Latest / CLEs You Actually Want to Hear / AI Without the Hype: A Practical Starting Point for In-House Counsel
Transcript
- 0:02Welcome to AI Without the Hype, a practical starting
- 0:06point for in -house counsel. I'm Anne Jolly.
- 0:10I'm chair of Beverly Hills Bar Association's
- 0:13in -house section. I'm delighted to introduce
- 0:16to you our speaker, Sonia Sigler. Sonia is the
- 0:20best -selling author of Welcome to the Next Level
- 0:23and What's Next for My Career. She's the founder
- 0:26and CEO of PractiGal Coaching and Consulting
- 0:30and has been an IP and corporate attorney for
- 0:33over 25 years. She started her career at Sega
- 0:37and Intuit before moving to startups, including
- 0:40Catafora, which pioneered AI software for the
- 0:43legal industry. And now let's get started. Thank
- 0:48you, Anne. I appreciate it. And let's start the
- 0:52slides and I'll welcome you all. I am so glad
- 0:55to be here. Welcome, everybody. My name is Sonya
- 0:58Sigler. And as Anne said, I founded PractiGal
- 1:01Coaching and Consulting. I work with lawyers
- 1:04and entrepreneurs to help them get to the next
- 1:07level in their business and in their life. So
- 1:09if you're here, hopefully you're in -house counsel,
- 1:11probably either using AI or maybe even been told
- 1:14to use AI. but are unsure how to make the best
- 1:18use of it in your job and your role and your
- 1:21function. So today's session is going to be interactive.
- 1:25So I want you to use the Q &A function. And I
- 1:29want you guys to answer this question to begin
- 1:32with. So what is your biggest challenge as in
- 1:38-house counsel? And are you being asked to or
- 1:42forced to use AI or using it on your own? So
- 1:46go ahead and put answers in the Q &A. I think
- 1:51that using AI effectively and efficiently is
- 1:54really an important topic and it's gonna be part
- 1:57of getting you to the next level in your role.
- 1:59And I think that lawyers who don't learn how
- 2:02to use these tools are kind of gonna be left
- 2:04behind. So let's go to the next slide and we'll
- 2:06chat about the agenda and what we're gonna do
- 2:08today. So why AI? So, why are we here today?
- 2:19Well, first of all, Belinda McCulley, the executive
- 2:21director of the Bar Association, asked for help
- 2:24a couple of years ago, I guess. She wanted to
- 2:26understand how to use chat GBT because her staff
- 2:29was using it, and she wanted to know more about
- 2:31it. So I did a little training session for her
- 2:35and it was life -changing kind of for both of
- 2:37us because we realized we just need to get the
- 2:39word out on how to use these tools and that there's
- 2:41no magic or mystery to them, that you just need
- 2:44to use them and test them out. So let me give
- 2:47you a little brief overview of what we're going
- 2:49to cover today. We're going to talk about YAI.
- 2:52We're going to talk about the spectrum of impact
- 2:55from a little impact like whiskey tasting notes
- 2:59or, you know, using it to do vacation plans to
- 3:02all the way of filing something with the court.
- 3:05We're going to talk about prompting basics. And
- 3:07then I'm going to do a demo of four different
- 3:09tools. And then I'm going to talk about the cautionary
- 3:11notes in here. So I just want to give you a Cautionary
- 3:18note at the beginning that AI programs are a
- 3:21tool. They're not a panacea. They're not a replacement
- 3:24by lawyers, you know, for lawyers, but lawyers
- 3:27who know how to use the AI tools will rise above
- 3:29the others. So I mentioned that and I will probably
- 3:31keep mentioning that as we go through this whole
- 3:33session. This session is really about the practical
- 3:37uses, not just the theory. And like I said, I'll
- 3:39show you those four tools and we'll get started
- 3:41on that. So let's see, biggest challenge, getting
- 3:44answers from LLMs whose providers want to restrict
- 3:47me from asking questions I need to ask about.
- 3:50Yes, if vendors are not answering your questions,
- 3:53we should try different vendors. Looking for
- 3:55AI platform selection criteria besides security
- 3:58for client data and hazards of erroneous responses
- 4:02and everything. Okay, so we'll talk about the
- 4:04criteria for choosing tools because it's an important
- 4:06part of the process. Let me just write that down
- 4:09so I be sure to cover that when we get there.
