Latest / CLEs You Actually Want to Hear / Beyond the AI Notetaker: How to Use AI Responsibly in Client Conversations
Transcript
- 0:03Hello, everyone. Welcome to today's program,
- 0:06Beyond the AI Note -Taker, How to Use AI Responsibly
- 0:10in Client Conversations. I'm Alapoli Castro,
- 0:13Director of Education and Professional Development
- 0:15at Beverly Hills Bar Association, and today I'm
- 0:18delighted to introduce our speaker, Hilary Bowman.
- 0:21Hilary is the founder and CEO of Curious, an
- 0:25AI platform specifically designed to enhance
- 0:27attorney -client communication. Prior to starting
- 0:30the legal tech company, Hillary advised over
- 0:3340 health care startups on the federal anti -kickback
- 0:36statute, Stark Law, and HIPAA as health care
- 0:39counsel for redesign health. Hillary also previously
- 0:42served as product counsel and health care regulatory
- 0:45counsel at IBM Watson Health. Hillary developed
- 0:48her subject matter expertise through prior roles
- 0:51at global law firms, K &L Gates and Womble Bond
- 0:54Dickinson. Recently, Hillary was chosen by Legal
- 0:58Week as her Innovator of the Year for 2026. We're
- 1:02so excited to hear what she has to share with
- 1:03us today. So I'm now going to turn it over to
- 1:05Hillary. Thank you so much for having me. All
- 1:10right. Well, I will share my presentation and
- 1:13we will get started. OK. All right. Thank you
- 1:22everyone for joining me for the next hour on
- 1:25the presentation Beyond the AI Note -Taker, How
- 1:27to Use AI Responsibly in Client Conversations.
- 1:31A few notes before we get started. So if you
- 1:34have questions, please collect them. You can
- 1:38email them to the organizers as well as save
- 1:40them at the end. There is some time reserved
- 1:43for question and answers at the end of this presentation.
- 1:46Another housekeeping note before we start, this
- 1:48is a general credit CLE. So even though we will
- 1:52talk about the rules of professional responsibility
- 1:55extensively, there's obviously no way that you
- 1:57can talk about AI technology without addressing
- 2:00the This is not a formal ethics CLE. All right,
- 2:04so let's dive right in. By the end of the hour,
- 2:13I want you to be able to... feel comfortable
- 2:16discussing the following three topics. So the
- 2:19first is having a sense of the technology options
- 2:23that are available to you using AI in client
- 2:26conversations. The second is that I want you
- 2:29to come away from the next hour with an idea
- 2:32of how these tools actually work and how there
- 2:35are variations between different types of tools
- 2:38that are within the AI note taking or conversational
- 2:41space. Lastly, I want you to walk away from this
- 2:45present presentation with a repeatable framework
- 2:47that you can evaluate any technology, not just
- 2:51AI that's used in attorney client conversations.
- 2:54This is a presentation that is primarily for
- 2:58attorneys that are licensed in California, but
- 3:01as you'll see in the course of the next hour,
- 3:03there are privacy laws in Illinois and other
- 3:07states that affect the landscape of the legal
- 3:11and the ethical issues that we consider when
- 3:13determining whether or not you can responsibly
- 3:16adopt AI and client conversation. So you'll see
- 3:19those pop up as well, and we'll talk about how
- 3:21they relate to your work in California. A quick
- 3:27note before we get started, I am the founder
- 3:30and CEO of Curious, which is the first legal
- 3:32conversational intelligence platform we build
- 3:35using AI to support attorneys in client conversations.
- 3:40Our business also has relationships with malpractice
- 3:43insurance providers and state bar associations,
- 3:46including the Beverly Hills Bar Association.
- 3:48So I want to be transparent about that. I use
- 3:51some of the data from legal malpractice insurers
- 3:54in this conversation and would like to be transparent
- 3:57about that before we get started. Okay, so let's
- 4:03talk about where this all began. So I practiced
- 4:07healthcare law for over 12 years. I was half
- 4:12in large law firms, half as in -house counsel,
- 4:15including at IBM Watson Health. And when I moved
- 4:18from a large law firm to in -house counsel, that
- 4:22transition was a lot. more difficult than I had
- 4:25expected. And the reason was the scope. Suddenly,
- 4:28I went from being a very focused healthcare attorney
- 4:32working on reimbursement and licensure issues
- 4:35to having client conversations that were fast
- 4:39-paced, very broad, where I was expected to flag
- 4:43employment issues, IP issues, et cetera. It was
- 4:47around that time as I was making the transition
- 4:50when I was talking to the CEO of Healthcare Startup.
- 4:54And one question that stands out in my mind is,
- 4:59is the CEO asking me one day, can I use AI to
- 5:02summarize medical records? And this was before
- 5:05chat GPT was at the tips of our tongue before
- 5:08we were experimenting with it. And that question
- 5:12took me off guard because as a healthcare lawyer,
- 5:15I knew HIPAA forwards and backwards. But in the
- 5:19moment, I didn't have a good answer for how to
- 5:22respond. So I gave the typical lawyer answer
- 5:24and I said, let me do some research and follow
- 5:27up with you. After that conversation, I was reflecting
- 5:31on it and I was talking to a colleague and I
- 5:34said, I wish I could have asked better questions.
- 5:37I wish I could have said in the moment to that
- 5:39CEO, well, what do you mean by AI? What type
- 5:43of technology are you talking about? Or how would
- 5:45you apply AI to the medical record? What part
- 5:49of the medical record would you be summarizing?
- 5:52Maybe even how would you use that summary? Who
- 5:55would you share it with? In the moment, these
- 5:58questions didn't come to mind, but it started
- 6:00me off in thinking. what technology is safe enough
- 6:04for attorneys to use in conversations to both
- 6:08help an attorney brainstorm in the moment, what
- 6:11are issues, what are questions being raised by
- 6:13this conversation, and how to have secure documentation
- 6:18of that conversation afterwards. Either a few
- 6:22months years past and chat GPT came out I started
- 6:27experimenting with it and I realized very quickly
- 6:30that Through that experience using chat GPT that
- 6:34AI is an incredibly powerful brainstorming and
- 6:37summarization tool And so I think you'll see
- 6:40that come out as we talk about the different
- 6:42tools that are available to use in and around
- 6:45client conversations why that question, why that
- 6:49conversation set me off on this path of being
- 6:52involved in legal tech. All right, so I have
- 6:59given webinars like this presentation since Curious
- 7:03launched almost, I think it's two dozen times,
- 7:06maybe even more. And what I have learned through
- 7:10talking with hundreds, thousands of attorneys
- 7:12is that selecting a technology to solve a problem
- 7:16is often the easiest part of the evaluation.
