. And I'll tell you how
- 30:16I know that. I've actually gotten a video demand
- 30:20letter in one of my firm's personal injury cases,
- 30:23I'm on the defense side, that was created. using
- 30:26Google Notebook L .M. They fed in the case materials,
- 30:31kind of the story of why they believe their client
- 30:34was extremely sympathetic, why they think they're
- 30:36going to win the case and win gazillions of dollars.
- 30:39And Notebook L .M. put together slides and a
- 30:42video demand. And it was... it was actually pretty
- 30:46clever. And I will say it really came across
- 30:49extremely professional and persuasive. So I think
- 30:54people are already using this tool. And that's
- 30:57just one example that I experienced firsthand
- 30:59in one of my cases. Let me show you what it put
- 31:00together in this Johnny Depp, Amber Heard case.
- 31:03And I may pause on one of them that I thought
- 31:05was particularly. Keep in mind, I didn't tell
- 31:08it how to design any of this. It just kind of
- 31:11proactively went out and started grabbing pieces
- 31:13that would have been uploaded and started creating
- 31:16different types of content based on that. And
- 31:19the one, I'm going to pause on this one. This
- 31:21is the one that I thought was very well done.
- 31:25And part of why I think it's so well done is
- 31:27because it really laid things out in black and
- 31:29white, literally and figuratively, by comparing
- 31:32and contrasting two different stories. by the
- 31:36two witnesses. Here is what Johnny Depp's alleged
- 31:40reality was based on what he said happened versus
- 31:44amber herds and it was really interesting to
- 31:46see those two things kind of put right side by
- 31:49side and the two claims contrasted that way.
- 31:52So this is just one example of kind of what an
- 31:56AI tool can help you do even as you begin to
- 31:58get ready to either take cases to mediation where
- 32:01sometimes infographics and slides can be impactful
- 32:05or take it to final hearing or trial and you
- 32:08need to put together some type of demonstratives.
- 32:10These tools can can do some pretty amazing things
- 32:12for you. All right. We're going to keep going.
- 32:16More about depo prep. And so I use Claude a lot.
- 32:20I also use a legal tech specific solution built
- 32:23by Clio called Clio Work, sometimes called Vincent.
- 32:27I don't know which one they're ultimately going
- 32:29to settle on. They're very similar, seem to overlap
- 32:31to me. Clio's tool, for those of you who may
- 32:34use Clio as your practice management system,
- 32:36Clio's tool is built on OpenAI's latest version
- 32:40of ChatGPT. Um, as of the date of this presentation,
- 32:44I believe that's GPT 5 .6 sole. If you're keeping
- 32:48track, it's hard to keep track because things
- 32:50keep improving. Um, you know, certainly every
- 32:53month, if not every week. So how can you let
- 32:56me also say what I'm going to talk about can.
- 33:00Most of this stuff can be used with any of these
- 33:03tools. They're all very, very similar. So which
- 33:06one you pick, which one your firm authorizes
- 33:08you to use is, of course, ultimately up to you
- 33:11and your firm. But the concepts and the way that
- 33:13you work with them is going to be very, very
- 33:15similar, regardless of which tool you're using.
- 33:18So we're going to use Claude, and we're going
- 33:21to use it in a... Well, I'll explain in a minute.
- 33:27This is a real case. This is a fictitious image.
- 33:30Obviously, I didn't actually hire Dr. Strange
- 33:33or Iron Man to help me. But it is based on a
- 33:36real case. It's based on a forklift injury case
- 33:38that I was defending for one of my clients. And
- 33:40it really did become a battle of the experts.
- 33:43And so I used AI to put together this image for
- 33:46this presentation. So how can you use an AI tool
- 33:50to help you get ready for depositions? I think
- 33:52expert depositions are a great use case. And
- 33:55here's why. Because you typically are going to
- 33:58have the benefit before the deposition of the
- 34:01report, depending on the type of work you do.
- 34:03If you do personal injury like me, and it's non
- 34:06-retained experts, you're probably going to have
- 34:07the records, right? So think medical records,
- 34:10other types of records that they relied on in
- 34:13reaching their conclusions and opinions. And
- 34:15so once you get all that content, and you make
- 34:18sure you're using a tool that protects the privacy
- 34:21of whatever you put in, which is important, then
- 34:24you can do some pretty cool things. And so there's,
- 34:26again, a couple different ways to work with this.