- 4:16Okay. Great. Let's go to the next slide and we
- 4:21will give you a little background about how I
- 4:23got here. I've been an IP and corporate lawyer
- 4:27for, gosh, 30 years now. And I started, like
- 4:32Ann said, at Sega and Intuit. It was nice to
- 4:35be at companies that were 800 -pound gorillas
- 4:37and that had negotiating power. And then I moved
- 4:39to the opposite end. I moved to startups for
- 4:45the last 20 years, including It was there when
- 4:50I was at a company that I founded, Catafora,
- 4:53way back in 2002. It was there that I discovered
- 4:55I'm really an operations person with a law degree.
- 4:58So I discovered it because I put a contract management
- 5:02database. from design all the way to implementation.
- 5:05I put that in place everywhere I've ever worked.
- 5:07So that's how I got into this. Then I started
- 5:10PractiGal in 2018 and along the way published
- 5:13the books that Anne mentioned. Welcome to the
- 5:18next level and what's next for my career. And
- 5:20my most recent book that I published is 30 days
- 5:22to better self -care. So I talk about wellness
- 5:24a lot too. But AI is a subject that's near and
- 5:28dear to my heart because I spent most of the
- 5:30last 20 years teaching thousands of lawyers about
- 5:33AI tools. So how to use them, how to use statistics
- 5:36and linguistics. So the question that you guys
- 5:38asked about how do I get the answers out of the
- 5:42vendors, this is important because In teaching
- 5:46people about the tools, I can teach them what
- 5:49questions to ask in terms of how to really understand
- 5:51what's going on with the tool. And then from
- 5:53a privacy security standpoint, how are things
- 5:55handled on the back end to keep things confidential
- 5:59and private as needed? So now I'm helping lawyers
- 6:02use these tools and manage their career as they
- 6:05go along. Let's go to the next slide. And we'll
- 6:12talk about why AI, but let me just check the
- 6:15questions. I'm concerned that solos don't have
- 6:20the resources to compete with big firms and using
- 6:22AI effectively so that clients really get the
- 6:24benefit of the productivity enhancements, shorter
- 6:26time spent on true legal work. Okay. I'll talk
- 6:29about that in just a minute. And then another
- 6:31big challenge, how to run an open source LLM
- 6:33in a secure environment. Right. Okay. So we'll
- 6:35talk about that when we get to the cautionary
- 6:37notes. to talk about what that actually can look
- 6:41like. Great. Okay. So why do we want to use these
- 6:46AI tools anyway, like ChatGPT or Claude or Perplexity
- 6:49or GC AI or any other tool for that matter. But
- 6:52those are the four that I'm going to talk about
- 6:53today. So are we afraid life will just pass us
- 6:57by? Maybe our careers will stagnate because we're
- 7:00not using these tools. And then how do I, as
- 7:03a solo or as a practitioner in a small, place,
- 7:08compete against those bigger resources. Is everybody
- 7:12else using these? So I should use them too, or
- 7:14the company is actually forcing everybody to
- 7:16use the tool. So I just want to back up for a
- 7:21minute. So what are lawyers good at? What is
- 7:24your highest and best use of your time? So this
- 7:27goes to the answer with the solo question and
- 7:29how do I use these resources? So if your highest
- 7:32and best use is arguing or persuading or putting
- 7:35the pieces of the puzzle together to tell a cohesive
- 7:37story, or maybe it's making a game plan or setting
- 7:41the tone and direction and strategy for a case
- 7:43or negotiation, or maybe it's strategic thinking
- 7:47or negotiating, drafting, researching. Like I
- 7:50know a lot of people who love the researching.
- 7:52I'm writing part of it. That is not me. So where
- 7:57do we want to use these AI tools? So what is
- 7:59the purpose? What is the why for using the tool
- 8:01in the first place? Using these tools wisely
- 8:05will free us up to do the good stuff that we
- 8:08want. It'll free us up to do the highest and
- 8:10best use of our time. So if you know that ahead
- 8:13of time, then we can look at these tools and
- 8:16what can they take off our plate that is a lower
- 8:20level use of our time and free us up for the
- 8:23good stuff. I'm calling it the good stuff, but
- 8:25really it's the higher level thinking and it's
- 8:27using our judgment because that's where we will
- 8:29never be replaced is in that thinking and the
- 8:32judgment that we have. based on our experience.