- 7:20What is really challenging is finding a tool
- 7:23that can be trusted to be used in attorney flying
- 7:25conversations and then implementing it in a way
- 7:29that doesn't introduce any unexpected legal or
- 7:33ethical risks. So quick story, I was having a
- 7:37conversation, it was, I had given a presentation,
- 7:40someone reached out afterwards, and I was having
- 7:42a conversation with a very experienced attorney
- 7:45who does government investigations work, and
- 7:47he said, I listened to your presentation. I see
- 7:51absolutely how this technology can improve efficiency,
- 7:56document details, but I'm just thinking about
- 7:59how practically I would implement this in my
- 8:01practice. And so he posed to me this question
- 8:04and he said, what happens if I do all of the
- 8:06right things? I get the client to give the consent
- 8:09to use the technology in the conversation. And
- 8:13then mid conversation, the client says, Can I
- 8:16tell you something private? Can I tell you something
- 8:18off the record? And it was in that moment where
- 8:22I thought, yes, this is the hardest part about
- 8:25using technology in legal practice generally.
- 8:30It's having to decide how the features affect
- 8:35the workflow. And so we'll talk about things
- 8:38like this. In this particular instance, what
- 8:42transpired was a conversation about a pause button
- 8:45in our product in Curious. And so that is now
- 8:48in production, but it was because of an actual
- 8:51conversation I was having with an attorney who
- 8:53was thinking through and picturing themselves
- 8:55using this technology that surfaced a feature
- 8:59that became part of our product. All right, so
- 9:06here's a little bit of framework for how we are
- 9:11going to cover the next hour of this conversation.
- 9:15I hope that by attending this webinar, you will
- 9:18have an understanding of the general categories
- 9:22of tools that can be used in and around conversations.
- 9:26So here's the three topics. Platform transcription
- 9:30and summarization tools. I think as we talk about
- 9:33this category, you'll realize that some of these
- 9:35may already be on your computer in the software
- 9:39that you're already using. So we'll call that
- 9:42one category. The second, general purpose AI
- 9:45note takers, or sometimes I refer to them as
- 9:47generic AI note takers. These are... tools that
- 9:51leverage AI to summarize, transcribe your conversations.
- 9:55They are not specific to a particular industry.
- 9:58They are used generally by lots of professionals.
- 10:01That is another category that we will talk about
- 10:03and we will address the legal and ethical considerations
- 10:07of those tools. The last category is legal conversational
- 10:11intelligence. This is where we'll focus on the
- 10:14tools that are purpose -built for attorneys to
- 10:16use in privileged conversations. Again, we'll
- 10:19use that same legal and ethical framework that
- 10:21we've been talking about through the entire webinar
- 10:24to evaluate those tools. Okay, so let's talk
- 10:33about... the reaction that you may have felt
- 10:36when you heard about the topic of this webinar,
- 10:39or you've thought about incorporating AI note
- 10:41takers into your practice. So whenever I mention
- 10:45the topic of using AI responsibly in client conversations,
- 10:51I typically get... one of two reactions and they're
- 10:55described here and they're pretty visceral. People
- 10:58feel very strongly if they're in the champions
- 11:00camp or the skeptics. But if you're in the champions
- 11:03camp, what I typically hear is, of course, I
- 11:06see the value of using AI in my conversations
- 11:09to take notes for me to capture the critical
- 11:12details because it allows me to finally make
- 11:15eye contact again with my client or to capture
- 11:18the critical details. It allows me to pass off
- 11:21notes to an associate afterwards. Maybe even,
- 11:25and I heard this recently with someone using
- 11:27a desktop app version of a note taker, my clients
- 11:31think this is really cool and helpful and they
- 11:33get more timely and thorough follow -ups. So
- 11:35that's the champions camp. The skeptics camp,
- 11:39here's the arguments that I typically hear from
- 11:41them and they'll say something like, my handwritten
- 11:44notes are perfect. I don't need any help. I know
- 11:47exactly what I need to capture and I just type
- 11:50it into a draft email to myself. Sometimes I
- 11:54also get pushback about my clients will never
- 11:57allow this. I hear that a lot. The funny part
- 12:01about that is a lot of times when I follow up
- 12:04and say, well, what clients did you ask and what
- 12:06was their objection? It's just a sense they've
- 12:09had. They haven't actually had the conversation
- 12:11yet with their clients. So these are two different,
- 12:15very different viewpoints on whether or not attorneys
- 12:19feel comfortable incorporating an AI technology
- 12:23into their client conversation for note -taking
- 12:25purposes. Interestingly, I think whether or not
- 12:30which camp is true, correct, more persuasive
- 12:36really depends on What technology are you picking
- 12:40up? What specific tool are you going to use and
- 12:44how do you implement it? And so I want you to
- 12:47think about that as we talk about three different
- 12:49categories of tools in this webinar, why you
- 12:53could see someone making the arguments for being
- 12:55a champion and why there are skeptics out there.
- 13:02One more slide here about framing the conversation.
- 13:07Up until this point, I really talked about using
- 13:11AI in client conversations, the attraction of
- 13:14that as being an efficiency tool, right? Being
- 13:17more efficient at capturing critical details,
- 13:21handing off that information to a colleague,
- 13:23et cetera. But another really important reason
- 13:27for attorneys having an interest in using AI
- 13:31tools in conversations, is actually a malpractice
- 13:35risk. So we are partnered with the Wisconsin
- 13:38Lawyers Mutual Insurance Company. They recently
- 13:41did an annual review of their claims from 2025,
- 13:452026. And what they highlighted was that 39 %
- 13:50of the malpractice claims that they handled from
- 13:53their policyholders this year could be attributed
- 13:56back to three reasons. And they're listed out
- 13:59here. The first is that the attorney just failed
- 14:02to follow the client's instructions. The client
- 14:05said exactly what they needed to do. For whatever
- 14:09reason, the attorney didn't capture it. They
- 14:12didn't carry it out. The second reason is failure
- 14:15to follow up or just procrastination. They took
- 14:18entirely too long to get back to the client,
- 14:21and that led to a malpractice claim. The third
- 14:24is the attorney just failed to get consent or
- 14:28to inform the client of something extremely important
- 14:32in the representation. What I think is so notable
- 14:36about these three reasons is that it wasn't because
- 14:41like these malpractice claims didn't come about
- 14:44because the attorney misinterpreted the law.
- 14:47It wasn't an error in judgment. It wasn't an
- 14:51error in interpreting the law. They were really
- 14:53administrative errors where better documentation
- 14:57could have gone a long way, right? It may not
- 15:00have solved every single issue here listed here,
- 15:04but having better documentation after every conversation
- 15:08could have definitely been a risk mitigation
- 15:11strategy. So I want you to have that in the back
- 15:14of your mind as we talk about the different tools
- 15:16that you could potentially leverage. Think about
- 15:19the legal and ethical implications of doing that.
- 15:25So for the remainder of the presentation, I am
- 15:28going to use three different types of errors,
- 15:31right? These administrative errors we just talked
- 15:33about that gave rise to malpractice claims. We
- 15:37have accompanying hypothetical scenarios for
- 15:41each of those. And within each of those hypotheticals,
- 15:45we'll talk about a category of AI tool that could
- 15:49have been leveraged in this situation to reduce
- 15:53the risk of a malpractice claim. So there's going
- 15:57to be kind of building on three hypotheticals
- 16:01and weaving in the legal and ethical implications
- 16:04so that hopefully again by the end of this presentation
- 16:07you will have a framework for how to comfortably
- 16:10evaluate technology. Here's the first hypothetical.