- 34:28I'm going to show you chat interface. Just know
- 34:32that with Claude Cowork and other options, again,
- 34:37Grokbot just launched last week. I don't know
- 34:40if you've heard about it yet. There are more
- 34:42and more agentic options, which means you're
- 34:46kind of telling it what to do. Here's the source
- 34:49material in this folder or in this API. Just
- 34:52go do it for me. If you're just dipping your
- 34:57toe in the water, I would start with the chat
- 34:59experience first. It's a very traditional approach
- 35:03to AI. I guess we call it traditional at this
- 35:05point as fast as things are moving. And you also
- 35:08get the benefit of kind of back and forth and
- 35:10back and forth discussion as you're chatting,
- 35:12as you're refining, instead of just delegating
- 35:15everything to some type of AI agent to go and
- 35:18do it for you. So. We're going to tell it, hey,
- 35:22take the content of these expert reports and
- 35:25help me depose the other side's expert. And I'm
- 35:29a big fan of being specific. I think you got
- 35:31to really tell the tool, how do you want your
- 35:34outline organized? What topics do you want it
- 35:36to cover? I really think about prompting the
- 35:38way I think about assigning a task. So if I was
- 35:41sitting down with a senior associate or a paralegal,
- 35:44and they were going to help me put together an
- 35:46outline or whatever I'm getting ready for, That's
- 35:50the level of detail I'm looking to give the system
- 35:53so that the system can give me back work product
- 35:57that I actually want to use. So we're going to
- 35:59do that on this one. I've kind of got this preset
- 36:04up so we can talk through it. We're telling Claude,
- 36:08here's how I want you to organize my outline.
- 36:11Of course, I want to undermine the credibility
- 36:13and the foundation of the claimant's expert,
- 36:15I want to expose any flaws. If at all possible,
- 36:18I want you to leverage my expert's opinions and,
- 36:21you know, the way that my expert is thinking
- 36:23through. And so, you know, again, within 30 seconds
- 36:26to a minute, it spits out a first draft. Again,
- 36:30key words there, first draft of a deposition
- 36:34outline. Again, this, you know, when I first
- 36:36started presenting on AI a couple of years ago,
- 36:38this was maybe a novel use case or kind of people
- 36:42go, wow, I can't believe it can draft an outline.
- 36:44I think that's hopefully in your practice is
- 36:47becoming more and more commonplace to use generative
- 36:49AI to help you draft. And if it's not, you really
- 36:52should. It's probably one of the easier things
- 36:55that AI can do. And as you read through samples,
- 36:58Or if you read through outputs and you think,
- 37:01gosh, AI has missed the mark on this, or it feels
- 37:04a little light on this issue or that issue, just
- 37:07tell it. That's the beauty of these tools. You
- 37:10can critique it in real time. You can tell it
- 37:12to go deeper on this issue or that issue. If
- 37:16it misses something or... comes at it from the
- 37:18wrong angle. You can redirect it. And so the
- 37:21more you work with it, the better it's going
- 37:23to do. And the memory on these tools is improving
- 37:27dramatically. What I mean by memory? I mean,
- 37:30literally, if you tell it your styles, your preferences,
- 37:32how you handle certain issues, it is starting
- 37:35to get much better at remembering that in the
- 37:38future if a similar issue comes up. That's the
- 37:40layman's explanation. So depo prep. Okay, hopefully
- 37:46none of you have had a witness actually turn
- 37:48green in a deposition, but this is another fun
- 37:52AI -generated Incredible Hulk, this one, you
- 37:54know, Mark Ruffalo. So what I want to talk about
- 38:00here is this is a deposition I was getting ready
- 38:03for where I needed to prepare this forklift driver
- 38:07to be deposed. And one of the things that I think
- 38:10maybe sometimes we forget, especially attorneys
- 38:12that have been doing this for a fair amount of
- 38:14time, is a lot of times these witnesses have
- 38:17never been deposed before. And they didn't go
- 38:19to law school. And they don't necessarily speak
- 38:21the exact same language and technical jargon
- 38:25that you and I do. They don't even necessarily
- 38:27remember what a plaintiff is versus a defendant
- 38:29versus a respondent or a claimant. What I will
- 38:33often do when I'm using an AI tool to help me
- 38:36prep a witness and build a prep guide for me
- 38:39is I will tell it what the education level is
- 38:44of my witness, what the background is, right?
- 38:47Where have they worked? What have they done?
- 38:49What type of schooling do they have? And then
- 38:51the AI will tailor your outline to your audience,
- 38:56to the person that you're prepping. So in this
- 38:58example, We're going to have it draft based on
- 39:01the content that I've given it. I've uploaded
- 39:03copies of the pleadings. I've uploaded copies
- 39:05of key facts. But we're going to have it draft
- 39:07a prep guide. And all the names have been changed
- 39:10to Marvel. My kids like Marvel, so it's fun.