- 8:35So there are so many tools available. We want
- 8:38to kind of use them for the repetitive boring
- 8:40stuff so that we can concentrate on that fun,
- 8:42that good stuff I mentioned. And there are so
- 8:45many tools. So we're going to chat about four
- 8:47of them today, but I just want to say there are
- 8:49so many tools out there that it's hard to keep
- 8:52up with everything. I recommend choosing one
- 8:56or two or three tools and evaluating them and
- 8:58really understanding them rather than trying
- 9:00to like boil the ocean and figure out all the
- 9:02tools. And these tools are updated almost daily.
- 9:05I feel like every time I open chat GPT, it says
- 9:07new version available. Do you want to install
- 9:10it? So that's the other thing that can keep happening
- 9:13is these tools just change so quickly and they
- 9:16are getting smarter and smarter as we keep using
- 9:20them. So let's take a look at the what and the
- 9:25why behind these tools. Go to the next slide.
- 9:34Go to this one that's marked AI tools, please.
- 9:39Okay, so these tools are available 24 -7. And
- 9:45they never get tired. So this is where they have
- 9:48an advantage over us as people. They help us
- 9:52externalize our thoughts and get everything out
- 9:54of our head on kind of on to the paper. And they're
- 9:57great at organizing, simplifying, starting and
- 10:02hopefully finishing tasks or projects. The difference
- 10:06about these tools from prior AI tools, like the
- 10:11technology assisted review and all those kind
- 10:13of tools, is that those tools focus on the data
- 10:19that you have in front of you and analyzing it
- 10:21and making sense of that data and drawing connections
- 10:25in that data and looking for things that go across.
- 10:28It could be themes, it could be data to support
- 10:30something, but those tools really focus on data
- 10:36that exists. The latest AI tools, the tools that
- 10:39have really blossomed since let's say 2022, 2023,
- 10:45meaning they generate a response as opposed to
- 10:50like looking at the existing information and
- 10:52then serving you the results that it finds. So
- 10:54I'm gonna put this in context of these tools
- 10:59can be really powerful if they're used as intended.
- 11:03And for us as lawyers to look at these tools
- 11:06and from a technology and technologist standpoint
- 11:08to say, what do they even do? What are they good
- 11:10at? It can feel like we're kind of behind the
- 11:14eight ball or whatever. So I'm going to give
- 11:15you one example. Dolphin search. Great search
- 11:18engine. You put your data in it. It's going to
- 11:22give you results to say these documents are relevant.
- 11:28Documents are not relevant. These documents may
- 11:31be privileged. Take a closer look. Could be hot
- 11:33documents. However you set up your system. People
- 11:36thought the tool didn't work because every time
- 11:38they added information, it gave you different
- 11:40results. That's because the system got smarter.
- 11:43But that tool was very powerful if used correctly.
- 11:46And then the generative AI tools are the same
- 11:50way. They have to be trained. You can make up
- 11:53different personas. Like if you are a cranky
- 11:56I'm going to say this persona, Cranky Senior
- 11:59Partner, who's got a red pen in hand to look
- 12:03at your materials. It could be a marketing one.
- 12:06It could be if you're in HR, make an HR executive
- 12:09or CEO, whatever it is. You make one that's your
- 12:12brand, your voice. And I'll give you some examples
- 12:14of this as we go further into this. Let me just
- 12:19take a look over here at the questions. Resources,
- 12:24open source. So I saw an opinion out of New York
- 12:31that a client's personal AI logs related to their
- 12:33attorney representation, the client using AI
- 12:35to draft emails to their attorneys, destroy the
- 12:40privilege. How would you go about using the other
- 12:42side of AI against them in discovery? Okay, we'll
- 12:45talk about that in the cautionary section. Okay,
- 12:48let's go on. I'm gonna explain a little bit more
- 12:50about the tools themselves. So I suggested that
- 12:54prior tools in the... use of AI tools before
- 13:03this, before the generative tools. They're based
- 13:06on search and retrieval, right? You're used to
- 13:08all these tools like LexisNexis, Westlaw, Google,
- 13:13whatever your favorite search engine is, Bing,
- 13:17DuckDuckGo, whatever that is. those were all
- 13:21about finding the right search to find the information
- 13:24you're looking for, right? So that was all like
- 13:27plumbing the depths of the ocean, pulling up
- 13:29a result and saying, is this what I was looking
- 13:31for or not? The difference with these tools that
- 13:35are all actually, if you can go to the next slide,
- 13:38please. These are large language models, meaning
- 13:42there is a plethora of information. Think about
- 13:45Google and all the results that it's looking
- 13:47through and pawing through to give you information.