- 16:16So this is all around an attorney that fails
- 16:19to follow up or maybe just procrastinates and
- 16:21takes too long to get back to their client. So
- 16:24imagine that you are an entertainment attorney
- 16:28and you're handling the music rights for Toy
- 16:32Story 5. One song is licensed only for theatrical
- 16:36release. Your client tells you to finalize the
- 16:39additional digital. distribution rights for the
- 16:42film in a streaming platform. You draft the amendment,
- 16:45but you never follow up with a rights holder.
- 16:47Well, the film goes digital, the song's distributed
- 16:50without the necessary rights, and the client
- 16:53is hit with an infringement claim. Again, this
- 16:56is not a situation where you got the law wrong.
- 16:59You just simply did not finish the work in the
- 17:02appropriate amount of time. So Let's talk about
- 17:06what technology is available that an attorney
- 17:10can leverage to reduce the risk that this happens
- 17:13in their practice. One logical place where I
- 17:17see lots of attorneys turn to is a platform transcription
- 17:22or summarization tool. The reason for that is
- 17:25it's very convenient, right? It is something
- 17:28that an attorney has at their fingertips, right?
- 17:32It's something that is already in the software
- 17:35they have on their operating system. They've
- 17:38already purchased it and they've integrated it
- 17:41into their practice. Here's an example of me
- 17:45actually using Microsoft Copilot. I opened up
- 17:49a Microsoft Teams meeting, I enabled Copilot,
- 17:52and what you see on the left is a summary here
- 17:56of a conversation of me just. talking into the
- 17:59product, and you see a summarization, a partial
- 18:03transcription of the conversation. What you see
- 18:06on the right is an actual screenshot of after
- 18:09the conversation when I ended the meeting. I
- 18:11said, OK, I'd like to see the AI summary. And
- 18:14it actually prompted me to upgrade. So remember
- 18:17that note here when we talk about the next few
- 18:21slides here. So how is a tool, a platform tool
- 18:27like Microsoft Copilot operating? Here's what's
- 18:30happening. Behind the scenes, it is capturing
- 18:33audio from your virtual meeting. So the technology
- 18:37is listening to the conversation. It's transcribing
- 18:40every word that you're saying. In some cases,
- 18:43it's displaying a verbatim transcript. or it's
- 18:46summarizing it and making that available to you,
- 18:49in some cases during the meeting itself or afterwards.
- 18:54So these are tools that are very easy to pick
- 18:56up. They're very easy to enable in the tools
- 19:00that you're probably already using to meet with
- 19:02your clients, right? If you're meeting on Microsoft
- 19:05Teams or Zoom, these are very convenient tools
- 19:08to pick up. These last two columns here, the
- 19:11retention in the cloud, the secondary data use
- 19:14by the vendor, these are, while I consider them
- 19:18to be features, and you'll see why, they're not
- 19:21readily apparent by using the technology. So
- 19:25let's talk about retention in the cloud. After
- 19:28your conversation ends in a typical platform
- 19:31model, what happens is... an audio file, a transcript
- 19:36file, sometimes even a recording of the screen,
- 19:40depending on what vendor you're using, is saved
- 19:42to the vendor's cloud. And oftentimes, if you
- 19:46look closely at the terms and conditions, it's
- 19:48saved indefinitely unless you reach out to delete
- 19:51it. So that's what's happening behind the scenes
- 19:54a lot of times when you are ending a conversation
- 19:57that's leveraging one of these platform tools.
- 20:00What happens from there? That's the last part
- 20:03here, the secondary data used by the vendor.
- 20:06Once that information is in the cloud, you should
- 20:09be very careful about what vendors can do with
- 20:12that information. A lot of times what you will
- 20:15find in looking through the Terms of Service
- 20:18is that vendors can use your audio, your transcript,
- 20:23even your screen recording to do independent
- 20:26commercial purposes, use it for independent...
- 20:30commercial purposes like training AI tools or
- 20:33aggregating that data and analyzing the trends
- 20:37and publishing that information. It's always
- 20:40it's in the fine print about how the vendor can
- 20:45use your own conversational data once they've
- 20:48saved it to their cloud environment. And the
- 20:52answer to this really depends on what version
- 20:56of the product you are using, what tier you have,
- 20:59et cetera. So remember I said earlier when I
- 21:02was showing you a picture of Microsoft Copilot,
- 21:05remember that screenshot where it said, do you
- 21:07want to upgrade to see your AI summaries? I took
- 21:10this language right from the Microsoft Privacy
- 21:13Policy. And it says that Microsoft products are
- 21:18released in families, meaning You might have
- 21:22Microsoft Copilot, but there's lots of different
- 21:25versions of Microsoft Copilot. And you may have
- 21:28heard or read on LinkedIn about enterprise tools
- 21:32versus consumer tools. That is a proxy for, typically
- 21:38when someone uses the term enterprise, they're
- 21:40saying it's safer. There's better privacy and
- 21:43security standards compared to a consumer tool
- 21:46that doesn't have those or you're not paying
- 21:48as much or maybe you're getting it for free.
- 21:50While that distinction is important, you cannot
- 21:53rely solely on those categories. Those categories,
- 21:59they're not defined by the industry. They're
- 22:02not, there's no standard terminology. And so
- 22:05you really have to look at What version of the
- 22:09product am I using on my computer in my software
- 22:12and what terms of service apply to that? And
- 22:16actually finding the answer to that can be incredibly
- 22:19difficult. A quick story about this. So when
- 22:21we are building our product in a Microsoft Azure
- 22:25environment, periodically we are adding technology
- 22:30from different third party providers in order
- 22:33to build this dynamic product that joins your
- 22:35virtual meeting. And so at the beginning of the
- 22:38year, we were doing a typical audit of just the
- 22:42technology we're using and the applicable terms
- 22:45of service. And I cannot tell you how long it
- 22:49took to find the relevant with certainty, the
- 22:53relevant terms of service that applied to one
- 22:56of our cloud related. technologies that was part
- 22:59of our tech stack. And so as an attorney, this
- 23:02is very important to me, and we track this and
- 23:04even use outside counsel to evaluate it. But
- 23:07you just have to recognize that it's simple to
- 23:12hear, it's simple to hear, oh, just simply make
- 23:15sure the right terms of service apply to your
- 23:18product. But in technology where things are moving
- 23:21so fast, or maybe you have an existing relationship
- 23:24with a vendor, but then you add a new feature
- 23:26with that vendor, every single one of those changes
- 23:30should trigger what terms apply to this product?