- 39:15But it's giving me a guide that I then either
- 39:18print or create a PDF version of to walk through
- 39:22sample Q &A, key points or themes, possible traps,
- 39:27all based on kind of the content that I gave
- 39:30it and anticipated maybe areas where I think
- 39:34the witness could get tripped up. Many of us
- 39:38have prep guides for witnesses. Many of our firms
- 39:40have kind of deposition prep 101 for a lay witness,
- 39:43for a corporate rep, whatever the issue is. The
- 39:46nice thing about using generative AI to create
- 39:48a prep guide is you can customize it so fast,
- 39:51right? You can get it to your facts, your witness,
- 39:56framed a lot of the same concepts that many of
- 39:58us have learned, whether it's through NEDA or
- 40:00our own law firm training, a lot of the same
- 40:02concepts, but all customized and kind of purpose
- 40:06built. for your case and your witness. And I
- 40:08think it'll resonate a lot, lot better. So here's
- 40:12an example of actually using Scribe to do that.
- 40:15So this is going to be after one of the depositions
- 40:17has already happened. And I'm going to show you
- 40:19what I mean by that in just a minute. But here's
- 40:21the prompt. Here's the concept. We're going to
- 40:23say, hey, look, plaintiff's been deposed. Based
- 40:26on this plaintiff's depo, I want you to identify
- 40:29topics that may be probed in a corporate rep
- 40:31depo. 30b6 depot and there are areas where the
- 40:35plaintiff's testimony is going to create potential
- 40:37traps for my witness. And if so, how should I
- 40:40prep them? So here's kind of what it could look
- 40:42like in practice. So now you're logged in. Okay,
- 40:45this is just through the web browser on Google
- 40:47Chrome. And we're going to go into this deposition
- 40:51of Ms. Mitsubar that's already happened. And
- 40:56so we're going to talk, we're going to click
- 40:57into it. We have our let me kind of explain the
- 41:03layout real quick. So we have our transcript
- 41:06on the left Rough transcript for the first week
- 41:09and then after a week the final drops in We have
- 41:12our video pane in the middle with the video from
- 41:15the deposition clips if we've created any video
- 41:18clips Playlist if we've created any playlist
- 41:21of video clips, right kind of highlight reels
- 41:23multiple clips stitched together And then files
- 41:26and if there were any exhibits there'd be an
- 41:28exhibits tab right here And then we have an AI
- 41:31panel off on the right. When I mentioned optional
- 41:34AI tools, this is where those optional AI tools
- 41:37live. It's over on the right -hand side. If you
- 41:40don't want to see them, you don't want to use
- 41:41them, you can collapse it. But you probably do.
- 41:44It's probably one of the reasons you attended
- 41:45this presentation was to learn how to leverage
- 41:48AI. Today's date is November 21, 2020. So we're
- 41:53going to drop the prompt in that we talked about.
- 41:56and we're going to run the AI on the right -hand
- 41:59panel. So here it is. We've asked it the question.
- 42:06It's going to show you as it thinks about the
- 42:09question that we asked, and then it's going to
- 42:12start populating the answer to the question.
- 42:16could get probed what some this was a med mal
- 42:19hypothetical deposition and so it's going through
- 42:23all the different topics identifying based on
- 42:25what this plaintiff zoomed in to make it a little
- 42:28easier for to read based on what this witness
- 42:31said here are the key documents that are important
- 42:33here are the key issues that are important here
- 42:35are things that you need to worry about potential
- 42:38traps for corporate rep based on what was testified
- 42:41to really all very very specific to the content
- 42:45from this deposition. So it's a very easy and
- 42:49fast way to work with the tools. There's a question,
- 42:53does it cost any more for these optional features?
- 42:56No, it doesn't. Not with Scribe. I can't speak
- 42:59to what other digital reporting or court reporting
- 43:01firms charge, but with Scribe, everything is
- 43:04included for the base price. Which is nice for
- 43:08those of us who run firms or manage firms, you
- 43:11don't have to worry about a subscription fee.
- 43:12This is not subscription fee to death for all
- 43:15these AI tools. It's just a part of the deposition
- 43:18cost, which for many of us is a pass through
- 43:20expense anyway. Let's keep going. Good question.
- 43:25Glad I was able to notice it. Some things to
- 43:28look for, okay? I'm gonna make a few recommendations.
- 43:31Again, as of the date of this presentation, I
- 43:34keep weaving that in because the world of AI
- 43:36is changing so fast, right? You blink your eye
- 43:40and you've already got a new model for Claude
- 43:43or for Clio or for OpenAI or whatever. But here's
- 43:47what I recommend currently. I would not put your
- 43:50entire case file. into the deposition system.
- 43:54You're going to want to pick out what's relevant.
- 43:55You're going to want to pick out the materials
- 43:57that you would pick out back in the old days
- 43:59when you would print it to read it for depo prep.