- 13:50So generative AI tools generate the output. So
- 13:57that's the big difference. They're not just interpreting
- 14:00the information that it has in front of them.
- 14:01It's generating you some results. So I'm just
- 14:05going to talk about the difference between tools.
- 14:08that have gone before us and tools that we're
- 14:10using now. So extractive tools, that's what I
- 14:12call technology assisted review or predictive
- 14:15coding. They are based on the data that you have.
- 14:18So you feed it discovery results. The generative
- 14:21tools make up something based on the underlying
- 14:25model. So this is where the models are important
- 14:27because the different tools are based on different
- 14:29large language models. GPT stands for generative
- 14:35pre -trained transformer. It's a family of large
- 14:39language models. It's developed by OpenAI. They
- 14:42generate the word or phrase or paragraph based
- 14:45upon the language used, based on the frequency,
- 14:48based on the phrases, and what it is fed. So
- 14:52it assigns a number to each word and then they
- 14:55look at the frequency, the pairs, the phrases,
- 14:58the sentences. labeled vector space modeling.
- 15:01So basically it's an algebraic way to model a
- 15:04document or text. So it's basically translating
- 15:07those words into numbers and then looking at
- 15:09the frequency and the patterns in there and then
- 15:11being able to generate language. So back from
- 15:15the number to the letter, generate language and
- 15:17a result for you. So if you use, you know, AI
- 15:21tools, you use these every day, you probably
- 15:23just don't even know it. So like AI tool, AI
- 15:26results come in Amazon or eBay, they make suggestions
- 15:28all the time. So those are our predictive kind
- 15:33of tools that are going to. predict your behavior
- 15:35based on prior purchases. Another use would be
- 15:39when you talk to Siri or you start a text and
- 15:41it gives you predictive words based on what other
- 15:44people have done and what you have done. But
- 15:46I just want to say these are all tools, like
- 15:49calculator or Microsoft Word. You know what?
- 15:52There's no easy button in this whole equation,
- 15:55okay? This technology is growing by leaps and
- 15:57bounds and it is, I'm going to say, continuously
- 16:02use it and explore it. and iterate on it and
- 16:06make use of it. So let's go to the next slide.
- 16:10So what about this context? So large language
- 16:18models are pretty amazing. They can generate
- 16:21almost anything, and especially if you train
- 16:23it to your voice and tone and your expertise.
- 16:25But the English language is hard. So when you've
- 16:28got something with the same name, George Bush,
- 16:31the dad or the son, which one is it? Words spelled
- 16:34the same, tear or tear. Those can like wreck
- 16:38the models. So those are things that can make
- 16:41tools act differently, even though you're putting
- 16:45the same information in there. So I just want
- 16:47to set the stage here. Who are you and why are
- 16:52you using an AI tool to generate something for
- 16:56you? Okay, so. As you use these tools, are you
- 17:03an associate? Are you in -house counsel that
- 17:08represents the sales team? Are you like I was,
- 17:11I did IP, so I negotiated agreement after agreement
- 17:14after agreement. So who are you and why are you
- 17:18using the AI tool? Okay, so that context is really
- 17:21important because what context you give it is
- 17:27going to give you a different result. So I'm
- 17:29going to ask you, what hat are you wearing today?
- 17:31Are you an intern, associate, cranky senior partner?
- 17:33Are you the general counsel? Are you an associate?
- 17:37Are you a marketing director drafting a flyer
- 17:40or drafting an ad or webinar description? Are
- 17:44you a writer? Now you've got to write a bio and
- 17:46a presentation like this. Do you need to generate
- 17:51a byline because you just put a proposal together
- 17:54for a presentation? Are you generating a headshot
- 17:58or a dating profile picture? Are you summarizing
- 18:01a deposition or drafting a release? One of my
- 18:04first uses of AI was to draft a release for a
- 18:07group I was on the board of and I didn't have
- 18:10a good win in my templates. So I was like, okay,
- 18:13I need help doing this. What's the minimum I
- 18:15can put in there? Are you drafting a set of interrogatories?