- 23:33What version do I have? And have I just... quickly
- 23:37given it a check of maybe five key questions
- 23:40that we'll talk about in the presentation. Okay,
- 23:43so that's the tier problem. While we're here,
- 23:46why don't we also talk about the ethical implications
- 23:48of using a tool in a virtual meeting such as
- 23:52Microsoft Teams? So the first rule that comes
- 23:56to mind is competence. So you're probably familiar
- 23:59that in California rule 1 .1 is not only about
- 24:02competently representing your client, but also
- 24:06being competent about the technology that you're
- 24:09using in representation. So there is a brief
- 24:11comment about this. There is also proposed language
- 24:15that will make this more explicit related to
- 24:18AI and so that QR code can be used to just quickly
- 24:21look up those proposed amendments. I'm also happy
- 24:24if you email me afterwards. I have my email address
- 24:27at the end. Happy to share a link for that as
- 24:29well. While we're here, let's also cover some
- 24:32of the other relevant ethical obligations that
- 24:34come up when you're bringing AI into a client
- 24:36conversation. That includes communication. So
- 24:39are you being really transparent with your client
- 24:41about the technology that you're using to represent
- 24:44them? In the case of bringing AI to a client
- 24:47conversation, it's pretty transparent often when
- 24:51you have a bot joining a virtual meeting. However,
- 24:54there are versions of these products where it's
- 24:57a desktop app. And maybe it isn't visible to
- 24:59the client. And so we'll talk about the importance
- 25:01of being transparent about that. Another ethical
- 25:05obligation that I'm sure you've guessed already
- 25:07is confidentiality. So attorneys have an obligation
- 25:10to make sure that they protect the confidentiality
- 25:14of their client information. And that includes
- 25:18having reasonable assurances that the technology
- 25:21you're using in your representation also protects
- 25:24that confidentiality. And then lastly, I'll mention
- 25:27here is supervision. So whenever you are using
- 25:31this technology and implementing it in your practice,
- 25:35these obligations not only apply to you, but
- 25:37to everyone on your team. All right. So to wrap
- 25:45up that section, remember, we started with a
- 25:49hypothetical about helping your client with the
- 25:54the digital rights to a particular song. So how
- 25:58could we have avoided missing the critical deadline
- 26:01of protecting those rights before the movie went
- 26:05into streaming? Here's, you know, we talked about
- 26:09a platform tool. A platform tool can summarize
- 26:13transcribed conversations that happen in Microsoft
- 26:16Teams, in Zoom. There's also Zoom. Zoom transcription,
- 26:21which is another platform tool. These tools can
- 26:25absolutely capture critical details. That includes
- 26:28deadlines. It can help you. Then if you have
- 26:31it configured, it may potentially auto -populate
- 26:34calendar invites in your Outlook meeting, for
- 26:38instance. This is a tool that, you know, without
- 26:42a lot of imagination, you can see how you can
- 26:45reduce the administrative errors that give rise
- 26:48to malpractice risks that we talked about. However,
- 26:52before you pick up one of these tools, I think
- 26:55the biggest thing I want to emphasize in this
- 26:57section and the biggest error that I see is this
- 27:01thought of attorneys say, well, I already have
- 27:04Microsoft on my computer. I already use Outlook.
- 27:08I already use their email. Why do I need to do
- 27:11additional due diligence before using something
- 27:14like Copilot in my meetings? And it goes back
- 27:17to that point I made about there's product families,
- 27:21there's tiers within those families, there's
- 27:24changes that happen to the product. And so at
- 27:27every turn, you cannot just assume because of
- 27:31existing use, because you have access to other
- 27:33products by that company. that every new feature
- 27:37that's rolled out or new product released is
- 27:40going to satisfy the critical questions that
- 27:43attorneys need to be asking about whether these
- 27:45products protect confidentiality and preserve
- 27:48attorney -client privilege. So if you're interested
- 27:51in a more comprehensive framework about evaluating
- 27:55vendors, this QR code actually points to a handout
- 28:00that I helped develop. co -presented with the
- 28:04Chief Innovation Officer of Legal Aid of North
- 28:06Carolina, Cherie Gilchrist, in February at Vanderbilt
- 28:10Law School. And through that session, we developed,
- 28:13with the help of the attendees, a framework of
- 28:16the questions that you can ask to help evaluate
- 28:19vendors. So I hope that's a valuable resource
- 28:21to you. All right, let's move on to our next
- 28:25hypothetical. So this hypothetical is centered
- 28:29around one of those other administrative errors
- 28:33that has been identified to lead to several malpractice
- 28:37claims, and that is failure to follow a client's
- 28:40instructions. So imagine that you represent a
- 28:44high profile musician named Dolly, and she tells
- 28:47you that she wants her tour bus, that's bejeweled
- 28:50in sequence, to be donated to a charity rather
- 28:52than passing to her residuary beneficiaries.
- 28:55You understand these instructions, but then suddenly
- 28:58she passes away. Before her death, you were not
- 29:02able to change the specific language that would
- 29:05have satisfied her wishes. So now that bus is
- 29:09up for sale. It's at an auction for one point
- 29:11five million dollars and a lawsuit ensues. So,
- 29:16again, not an example of an allegation where
- 29:19you misread the law. You just were not. consistent
- 29:23and enabled to follow the client's instructions.
- 29:27So again, let's ask what tools are available
- 29:30to you in the market where In a conversation
- 29:34with Dolly, you could have captured those client
- 29:36instructions. You could have shared them easily
- 29:38with colleagues. So we'll use another category
- 29:42of note taker, of AI -inclined conversations.
- 29:45We're calling this the generic note taker section.
- 29:49So this category of tools, you may have heard
- 29:52of these before. You may have seen them in meetings.
- 29:56I typically have attorneys tell me, oh, my clients
- 29:58have brought those to conversations as well.
- 30:01They include Otter, Firefly. Granola, Plaid.
- 30:05These are multi -platform, right? They're not
- 30:07platform specific. They are multi -platform tools
- 30:11that are either software, desktop apps, or even
- 30:14hardware. So in the case of Plaid, that's actually
- 30:17hardware that attaches to your phone, listens
- 30:19to audio that it captures from the environment,
- 30:23and then summarizes that and provides output
- 30:25to you after the meeting. The kind of differentiator
- 30:30here from a platform tool to the generic note
- 30:33takers is how it's integrated into your practice.
- 30:37So remember I mentioned the platform tools are
- 30:39very specific to maybe Microsoft or Google or
- 30:42to Zoom. These tools really cut across all different
- 30:46platforms. They're very inexpensive. That's one
- 30:49of the benefits, the draws of them. The fact
- 30:53that they can... perform another platform agnostic.
- 30:56That's also a benefit. But by and large, I have
- 30:59not seen any very legal specific integrations
- 31:03for these tools into, say, practice management
- 31:06tools in the legal space. So let's quickly talk
- 31:10about how these work. And you'll recognize some
- 31:12of this product architecture from our previous
- 31:15section. How do these tools work in practice?
- 31:19When you have a bot from, let's just use the
- 31:23software example. When you have a generic AI
- 31:25note taker that is software that is synched with
- 31:28your calendar, it shows up in a virtual meeting
- 31:31as a bot, a little rectangle. And it's around
- 31:34the faces typically in the meeting. When that
- 31:39happens, what is happening behind the scenes
- 31:42is the tool is recording the audio from the conversation.