- 44:03I would be very hesitant to just drop an entire
- 44:08case file into any AI system. You want to go
- 44:11through it. You want to understand what's in
- 44:12the file. You want to try to figure out what
- 44:14is really pertinent. to the deposition that I'm
- 44:17getting ready to take. And so I always say garbage
- 44:20in, garbage out. You want to be specific with
- 44:22your questions. You want to try to cultivate
- 44:24the relevant information as best you can. That's
- 44:27not to say there couldn't be other tools out
- 44:29there that can help you narrow it down. Of course
- 44:32there are. I mean, there have been, for those
- 44:34of you who do really document intensive work,
- 44:36I mean, there's been eDiscovery. platforms forever.
- 44:39So I realize some of you may have very document
- 44:41intensive cases, but for your average case, your
- 44:45average deposition. I would not drop your entire
- 44:48case file in. I would try to pick a subset of
- 44:51information that's pertinent to the witness.
- 44:54You also, of course, you want to check any AI
- 44:56tool you're working with. You want to check for
- 44:58the source of truth. What I mean by that is,
- 45:00if it's summarizing something, if it's citing
- 45:02testimony, make sure that source of truth is
- 45:05one click away and that you can get to the source
- 45:07material and validate it. And keep in mind, at
- 45:11the end of the day, it's your deposition, whether
- 45:13you're taking it or you're defending it. you
- 45:16need to be in the driver's seat. Yes, AI can
- 45:20be extremely helpful, and we're going to show
- 45:21more ways that AI can help, but you've got to
- 45:24have the judgment to decide what actually gets
- 45:27asked, what goes into the outline or doesn't
- 45:29go into the outline. And for those of you who
- 45:32may be more junior, such as junior associates,
- 45:35if you don't know, if you don't have the human
- 45:37judgment because you just haven't lived it enough
- 45:39yet, Don't rely blindly on the AI. Ask a senior
- 45:43lawyer. Ask a partner. Ask someone who's been
- 45:46there before, because the AI is not always 100
- 45:49% accurate. So if you want some sample prompt
- 45:54guides or if you want the sample engagement letter
- 45:57or any of the stuff that I mentioned earlier,
- 45:59I put my scribe email on the screen because it's
- 46:01shorter, honestly. Carl with a K. at scribewithak
- 46:04.ai. I'm happy to share any of it or if you have
- 46:08follow -up questions, I'll do the best I can
- 46:10to answer it. Again, I genuinely enjoy this stuff
- 46:13and I'm trying to stay caught up on all the latest
- 46:17developments myself because things are moving
- 46:19fast. So before you take a deposition now, one
- 46:25of the things that I want to talk about is You
- 46:27can start uploading before the before the deposition
- 46:30goes live. You can start uploading the material
- 46:33that you're going to want to analyze with AI
- 46:35against the live testimony before the deposition
- 46:38starts. And really that's key because you don't
- 46:41I don't think you want to be hassling with uploading
- 46:45exhibits and your outline and prior transcripts
- 46:48all kind of like as the depositions taking place.
- 46:51To me that would feel a little stressful. So
- 46:53you can log in. you know, an hour before, a day
- 46:56before, and you can start or your paralegal can
- 47:00start uploading some of the key information into
- 47:03the system before the deposition ever takes place.
- 47:06Here's kind of what that looks like in the Scribe
- 47:08system. There's only, I think, a couple tools
- 47:11in the market that I'm aware of that do this
- 47:13type of stuff, Scribe being one of them. In the
- 47:16scribe system, you just log into the matter.
- 47:18So here we just have this one called Carl and
- 47:20Laura. But if this were a deposition, it would
- 47:22say deposition of witness name. And once you've
- 47:26logged in, you just go to this file section and
- 47:29you upload the files. You upload those prior
- 47:32transcripts, maybe taken with other court reporting
- 47:34companies. You upload potential exhibits. You
- 47:37upload a copy of your outline. Why are you doing
- 47:40that? I'm going to show you in just a minute.
- 47:43So that gets us to the second part. working the
- 47:46record live, right? We're trying to get as much
- 47:48as we can out of the AI tool, the software tool
- 47:53to make sure that we take the best deposition
- 47:54we can possibly take. So how are we going to
- 47:57do that? And why did we upload those things?
- 47:59Well, it's because in my opinion, the world of
- 48:03court reporting is changing very, very fast.
- 48:05And what used to sound be very, very expensive
- 48:08or sound impossible is now possible. So you should
- 48:11get a real -time transcript with every deposition
- 48:14you take for no extra cost. You should find a
- 48:17tool that can help you leverage AI to analyze
- 48:20the real -time transcript against key documents
- 48:24in the case against your outline to make sure
- 48:26that you don't miss things. All of those things
- 48:28are possible now. It really is like having a
- 48:31second set of eyes or a superpower to be able
- 48:33to have an assistant or a second chair watch
- 48:37the deposition unfold and cross -reference that
- 48:40testimony against the content that you've uploaded.