- 18:19What is the context of what you're doing? Because
- 18:22it's super important. So let's go to the next
- 18:24slide. It's super important because the uses
- 18:33for these tools are endless, okay? The sky's
- 18:37the limit. Whether you're drafting, whether you're
- 18:39putting a template together, whether you're saying,
- 18:40should I even sign this agreement? whether you're
- 18:44drafting a policy or reviewing a policy, whether
- 18:47you're writing briefs, summarizing, writing an
- 18:49article, approving marketing materials that are,
- 18:53you know, coming out of the product team, you
- 18:56know, headshots, like everything on here. Use
- 19:01it to write client letters or Communications.
- 19:07My husband calls it the make Sonia friendlier
- 19:09tool because I just usually put the line that
- 19:12I need to convey in my emails and then I go back
- 19:15and I put the niceties in it. So he calls it
- 19:17the make Sonia friendlier tool. Do you use it
- 19:20to fight the blank page syndrome? So I'll tell
- 19:23you more about that one in a minute. Do you use
- 19:26it for processes, procedures, checklists? Do
- 19:28you have playbooks or templates that you need
- 19:31to evaluate agreements with? or red line agreements.
- 19:35So it's only limited by your imagination. And
- 19:38it's really just a tool to be tried and tested
- 19:40and vetted. And so before we launch into the
- 19:42prompting basics, I want to talk about the importance
- 19:45of the impact spectrum, as I like to call it.
- 19:50So let's go to the next slide. So it's crucial.
- 19:56And this is really important. Go to the next
- 20:00slide. It's crucial to understand the importance
- 20:03and the impact of what you're using the tool
- 20:05for. So remember I said, who are you? Why are
- 20:07you using this tool? So let's look at that context.
- 20:10Because if you're relying on the tool for something
- 20:13easy, like I'm writing an article that's going
- 20:15to be in a local journal, or I'm writing a law
- 20:19review article, is it for an industry trade public
- 20:24bin, or is it marketing? So where does it fall
- 20:26in this? spectrum of impact. For anything on
- 20:31that crucial side, like court filings, or they're
- 20:33going to be shared with the other side, I really
- 20:36want you to consider using a legal specific AI
- 20:40tool. And the reason I'm going to demo GCAI is
- 20:44that's a legal specific tool because they have
- 20:50taken that into account. into account and how
- 20:53they've written their tool. So let's look at
- 20:55these. I'm going to take each one of them. And
- 21:01I'll talk more about what's in these tools specifically
- 21:05in terms of the impact. So low impact things.
- 21:09Let's go to the next slide. Are anything like
- 21:13travel planning, menu planning, block scheduling?
- 21:17I used it for a whiskey tasting event I hosted.
- 21:21I used it in marketing materials, blogs, articles.
- 21:24These are all low impact. These are things that
- 21:29if you get something wrong, it's not going to
- 21:32change or alter the course of the world. So these
- 21:35kind of things, easy uses. I'll give you one
- 21:39example. I write a column for a local magazine.
- 21:42I've been doing it for seven years. I've got
- 21:45At the time, I started using ChatGPT to help
- 21:48me. I trained it. I do the cover interview, cover
- 21:53family, and then a column called making a difference.
- 21:57I only get 250, 300 words. So I trained the system
- 22:02to write in my voice by giving it prior articles,
- 22:05by giving it my interview notes, and I compressed
- 22:08what used to be a two -hour process into 15 minutes.
- 22:12So That was important and it made it much easier
- 22:17to get that work out the door and in my voice
- 22:20because I trained it that way. And then I have
- 22:23very specifics about no nonsense, practical advice
- 22:26that can be followed and I iterate those when
- 22:31I'm having it generate text for me. And then
- 22:34of course you have to review it and don't just
- 22:38take the output and put it out. there you actually
- 22:40have to review it. So let's talk about higher
- 22:41impact. Let's go to that next slide for higher
- 22:44impact. So these things can be anything that's
- 22:48going to someone else like client communications.
- 22:50It can be the client intake process where you're
- 22:52gaining information from clients or if you're
- 22:55taking you know work in and you need to track
- 22:58it in -house because being in my existence or
- 23:02you're measuring it against a standard agreement,
- 23:06or you're trying to summarize something for the
- 23:07business people, or let's say it's now you've
- 23:09got to research something, or you're redlining
- 23:12your agreements, or you need to update a bio
- 23:15resume cover letter, whatever. So these are higher
- 23:18impact because they're going to be relied upon
- 23:19by other people. I find that these require more
- 23:24review, more scrutiny. And I'll give you an example.