- 31:46It's transcribing that a lot of times into a
- 31:50verbatim transcript. It's saving that after the
- 31:53conversation. A lot of times it is also saving
- 31:57a complete audio file and the video from the
- 32:01conversation. So after the end of the conversation,
- 32:05there's lots of detail. There's lots of ways
- 32:07that you can go back and replay what happened
- 32:10in a conversation. That's all part of the user
- 32:14experience. As I mentioned before, there is what
- 32:18happens behind the scenes, right? I wouldn't
- 32:19necessarily call these features, but it's things
- 32:23you should be aware of before you pick up a tool
- 32:26and that includes data retention, data use. So
- 32:30let's focus on four right here, this column for
- 32:34a second. So similar to platform tools, what
- 32:36you will find if you really dig into the terms
- 32:39of service is that these tools save oftentimes
- 32:44As I mentioned, the entire conversation, the
- 32:46audio file, the transcript, summaries, even the
- 32:50video from the meeting itself into a cloud environment
- 32:54that the vendor controls. And there is not always
- 32:58a limited time frame, so it can be stored indefinitely.
- 33:02And I've seen in some examples where the... deletion
- 33:08of that data is a multi -step process. So if
- 33:10you decide, okay, can you please delete my data
- 33:12from that? There was one terms of service that
- 33:16said, okay, you can request the data deletion
- 33:19after a certain period of time that happens,
- 33:21and then you have to further delete it from a
- 33:24trash box or whatever their terminology was for
- 33:28the deletion area. So it's not always straightforward.
- 33:34How do I know this? True story, I was at the
- 33:37American Bar Association tech show earlier this
- 33:40year in Chicago. And I was asked given the work
- 33:44that we've done on curious if I would write the
- 33:47chapter for AI note takers for an upcoming ABA
- 33:51publication. So that that publication was actually
- 33:54published recently. Again, happy to share that
- 33:57information with you. But I actually spent a
- 33:59weekend painstakingly going through the terms
- 34:04of service of lots of common AI note -taking
- 34:07tools that I've encountered in having conversations
- 34:10with clients. Let's talk about the last category
- 34:14here. Secondary data use by the vendor. In this
- 34:18case, again, The questions you should be thinking
- 34:21about asking is, okay, I know that the vendor
- 34:24storing my data for a certain period of time,
- 34:26maybe indefinitely, but what are they allowed
- 34:28to do with that vendor? And it can show up in
- 34:31a lot of tricky ways. So there might be language
- 34:34that says something like, we de -identify it,
- 34:39we aggregate your data. Well, that's a little
- 34:42bit of a... deceptive language there because
- 34:46just because you take client A's data and client
- 34:49B's data and aggregate it together doesn't suddenly
- 34:53make it less risky if they're going to use it
- 34:55for an independent commercial purpose. So I think
- 34:57the emphasis here is look for language that sounds
- 35:01like giving the vendor rights to train AI models
- 35:06off your data, aggregate it, analyze it, de -identify
- 35:09it. Those can all be red flags. All right, so
- 35:16the other key thing that you should be aware
- 35:18of when we're talking about these generic AI
- 35:21note taker categories is the ongoing litigation.
- 35:25If you haven't heard about this yet, there are
- 35:28three active class action lawsuits against Otter,
- 35:32Firefly, and Granola. I think these next few
- 35:37slides will highlight the key issues that you
- 35:40should be aware of. I would like to emphasize
- 35:43that up until this point, the litigation has
- 35:46been focused on the vendor itself. So I often
- 35:48get that question about, is the litigation against
- 35:53the meeting hosts, the attendees? It's not yet.
- 35:57It's certainly conceivable. But the focus now
- 36:00is on going after the technology vendors for
- 36:04having products in the market that did not protect
- 36:07the privacy of the attendees and in some cases
- 36:11allegedly exploited it. So let's talk about the
- 36:15line of cases around Otter AI. This is a set
- 36:21of plaintiffs that Their complaints were all
- 36:24consolidated at the end of last year. What's
- 36:27important to know about this litigation is that
- 36:30the plaintiffs in the case are non -customers.
- 36:33They were meeting attendees where Otter AI was
- 36:36used in the conversation, but they never agreed
- 36:39to the terms of service for Otter AI. And what
- 36:43the plaintiffs are alleging is a few things.
- 36:47One, in the Otter AI case, they are alleging
- 36:49that defendant's product joined the meeting,
- 36:52started recording the conversation, took screenshots.
- 36:55Remember I said there was oftentimes pictures
- 36:58of what was being shared on the screen, stored
- 37:01voice prints, identifying people by their voices,
- 37:04and then emailed all the invitees notes from
- 37:08the conversation afterwards without their permission.
- 37:11And they said, Furthermore, not only did you
- 37:15collect all of this without our consent, on top
- 37:18of it, you used those recorded conversations
- 37:20to train your own AI models, to use it for an
- 37:24independent commercial purpose. So the main underlying
- 37:28laws that are being tested here are the Federal
- 37:34Wiretapping Act, the California Invasion of Privacy
- 37:37Act, the Illinois Biometric Privacy Act, California's
- 37:41unfair competition law and unjust enrichment.
- 37:44So we're going to focus on the SIPA, the wiretapping,
- 37:49the state law here, because there's a lot of
- 37:52great commentary, even though we don't have a
- 37:54final disposition in this case. we have a motion
- 37:58to dismiss that the court granted in part. So
- 38:01we know what claims went away. The claims that
- 38:04are listed here are what have survived, and this
- 38:06is ongoing litigation. So California, like 15
- 38:11states in the country, requires that before a
- 38:14technology or a person intercepts audio from
- 38:17a confidential or private conversation that you
- 38:21need all of the party's consent. And if you don't
- 38:24do that and you commit a wiretap, it is punishable
- 38:27by $2 ,500 of a fine or imprisonment up to a
- 38:32year. And what we know from case law related
- 38:34to this is that you need to get someone the consent,
- 38:38the requisite consent before the interception
- 38:41takes place. So think back to, again, the generic
- 38:45AI note taker that's joining the meeting. even
- 38:49if you're not the technology provider, it's conceivable
- 38:53that at some point someone could say, well, you've
- 38:57allowed this bot into the meeting, you didn't
- 38:59get my consent ahead of time, and it recorded
- 39:02the entire conversation, and that's an unlawful
- 39:05wiretap in California. So that claim is being
- 39:10tested in the Otter AI litigation. Another line
- 39:16of cases that you should be aware of is the Fireflies
- 39:20class action lawsuit. And this one focuses on
- 39:23a different privacy law, the Illinois Biometric
- 39:26Privacy Law. So let me quickly describe the allegations,
- 39:30the nature of the law, and then we'll talk about
- 39:32how it relates to attorneys that are practicing
- 39:35in California. So similar to the Otter cases,
- 39:39the Firefly litigation, the plaintiffs are non
- 39:42-customers. They never agreed to the terms of
- 39:45service, yet they attended meetings where people
- 39:47were using the Fireflies. product in the conversation.