- 48:43And I'm going to show you what that looks like
- 48:44in just a minute. I want to show you what the
- 48:47real -time, again, there's various real -time
- 48:50experiences out there. I want to show you what
- 48:51one looks like. So this is an example. Launch
- 48:53the real -time transcript and watch it update.
- 48:57It updates every few words and continues until
- 49:03the event ends. It streams the ASR generated
- 49:06text as the event happens. Keyword search. Change
- 49:11font size. Chat with the transcript when you
- 49:15need it. get your questions answered, and finish
- 49:20your depot strong. All right. I was trying to
- 49:25answer just a couple of those questions. Thank
- 49:26you for the questions. They do make it fun. If
- 49:30it's a really important question, I'll read it
- 49:32out loud. Some of those were just follow -up
- 49:35comments. Again, we talked about real -time on
- 49:40the left, AI panel on the right. and that can
- 49:44be available to you as you're taking a deposition.
- 49:47So what can you do with that? And so here's one
- 49:50thing that I like to do and that I train my team
- 49:53on, which is make sure you don't miss anything
- 49:57from your outline. I mean, I don't care how many
- 50:00depositions you've taken. We've all had the experience
- 50:03of leaving a deposition, driving home from a
- 50:05deposition. And one of two things can happen
- 50:08to any of us. We either forgot to ask that one
- 50:12extra question on whatever the topic was, or,
- 50:15and I've had this happen to me. we could have
- 50:18swore we got the testimony we like it came out
- 50:21a certain way and we thought we got the the witness
- 50:24to say X and we needed that you know in support
- 50:27of you know whatever motion we're getting ready
- 50:29to file and then once we get that transcript
- 50:32back weeks later the old -fashioned way of doing
- 50:35it we realize it didn't come out the way I thought
- 50:39it did. It's kind of wishful thinking on my part.
- 50:43And so with AI tools and using them in real time,
- 50:45you can kind of try to make sure that doesn't
- 50:47happen. If not, eliminate it, at least reduce
- 50:50the likelihood of it happening. So let's look
- 50:53at that. Let's drop a prompt in. And we're going
- 50:56to base it off an outline and just ask, were
- 50:58there any areas where this witness was evasive?
- 51:02And if so, suggest some follow -up questions.
- 51:07And actually, excuse me, we're not going to do
- 51:09the outline example yet. I'll do that in a minute.
- 51:11So let me kind of zoom in on this. We're just
- 51:15going to use AI to try to issue spot for us without
- 51:18any additional documents right now. Just has
- 51:21the witness been evasive? And if so, what should
- 51:23I ask to follow up with the witness? And so what
- 51:26the AI is doing, let me kind of pause for just
- 51:29a minute if I can figure out how to do that.
- 51:31Hold on. Let me zoom out to pause because let
- 51:34me explain what's happening. OK. Let me explain
- 51:37what's happening here. At the time that we hit
- 51:40submit, the entire real -time transcript is being
- 51:44analyzed top to bottom using the latest generative
- 51:47AI tool in a secure way, private way, and it
- 51:51is analyzing it for the question that we post.
- 51:56So again, it's doing something that a human could
- 51:58do. It would just take a lot longer to read the
- 52:00entire thing top to bottom and to critically
- 52:02think through. Are there any areas where this
- 52:04witness has been evasive or failed to answer
- 52:07a question? So it's doing it kind of in a split
- 52:10second, which would otherwise take most of us
- 52:13hours to do. So let me kind of zoom back in and
- 52:16we'll. Oh, this tool is a little finicky, but
- 52:20so it's analyzing the key issues. And one of
- 52:23the things that's really cool here is it's also
- 52:25suggesting some follow up questions and identifying
- 52:28questions that maybe I haven't even asked. that
- 52:31I want to think about asking. So again, it's
- 52:35you ultimately have to use your judgment on whether
- 52:38you want to follow up and ask the questions and
- 52:41cover the topics that it's identified. But it
- 52:44is extremely powerful way to get issues spotted
- 52:47for you. So let's keep going and let's talk about
- 52:51some of you may be thinking, oh my gosh. This
- 52:55is so much information. How am I going to remember
- 52:57this? How am I going to do this? I've already
- 52:59got to focus on the witness and my exhibits.
- 53:01Do I really need one more thing to focus on?
- 53:04I hear you. And that's one of the reasons that
- 53:06the Scribe team put together what I call the
- 53:09easy button. It's not technically what it's called.
- 53:11But the goal was, what if the AI could look at
- 53:15these issues and monitor them for you? So all
- 53:18you have to do is click a button. And so let
- 53:20me show you what that looks like with this example.