- 23:29I helped a client They were hiring paralegal
- 23:32and they kept getting candidates and they had
- 23:36asked for a cover letter. So they kept getting
- 23:37these cover letters from candidates. And I read
- 23:42every single one of them. And I'm like, these
- 23:44people have gotten lost in the gobbledygook that
- 23:47AI spit out. I couldn't even tell who they were
- 23:49as a paralegal. Like from a philosophical standpoint,
- 23:53how they approached their job or who they were
- 23:56as people. So those cover letters were like,
- 23:59okay, you wrote a cover letter, so you check
- 24:00the box for that. But it didn't convey the story
- 24:03that they needed to tell about themselves and
- 24:05who they were as a paralegal and how they could
- 24:06help the client who was looking to hire a paralegal.
- 24:09So don't get lost in the middle of this because
- 24:11this particular output has a higher use. So like
- 24:15I said, you can... train it with the voice that
- 24:18you want, and definitely part of the picture
- 24:21in what we need to do to train the tools. Let's
- 24:24go to crucial impact. Let's go to that next slide.
- 24:26We'll talk about crucial impact and what it is.
- 24:30This is anything that's going to be signed. This
- 24:32isn't waivers. This is App Davis. It's court
- 24:34documents. It's an agreement that you're going
- 24:36to the other side. These, and frankly, I would
- 24:40use a legal specific tool for this. Frankly,
- 24:44I would also say chat, GPT, Claude, Harvey, all
- 24:48these tools are a little too general and you
- 24:53don't know exactly how it was trained. So I would
- 24:57be very, very, very wary of using any of those
- 25:02tools for anything that was on the crucial impact
- 25:05side of that spectrum. This is where one of the
- 25:09cases in your materials, the Avianca case, where
- 25:12the lawyers lives in New York didn't do their
- 25:17due diligence and they just used the cases that
- 25:21chat GPT made up. I'm sure you've seen all of
- 25:25those cases that have cautioned about using stuff
- 25:29because it makes up case sites. Well, these lawyers
- 25:31did their due diligence in a way because they
- 25:34tried to find the case, couldn't find it, but
- 25:36they filed it. with those references and cases
- 25:40cited in it anyway. Then they lied about it,
- 25:42they withheld the truth, and then they finally
- 25:44delayed, delayed, delayed before coming clean.
- 25:46And obviously sanctions happened in that case.
- 25:49But one of the ways to prevent this sort of problem
- 25:52is always asking for sites and references in
- 25:54your prompts. And we'll get to that in just a
- 25:57minute. So let's talk about prompting basics
- 26:00before we get to the demo. Let's go to that next
- 26:03slide. And we'll talk about what this is. Okay.
- 26:07Remember I mentioned that spectrum of keyword
- 26:09search and trying to find your information using
- 26:12Lexis and Westlaw and you spent all that time
- 26:14learning, you know, search with, you know, pen
- 26:19within five of pal to hopefully find, you know,
- 26:22the word pen pal, something like that. Or then
- 26:25you used Google and you had to wade through pages
- 26:29and pages of results. So now the AI summaries.
- 26:32happen in Google and whatever tools you're using,
- 26:34and some of them are way off base, some of them
- 26:36are helpful, rarely are they 100 % spot on. So
- 26:40it's just like any other search and retrieval
- 26:41system, you have to be discerning. So what's
- 26:44different about these tools? Remember I said
- 26:46these are generative, they make up stuff, they
- 26:49generate information based on the underlying
- 26:51model. The prompting, is a way to put the context
- 26:59back into the system and what you're asking it
- 27:01to get a better result. So it's, like I said,
- 27:04we all learned how to manipulate keyword search
- 27:08and Google searches to get what we were looking
- 27:10for. I just want to say, let's go deeper in these.
- 27:16And the more details, actually let me just back
- 27:19up a second. I want you to think of these AI
- 27:21tools as an eager intern or first year associate.