- 39:53So the plaintiffs allege in this case that their
- 39:56biometric data, their voice prints were collected
- 39:59without their express written release and that
- 40:03was collected without their consent. It was collected
- 40:07by technology that didn't have the appropriate
- 40:10published retention schedule and that information
- 40:14was saved and then share, this is the third part
- 40:18of the allegation, was shared with unidentified
- 40:20third parties. So Illinois has the most, probably
- 40:25the most restricted biometric privacy law. The
- 40:30reason why it really has teeth is that it also
- 40:33has a private right of action. So Texas and Washington,
- 40:36as I noted here at the bottom of the slide, they
- 40:38both have biometric privacy laws, but they're
- 40:41enforced by the attorney general. In Illinois,
- 40:44a private individual can bring them. And the
- 40:47way that the fines are set out, you can get pretty
- 40:53high damages because they're per violation. And
- 40:57you see that there in the middle of the slide
- 40:59about a negligent violation versus a reckless
- 41:03violation, and you get awarded attorney's fees
- 41:05and costs if you're successful. So what is the
- 41:09biometric privacy law say in Illinois? As I mentioned,
- 41:12it is providing notice that you are collecting
- 41:16biometric data. It is getting an express release
- 41:19before you collect that data. Biometric data
- 41:22includes a voice print. So the identification
- 41:25that Hillary said this, John said that. That
- 41:28is biometric data, and that happens in virtual
- 41:31meetings or in conversations where you're using
- 41:34some other type of technology to distinguish
- 41:36between speakers. That technology also has to
- 41:39be transparent that it's being retained and to
- 41:43what extent it's being destroyed. Now, why is
- 41:46this relevant? We are talking in a CLE with California
- 41:50attorneys. Why should you care about this? Well,
- 41:54There is no express reach of the statute outside
- 41:58of Illinois, but we do have some case law that
- 42:02interprets or that is working on interpreting
- 42:05how far that law can reach. And so in this Amazon
- 42:10Web Services case, we saw the court say, well,
- 42:16the Illinois biometric privacy law only applies
- 42:20to activities that primarily and substantially
- 42:22occurred in Illinois. But then the Otter case
- 42:26is sort of testing that and saying, well, could
- 42:29it have greater reach? So we're keeping an eye
- 42:32on that to see how far it extends. But the takeaway
- 42:36here is that when you have a client, a party,
- 42:41an adversary that is located in Illinois or has
- 42:45substantial connections to Illinois, this should
- 42:47be in the back of your mind. Another issue to
- 42:52talk about here that we're going to introduce
- 42:55in the category of generic AI note takers. concept
- 43:01of whether or not GenAI waives privilege in work
- 43:03product. It applies to all the tools that we're
- 43:05talking about, but I'm sort of just adding a
- 43:07layered approach of what to evaluate. So earlier
- 43:10this year, we had two very informative cases
- 43:13that help us understand to what extent are courts
- 43:17going to find that use of generative AI in legal
- 43:21practice waives privilege. So when we started
- 43:24building Curious two years ago, We had a theory
- 43:28based on reviewing ethical guidelines about talking
- 43:32to a lot of attorneys and consulting advisors.
- 43:36We had a theory that if our technology collected
- 43:41information about a conversation and trained
- 43:44off of it, right, improved our product, that
- 43:47that would waive privilege. And it's really interesting
- 43:49because the case law is starting to catch up
- 43:52with that theory and validate what we thought
- 43:55was the case from the very beginning. So let's
- 43:58start quickly with USV Heppner. So USV Heppner
- 44:01is a case where a criminal defendant put privileged
- 44:06counsel communications into a public generative
- 44:08AI platform, Claude. And the court found in that
- 44:12case, when he was compelled to disclose the usage
- 44:16and the inputs and the outputs from that use
- 44:19of the technology, that That information was
- 44:23not protected by a turning client privilege or
- 44:25work product because there was no reasonable
- 44:27expectation of privacy using that version of
- 44:30Claude because the platform, it was the consumer
- 44:33version that trained off of his data. And he
- 44:36did that. The last part that they added in there
- 44:39was that it was not directed by counsel. So that
- 44:42was a case where use of generative AI waived
- 44:45privilege. The other side of the spectrum, a
- 44:48different scenario, also decided, I think it
- 44:50was on the same day, was a pro -state plaintiff
- 44:53using generative AI to prepare for litigation.
- 44:56She was acting, of course, as her own counsel.
- 44:58The court found that her use of that technology,
- 45:03she was not sharing. information about her representation
- 45:08with a person or an adversary. They really focused
- 45:11on that, that adversary language. And so they
- 45:14found in that case that she did not have to disclose
- 45:17her inputs and outputs to chat GPT. I think we
- 45:20can reconcile these cases while they came to
- 45:23opposite results. I think it's important, and
- 45:25we'll see this evolve too, of who's using the
- 45:29tool, for what purpose, and what were the relevant
- 45:32terms about confidentiality. Was there reasonable
- 45:35expectation of privacy? We know from Heppner
- 45:37that when the tool is training off of your data,
- 45:41that that can waive attorney -client privilege.
- 45:46Okay. Quick note here, I don't want to spend
- 45:48too much time on this. A lot of times when you're
- 45:52trying to answer these questions about what does
- 45:55the vendor do with my data? Or what data do they
- 45:58collect? Where do they store it? How do they
- 46:00use my data? What privacy and security standards
- 46:03are in place? We look to the website. We look
- 46:06to the marketing page. And it feels silly to
- 46:09say in a presentation to attorneys, Make sure
- 46:12you validate that in the terms of service. But
- 46:14we are all moving a mile a minute. And I think
- 46:17it's a good reminder that even though a marketing
- 46:20page or website says all the right things, you
- 46:24can't just rely on that for your due diligence.
- 46:26And I think a perfect example of that is SOC
- 46:282 Type 2. I have looked at lots of websites,
- 46:31lots of different technologies over the past
- 46:35two years. And sometimes all the right keywords
- 46:39can be on the website. If you're not familiar,
- 46:42SOC 2 Type 2 is a very lengthy process where
- 46:44you put in place certain policies, you train
- 46:47your team to act in a certain way when they're
- 46:50developing software, responding to a potential
- 46:52security incident. In Type 2, that Type 2 part,
- 46:56you are also audited by an independent auditor
- 46:59about whether or not you're following those policies
- 47:02and procedures. And so there's a very clever
- 47:05language trick that I've seen vendors say where
- 47:09it's like, we are compliant with SOC 2 Type 2,
- 47:13meaning we have some of those policies in place
- 47:16and we're following them, but what you're really
- 47:19looking for in SOC 2 Type 2 is that they are
- 47:21certified SOC 2 Type 2. That language matters,
- 47:25and it should also appear in the terms of service
- 47:28in the contractual language. Okay, there's more
- 47:32we could say on this, but that's just a reminder,
- 47:35so we'll leave it there. I want to be able to
- 47:37cover the rest of the content here in the webinar.