- 53:23So again, real time on the left, it's streaming
- 53:26out as the deposition is taking place, AI panel
- 53:29on the right. And now we have a new panel called
- 53:32Insights. And so if you click the button at the
- 53:35bottom, which is currently off screen because
- 53:36I'm zooming in, the insights begin to generate.
- 53:41And the insights are going to fall into three
- 53:43buckets, inconsistencies, follow -up items, and
- 53:47unanswered questions. and it's done this for
- 53:50you. It's divided this by category with you just
- 53:53clicking one button and it's identifying the
- 53:55issue. The other thing, let me zoom out briefly
- 53:58to pause because I can't figure out how to pause
- 54:00when I'm zoomed in. I'm going to zoom back in
- 54:03if I can. One of the things I want to point out
- 54:05here is it's going to hyperlink you back to the
- 54:10real -time transcript. So as it summarizes what
- 54:13it thinks the witness said, the issue that it's
- 54:17spotted for you, it's going to link you back
- 54:20to that real -time text so that you can go see
- 54:23for yourself the section of testimony that it
- 54:26has spotted an issue in. Then, of course, it's
- 54:28going to include some suggested questions as
- 54:30well. Again, that source of truth always being
- 54:33one click away is a core concept that I try to
- 54:36keep in mind with any AI tool that I'm using.
- 54:41And then unanswered questions, it's going to
- 54:43spot things where maybe they haven't been answered
- 54:45at all. And then it's going to suggest some follow
- 54:49-up questions for you to get those answered as
- 54:51well. All right, and then I'm going to speed
- 54:55things up just a little bit in the interest of
- 54:56time. I know some of you may have a hard stop,
- 54:58so I'm going to cover a few of these concepts
- 54:59a little quicker. Why were we uploading all that
- 55:02stuff earlier, right, when we talked about pre
- 55:04-event uploading? Let me show you why. So what
- 55:06we can do if we upload the content in advance,
- 55:10and I'm just going to do it in real time, but
- 55:12pretend you did it before. We're going to upload
- 55:14an outline. Best is to do it before the event.
- 55:18But now when we ask our question, the large language
- 55:23model is going to have the benefit of the document
- 55:26plus the question plus the transcript. So here,
- 55:30we're just asking it, did I cover everything
- 55:32from my outline? And if not, is there anything
- 55:34I should follow up on? And so it's going to analyze
- 55:38that issue. And I think I must have hit pause.
- 55:41Hold on. Sorry about that. It's going to analyze
- 55:45that issue, and it's going to very quickly All
- 55:49right, here we resume now. It's going to very
- 55:51quickly go through and confirm, based on your
- 55:54outline, here's what you've covered and here
- 55:57are some things that maybe you missed or you
- 56:00haven't explicitly covered. And it's going to
- 56:02identify them by topic and give me sample follow
- 56:06-up questions to ask. So again, it's all designed
- 56:10kind of in the attorney perspective in mind.
- 56:15All right. Let me see if I can advance the slide.
- 56:20So I want to go in the interest of time. I want
- 56:24to talk post event. So what are some things you
- 56:25can do after this deposition has wrapped up?
- 56:29And so here, one of the things I love about digital
- 56:32reporting is. you don't just walk out with a
- 56:35yellow pad anymore. You walk out of a deposition
- 56:37and you have a full transcript synced to the
- 56:39video that you can keyword search, you can leverage
- 56:42AI against it. It just is such a better way to
- 56:45work with testimony. And so here's an example
- 56:48of what you can do, right? Deposition ends. Let's
- 56:51say you want to spot some gaps in discovery based
- 56:53on the witness's testimony. Here's the prompt,
- 56:56okay? And again, happy to share prompts with
- 56:58you if you want them. But I'm just going to ask
- 57:02it something simple. Look at this plaintiff's
- 57:04depot and identify any potential gaps in discovery,
- 57:06and if you see any, draft some sample RFPs, ROGs,
- 57:10and RFAs for me to bolster my client's case.
- 57:14You may be asking yourself, well, how's it going
- 57:16to know what discovery has already been sent?
- 57:18That would be an example of uploading a copy
- 57:21of the discovery that's already been sent. Right?
- 57:23You don't have to necessarily even upload all
- 57:25the responses. You could. You could just upload
- 57:27the request to see what new issues have come
- 57:30up in this deposition that haven't been covered
- 57:32by my prior discovery. And you could do that
- 57:34by uploading it. There's a little paperclip button.
- 57:37Let me see if I can zoom in on it. So there's
- 57:40a little paperclip right here. You could attach
- 57:42the discovery that you've sent. And then you
- 57:45could prompt it, just like we're about to do,
- 57:47and say, hey, based on the attached and based
- 57:49on the testimony, Are there any gaps in discovery?