- 47:40So the second part of the webinar, what are the
- 47:43key takeaways? Know who your vendor is, what
- 47:46data they have from you, where they store it,
- 47:49and what they do with it. If you want a comprehensive
- 47:52article that talks about wiretapping biometric
- 47:56privacy, the work product issue, this QR code
- 47:59in the corner can be a great resource. I wrote
- 48:02this article earlier this year with an attorney
- 48:05in Illinois who spends a ton of time thinking
- 48:07about the privacy issues there. So please take
- 48:11a look at it. I hope it's helpful. Lastly, I
- 48:14think what is probably boiling up or becoming
- 48:18more apparent is that Using these tools, what
- 48:22is so important about doing it in the right legal
- 48:24and ethical way is that you need your client's
- 48:27consent to do that, especially if you are in
- 48:30a state like California, where you have an all
- 48:33party consent wiretapping law and there's potential
- 48:36that you are encountering. Maybe you have a very
- 48:39broad practice or there's parties from lots of
- 48:42different states where you could potentially
- 48:44encounter someone that is covered by the Illinois
- 48:47biometric privacy law. So we'll kind of focus
- 48:51in on the consent piece in the next segment.
- 48:53So this is our last hypothetical. Again, just
- 48:55trying to have some fun with this. So remember
- 48:58the last administrative error that gave rise
- 49:01to malpractice claims, it was specifically related
- 49:05to failure to obtain informed consent. So this
- 49:09is an example of where you're helping your client
- 49:13and maybe you didn't get their input when you
- 49:15should have, or you weren't really transparent
- 49:17about the consequences of a particular decision
- 49:20or of inaction. So let's use this hypothetical.
- 49:24Imagine you represent a high profile newlywed
- 49:27couple, Taylor and Travis. They just purchased
- 49:29their new lakefront home in Ohio. This is a true
- 49:32story. I'm from Cleveland and I was really surprised
- 49:34to see the headline about this. Travis wants
- 49:37the property to be held jointly with a right
- 49:40of survivorship, meaning it would pass to the
- 49:43surviving spouse. Taylor, on the other hand,
- 49:46separately mentions that she wants 50 % interest
- 49:49in the house to pass to her trust, which means
- 49:51it would fund a charitable foundation. So these
- 49:54instructions taken together would be incompatible.
- 49:58You don't catch this in time. And so again, problems
- 50:02ensue. What is another type or category of AI
- 50:08that can be used in these client conversations
- 50:10to reduce the risk that this inconsistency happens?
- 50:16So we're gonna talk about legal conversational
- 50:18intelligence using Curious as the example of
- 50:22one of these tools. So what you're looking at
- 50:24here on the far left is, and I know it's really
- 50:27small, but what you're looking at is a little
- 50:29box, a little prompt that before you go into
- 50:31one of your conversations, on your calendar,
- 50:34we sync with your calendar, you are providing
- 50:37curious with some context about the conversation.
- 50:40Maybe it's what type of issues do you want to
- 50:43spot? What are some relevant facts that maybe
- 50:45came in through an email? And then there's also
- 50:48a checkbox here to remind you that you've you've
- 50:51requested consent from the parties to the conversation.
- 50:54So that happens beforehand. Once you're in the
- 50:57conversation, if you plan to ask certain questions,
- 51:00you sort of have an agenda in place. That's what
- 51:03you'll see front and center. This was a situation
- 51:06where I just set up a conversation. I had three
- 51:10planned questions about, you know, Client, tell
- 51:12me what you want to accomplish. Tell me about
- 51:14your family situation and what's your biggest
- 51:17concern about your estate plan. So I ran the
- 51:20Taylor -Travis hypothetical through it. And so
- 51:22you can see how curious is capturing the answers
- 51:25to the conversation as the conversation takes
- 51:28place. and it proposes questions, one of which
- 51:31was, how do you want the title held on the new
- 51:34poem? So you can see how by leveraging technology
- 51:37in the moment, it could surface potential issues
- 51:41that were relevant to that hypothetical, where
- 51:44two parties had differing interests and they
- 51:47were in conflict with each other. All right,
- 51:52so how does, let's talk about product architecture.
- 51:54This is always our product architecture slide.
- 51:57How does this compare or contrast to the other
- 52:00categories of tools? Remember the platform tools,
- 52:02the generic AI note taker tools. So in the case
- 52:05of legal conversational intelligence, curious
- 52:08as an example, Curious is taking a lighter touch
- 52:12to the conversation. So it accesses audio from
- 52:15a virtual meeting, a desktop app, but then it
- 52:18processes it in real time ephemerally, meaning
- 52:21it's not going to create a complete audio file
- 52:25or by default, it's not going to create a verbatim
- 52:28transcript of the conversation. That was really
- 52:30important for building a purpose -built tool
- 52:33for attorney -client conversations, because we
- 52:37wanted to ensure that the notes that come out
- 52:40of the conversation reflect your conversation,
- 52:44your impressions, but that don't capture every
- 52:47word you've ever said. So what's happening during
- 52:50the conversation is, again, we're accessing the
- 52:52audio, we're spotting potential legal issues,
- 52:55we're suggesting follow -up questions, we're
- 52:57even pulling in research, flagging admin tasks
- 53:00while the conversation is taking place. That
- 53:03allows you to have a more productive conversation
- 53:06while you get those critical moments with your
- 53:08client to demonstrate your subject matter expertise
- 53:11or in their trust. And then after the conversation,
- 53:14similar to the other tools, this is where the
- 53:16overlap happens, you are getting a summary email,
- 53:19a detailed set of notes, and in this case, a
- 53:22billing entry. Those are also integrated. That
- 53:25output then is integrated to legal practice management
- 53:29tools like Clio, Smokeball, and ADMI case. As
- 53:34you saw before, let's focus on two more aspects
- 53:37of the product that aren't as visible when you're
- 53:40using the tool, and that is data retention data
- 53:42use. So in the example of Curious, Curious has
- 53:47a very limited data retention period. It's 60
- 53:49days for the notes and the email. Remember, by
- 53:53default, it's not capturing audio or transcript
- 53:55from the entire conversation. That retention
- 53:59period can also be reduced to a shorter period
- 54:01of time, especially if firms have opted to use
- 54:04the API, which means you can after every conversation
- 54:08push notes from your conversation to your firm's
- 54:12environment. On the secondary data use front,
- 54:15as I mentioned, because of research we did two
- 54:18years ago, conversations we've had and then following
- 54:21what has happened in the legal, in the litigation
- 54:25space, we are both and a contractual standpoint
- 54:30prohibited from using conversation data to train
- 54:33AI models or really for any other independent
- 54:36commercial use. Because we know after USV Hefner
- 54:40that a vendor that does that would waive your
- 54:43attorney -client privilege. Okay, so last topic
- 54:48to cover here before we put it all together and
- 54:51give you the framework that I promised you would
- 54:53have coming out of this conversation. How practically
- 54:57do you satisfy those consent requirements? Remember
- 55:00when we talked about generic AI note takers,
- 55:03we talked about the otter litigation, we talked
- 55:06about fireflies. Otters focused on the wiretapping,
- 55:10right? Getting consent from all parties to ensure
- 55:13that an unlawful wiretap doesn't happen. Fireflies
- 55:16is more focused on biometric privacy and getting
- 55:19that express written release before collecting
- 55:22biometric privacy data. Here's the takeaway,
- 55:26and there's lots of detail on this slide, and
- 55:28there's even more on the next, but this is what
- 55:30I want you to remember from using these tools
- 55:33in client conversations. Oftentimes, you can
- 55:37solve the consent problem with clients through
- 55:41engagement letters, other written signed communications.
- 55:45You can also do that with colleagues, right?
- 55:47Think of your associates at your firm and other
- 55:50partners. you can have a firm policy that people
- 55:53acknowledge that these tools are gonna be used
- 55:55in conversation. The trickiest group of individuals
- 55:58that you will encounter from a consent perspective
- 56:01is the non -client, the non -colleague. It's
- 56:04the one in the center here at the third parties.
- 56:08Every time before you start a conversation using
- 56:11one of these tools, whether it's using Curious,
- 56:13using a generic AI note taker or a platform tool,
- 56:17you either need to have a visual notice ask consent
- 56:21and document it or give a verbal notice. And
- 56:25then the options differ because as you see here,
- 56:27this is the environment over here on this axis.
- 56:30I'm talking about where the meeting is taking
- 56:32place. This one could be an in -person meeting
- 56:35that's using a desktop app where the third party
- 56:38doesn't see anything, right? They just see you
- 56:40having your laptop out. So there's two different
- 56:43ways that you can get consent, but the consent
- 56:45has to happen at the beginning of the conversation.
- 56:49before the recording takes place or the listening
- 56:52takes place, and you have to document it. So
- 56:55I just, you know, there's a lot of detail, but
- 56:58just focus on the trickiest part is with the
- 57:01non -client, non -colleague third party. Okay.
- 57:05And then here's a little bit more information.
- 57:06You can have access to these slides after the
- 57:09presentation, but when you're writing a blanket
- 57:11consent, right, you have to describe the scope
- 57:14of the information you're collecting or intercepting
- 57:17and then Get that consent and then give someone
- 57:20an opportunity to rescind that or reject it.
- 57:24And then think about how often there's really
- 57:27not clear guidelines about this, but maybe on
- 57:29an annual basis, you're renewing that consent.
- 57:33That's for the blanket consent that you're getting
- 57:35from clients and colleagues. Again, that visual
- 57:38or verbal consent that happens before every conversation
- 57:42to capture the third parties. That also has to
- 57:45cover the same scope, but you can just do it
- 57:47in a very consolidated and concise way. And so
- 57:51for curious, for instance, we have a lot of different
- 57:55suggestions for our attorneys about how to satisfy
- 57:58that. And that includes operationalizing. you
- 58:02automatically add the bot to a virtual meeting,
- 58:05but then you send it to a waiting room so that
- 58:07it can't access audio until everyone consents.
- 58:10You can do that. I mentioned the pause button
- 58:12earlier. So you can auto add our product and
- 58:15then click the pause button and pause the listening
- 58:18until everyone gives consent. You can also manually
- 58:21add the bot to a meeting or use the desktop app
- 58:24and just don't start it until you get consent.
- 58:26So this is the... the level of detail that we
- 58:30have thought through in terms of what's happening
- 58:33in litigation and how does that inform our product
- 58:36architecture. All right, so putting it all together,
- 58:40I have two minutes left remaining in this presentation.
- 58:44Here are the key takeaways. So if you only screenshot
- 58:48one slide from the presentation or save it or
- 58:51print it, I think this one might be it. When
- 58:54you address, when you've identified a problem
- 58:57or a pain point in your practice that you want
- 58:59to solve, and that might be from earlier in this
- 59:03webinar, I want better documentation to reduce
- 59:06administrative errors, reduce my malpractice
- 59:08risks. Once you've identified the problem, that
- 59:12narrows what potential technology can solve it
- 59:15because. I have often received the question of,
- 59:18well, how do I even know what AI tools to adopt?
- 59:20Well, what's your problem? What problem are you
- 59:22trying to solve first? Once you do that, you
- 59:25can really narrow down the options. Once you've
- 59:28identified the potential solutions, the categories
- 59:32of legal technology, then focus on the vendors.
- 59:35Are these credible vendors? Are they handling
- 59:39my data in a responsible way? Do they understand
- 59:41the guardrails? You can tell that from not only
- 59:45the marketing page, that's your first step, but
- 59:47reading the underlying terms of service. I worked
- 59:50with a partner early in my career that I remember
- 59:52this partner would say all the time, I'm not
- 59:54techie, I'm not into technology. And I think
- 59:57of that comment a lot because I think that attitude
- 1:00:02now... cannot exist anymore. We are increasingly
- 1:00:06responsible as attorneys, the competence point,
- 1:00:09right, for understanding not only how to competently
- 1:00:13represent our client, but also be competent about
- 1:00:15the technology we are using to represent them.
- 1:00:19Once you've identified a great technology, great
- 1:00:22vendor, test the product, train your team how
- 1:00:25to use it responsibly, and really think about
- 1:00:28consent. So the tools that we talked about, if
- 1:00:31you bring it to a virtual meeting, we talked
- 1:00:32about it's very transparent. But there's probably
- 1:00:35lots of tools, maybe ones you're using right
- 1:00:37now in your practice where your client has no
- 1:00:40idea that you're using their sensitive client
- 1:00:43data in those AI tools. And so you have to think
- 1:00:47about how do I convey that to them? I know it
- 1:00:50can be kind of a difficult decision, but just
- 1:00:52like talking about bills, right? It is just part
- 1:00:55of practicing law now. Last few points here.
- 1:01:01Once you've identified the technology, you know
- 1:01:02you're going to adopt it. Make sure you configure
- 1:01:05it appropriately. That goes down to the settings.
- 1:01:08And then if there's a mistake, inevitably something
- 1:01:11is going to happen. Just make it right, right?
- 1:01:13If you miss a consent, something gets captured.
- 1:01:16You can always delete it afterwards. Tell the
- 1:01:18affected parties. It's better to be transparent
- 1:01:21about that. And then the last reminder is that
- 1:01:23all of this is changing so rapidly. I actually
- 1:01:27remade this entire presentation for this webinar,
- 1:01:31even though I've given it multiple times, because
- 1:01:33so much had changed in the past few months that
- 1:01:36I kind of just started from scratch. So I hope
- 1:01:39you found it valuable. If you want to talk further
- 1:01:42about this topic, about the decisions you're
- 1:01:46making at your firm about implementing AI, even
- 1:01:48if it's an AI use case that's separate from note
- 1:01:52takers, I'm always Happy to talk with you. I
- 1:01:55feel very strongly that if medicine can figure
- 1:01:58out how to adopt AI scribes, then lawyers can
- 1:02:02figure out how to adopt AI responsibly in their
- 1:02:05client conversations. So thank you very much
- 1:02:08for sharing this hour with me. I hope it helps
- 1:02:11you ask sharper questions when you're adopting
- 1:02:15technology, evaluating vendors, and implementing
- 1:02:18it responsibly in your practice. Thank you. This
- 1:02:24concludes today's presentation. We'd like to
- 1:02:26thank Hilary Bowman for an excellent presentation
- 1:02:28today, and also great thanks to our audience.
- 1:02:31Have a great day, everybody.