- 57:52And if so, can you draft some follow -up request
- 57:56for me? So here it is. It has identified some
- 57:59potential gaps in the slip and fall case, and
- 58:02it is suggesting follow -up discovery for me
- 58:06to send. So it's drafted it. I can then copy
- 58:10and paste it. There's a copy button at the very
- 58:12bottom where I can drop it into Microsoft Word
- 58:16and edit it. to my liking. So very simple way
- 58:19to create discovery in your cases. And then the
- 58:23other cool thing that, you know, this is honestly
- 58:25one of the reasons I even got into legal tech
- 58:27is I've always thought video is the most powerful
- 58:30form of evidence and it's underutilized. You
- 58:34know, there's pre -COVID, honestly, I think only
- 58:36about 10 % of depositions were video. Now that
- 58:40most depositions, I think it's an excess of 75
- 58:42% are either remote or remote hybrid. almost
- 58:46all of them have the capability of being videotaped.
- 58:50It just depends on, with a lot of providers,
- 58:52you have to decide. Do you want to pay double
- 58:54or triple to get it videotaped? Again, try to
- 58:57find a solution that's not charging you more
- 58:58just for someone to press a record button on
- 59:00a Zoom. So once you have video, here's how easy
- 59:05it is to work with now. You can either create
- 59:07a clip yourself just by highlighting text. I'll
- 59:10show you that. Or you can just ask AI to do it
- 59:14for you. So here's an example. Now we're post
- 59:18-event. So we've actually got a final transcript
- 59:20in the application. And we're going to ask the
- 59:23AI on the right -hand side here, how did this
- 59:26witness describe the impact on the vehicle? Was
- 59:30it heavier on the right or the left, driver's
- 59:32side or passenger side? It's going to find the
- 59:34answer. It's going to play. So was the impact
- 59:39heavier on either of those? driver side or the
- 59:42passenger side or was it even throughout the
- 59:44front part of the vehicle? I would say even.
- 59:48Okay. So it has found the answer for me to my
- 59:57question. It played the clip and then it gave
- 1:00:01me the option to save the clip for future reference.
- 1:00:04And not only did it do that, I could have done
- 1:00:06this with the rough transcript as soon as the
- 1:00:08final transcript is in the system. I'm going
- 1:00:11to zoom in on this. The page and line numbers
- 1:00:14show up. So page 33 line 13 through page 33 line
- 1:00:1916. So it's actually going to apply page and
- 1:00:22line numbers as you create your video clips.
- 1:00:26Clever implementation. Depot summaries, I'm gonna
- 1:00:30skip over that. It's very, very simple. It has
- 1:00:32a built -in summary tool that you can just click
- 1:00:34a button at the very bottom that says summarize
- 1:00:37here at the bottom, or you can just tell it the
- 1:00:39type of summary you want. I think there's so
- 1:00:41many companies out there that will happily charge
- 1:00:43you for a 20, 30, 40, 50 page summary of a deposition.
- 1:00:47I don't think anybody wants to read those. I
- 1:00:49doubt your clients want to read them. So you
- 1:00:51can build a custom executive summary in whatever
- 1:00:54format you want, just using the chat interface
- 1:00:57at the bottom or clicking the Summarize button.
- 1:01:00Scribe's default is an executive summary. And
- 1:01:03you can also, once you've created video clips,
- 1:01:07you can begin to leverage them in ways that maybe
- 1:01:10you've never really even thought about before
- 1:01:12that can make you a more impactful advocate.
- 1:01:15One is you can actually put them into your briefs.
- 1:01:18And so you can put video clips with hyperlinks
- 1:01:21into your motions in your briefs. Now, some of
- 1:01:24you may be saying, ah, not my jurisdiction. We
- 1:01:27can't put hyperlinks in anything we file. Well,
- 1:01:30you can put the full URL. You can type the URL,
- 1:01:33copy and paste the URL into a footnote where
- 1:01:37I promise you, if the briefing attorney's reading,
- 1:01:39staff attorney's reading your motion or your
- 1:01:41brief and they see that they can watch the video
- 1:01:43just by copying and pasting a URL, there's a
- 1:01:47good chance they're going to do it. Now, you
- 1:01:48should still file the transcript as well. But
- 1:01:51I've talked to a lot of judges that really find
- 1:01:55that fascinating that they can actually watch
- 1:01:58some of the key deposition clips. And you can
- 1:02:00use it for mediation statements, mediators being
- 1:02:02able to sit and watch highlight reels of key
- 1:02:06testimony from key witnesses really can help
- 1:02:09a mediator better, you know, push a party based
- 1:02:14on the performance of a witness, whether it's
- 1:02:16good for that party or bad for that party. We've
- 1:02:19also seen it put into settlement negotiations,
- 1:02:22back and forth offers and demands. And it's a
- 1:02:26very impactful way to either highlight a witness
- 1:02:28who's great for your team or highlight a witness
- 1:02:31that's really bad for your opponent. You can
- 1:02:34score your performance. There's so many ways
- 1:02:36you can use this tool. Literally, you can ask
- 1:02:38it, please rank my performance using established
- 1:02:42techniques by the National Institute for Trial
- 1:02:45Advocacy and help me identify ways I can take
- 1:02:48a better deposition at the next deposition. Just
- 1:02:50a simple prompt like that, and you can get a
- 1:02:53report out of the system that can help you take
- 1:02:56better depots. So we are unfortunately out of
- 1:03:00time. So I've got a couple more slides, but I
- 1:03:03don't think I have time to do them. So I'm going
- 1:03:05to say thank you very much for having me. I'm
- 1:03:09going to take a look to see if I have any questions
- 1:03:11I have not answered yet. Let me see if I can
- 1:03:14find that panel. There's a question of can you
- 1:03:20use scribe in a deposition that's being transcribed
- 1:03:23by a court reporter for the attorney's own use
- 1:03:25to analyze testimony in real time? Yes, the answer
- 1:03:28is yes, you can. Feel free to reach out to me
- 1:03:30and I can connect you with someone who can explain
- 1:03:32that. And then there's another question about
- 1:03:35a competitor court reporting company. You know,
- 1:03:39all it will say is I am familiar with that company.
- 1:03:41They still do traditional court reporting, but
- 1:03:44they do have, I would say, one of the more modern
- 1:03:46interfaces. There's a handful of companies, Scribe
- 1:03:49being one of them, a handful of others that have
- 1:03:52really upped the game on what you should expect
- 1:03:54for your deposition software compared to some
- 1:03:57of the older tools. You will get a link to...
- 1:04:02the materials. So look for that. I think they
- 1:04:06may distribute that by email or perhaps in the
- 1:04:08chat. And I do see one other question, which
- 1:04:12is, can Scribe bill us after the case settles?
- 1:04:15Scribe does have a defer pay option for those
- 1:04:17of you who are interested, who maybe do contingent
- 1:04:19fee work. If you're interested in deferring the
- 1:04:22payment until the end of the case, that is an
- 1:04:24option. And other than that, where does human
- 1:04:27judgment beat AI? That question's a really good
- 1:04:32question, and I would say that answer's changing
- 1:04:34every year, every quarter. When I give presentations
- 1:04:39on AI, when I first started a few years ago,
- 1:04:42I would say, think of AI like a kind of a legal
- 1:04:46assistant, right? You're just gonna get you a
- 1:04:49decent first draft that you're still the lawyer,
- 1:04:51you still need to go through and do a lot of
- 1:04:54the analysis. And as AI continue to get better
- 1:04:56and better and better, I used to say, well, think
- 1:05:00of AI like a paralegal. And then as it continued
- 1:05:03to improve more and more and more, I started
- 1:05:07to say, well, think of AI like a junior associate.
- 1:05:11Right now, it's an attorney -level analysis,
- 1:05:13but you've still got to, you know, the junior
- 1:05:15attorney doesn't know as much as you know. You've
- 1:05:17still got to check it and make sure it's accurate
- 1:05:19and that the analysis is on point. I will say
- 1:05:23that the speed at which things are moving, I
- 1:05:26feel like at some point I'm going to be saying,
- 1:05:29you know, the generative AI tools are kind of
- 1:05:32on par with a junior partner. And maybe for some
- 1:05:37use cases, it's already there today. It probably
- 1:05:41depends on your area of practice and what you
- 1:05:43do. For what I do, I still haven't seen it reach
- 1:05:46the level of replacing or fully augmenting the
- 1:05:51need for an attorney to do drafting. But it does
- 1:05:55speed things up quite a bit. It helps spot issues
- 1:05:57that may otherwise be missed. It also can help,
- 1:06:01it really can help you to develop an action plan.
- 1:06:04And so for me, it really helps me stay organized
- 1:06:06on my cases, confirm that the team, if I'm stepping
- 1:06:09in and kind of either supervising or providing
- 1:06:12a second set of partner eyes on a case. I will
- 1:06:15run it through some of my firm's authorized tools
- 1:06:17to see, is there anything the team's missing
- 1:06:20here? And then I'll go read the source materials
- 1:06:23that the AI references. So just as long as you're
- 1:06:26using the source materials and not purely relying
- 1:06:29on AI summaries, I think you'll stay in pretty
- 1:06:32good shape. All right, I've answered all the
- 1:06:34questions. I think I have time to answer. Thank
- 1:06:36you so much for joining and for listening. If
- 1:06:39you have follow up questions, my email is on
- 1:06:41the screen, Carl with a K at scribe with a K.
- 1:06:45Thank you again to the host for having me. It's
- 1:06:48been a pleasure.