- 27:25You have to train them, right? They don't know
- 27:27everything. They're eager to do the work, but
- 27:29they don't know what they don't know. You have
- 27:31to supervise them. You have to learn. They have
- 27:34to learn what you want and how you like things
- 27:37to be done, right? Like if you're going to give
- 27:39a Starbucks order of half calf, lots of foam,
- 27:42you know, not just plain black coffee. That takes
- 27:45time. It takes iteration, takes the experience
- 27:47of getting to know you, how you like to work,
- 27:49how you like your results. So when we look at
- 27:53the prompting, the more detailed you can be in
- 27:57your prompting and the querying, the better the
- 28:00tool is going to understand the assignment. So
- 28:03the better that intern is going to understand
- 28:04what it is you want. So let's talk specifically.
- 28:08Let's go to prompting. Let's go to the next slide,
- 28:11please. Prompting is an art, okay? It's not a
- 28:15science. What do we need to know about prompting?
- 28:20It's like searching for that elusive key in Westlaw.
- 28:25Thankfully, we don't have to do that anymore.
- 28:27But think about prompting the same as you would
- 28:29about crafting one of those effective searches.
- 28:33You have to practice, you have to try, you experiment,
- 28:35you look at the results, you adjust it, do it
- 28:36again. All these tools are the same. So AI tools
- 28:40act the same way. So if you see a prompt list
- 28:43to buy or whatever, don't fall for that because
- 28:46context matters. And things change so quickly,
- 28:51you're better off just experimenting on your
- 28:53own. So let's talk about some considerations
- 28:56in writing your prompts and making your ask.
- 29:00So key concept here, garbage in, garbage out.
- 29:06Okay, so vague prompts are going to give you
- 29:09vague answers. Specific structured prompts will
- 29:13give you a targeted and hopefully more helpful
- 29:16response. So I'll give you an example. Here's
- 29:19a poorly constructed prompt. Write a legal argument.
- 29:25Okay, too vague. What's it going to do? Right?
- 29:2715 pages. Who knows? A better prompt might be
- 29:30draft a 500 word persuasive argument for a breach
- 29:34of contract case involving a missed payment and
- 29:38cite relevant case law with references. Right?
- 29:40That gave you its breach of contract, it involved
- 29:44a missed payment, and I need you to cite case
- 29:46law with references. Right? Much better context
- 29:48for that. So other things we can do to help our
- 29:52prompting be more effective. First, role assignment.
- 29:55Remember I said, who are you? Why are you using
- 29:57this tool? So if your role, I'm acting as a legal
- 30:00assistant, I'm acting as in -house counsel, I'm
- 30:03acting as a chief marketing officer, or I'm acting
- 30:07as a CEO. Whoever the context is for the role
- 30:11that you're... Addressing in the results that
- 30:15you're looking for, that's the role you need
- 30:17to tell the tool about. Context. Provide background
- 30:20information details like the breach of contract
- 30:23case involved in missing payment. So those context
- 30:27bits matter for the results you're going to get
- 30:29back. Format. You can tell it what you want.
- 30:32Bullet points, draft an email, give me a checklist,
- 30:35give me a red line, give me a Word document,
- 30:37slides, whatever. You could also give it tone
- 30:40and style. I want it to be formal. I want it
- 30:42to be concise. I want it to be persuasive. I
- 30:44always say no nonsense. Don't give me any fluff
- 30:46language because the early use of the tools would
- 30:49give you fluffy language. I was like, is there
- 30:50even a sentence here that makes sense? So you
- 30:53can be specific about what you want the tone
- 30:56and style to be. So I'm going to give you one
- 30:57example here. Imagine you need a client email
- 31:00summarizing a legal memo on intellectual property
- 31:03and instead of saying write an email about IP
- 31:07rights, try this. Act as a legal assistant, draft
- 31:11a professional email, so you've got the context
- 31:14of the role, you've got what you want, you've
- 31:16got an email, to a client summarizing a legal
- 31:19memo on IP rights. The client is a small business
- 31:22owner concerned about trademark infringement,
- 31:24so use clear, client -friendly language. Okay.
- 31:28It gave it all of those things. It gave it the
- 31:29role, the context, the format, the tone, and
- 31:32the style. Okay. So let's go to the next slide.
- 31:37And I want to say, like, we can put all that
- 31:39in there and we can crap back, right? Like, I
- 31:42want you just to remember that prompting is iterative.
- 31:47So there is no one size fits all here. This is
- 31:50one big colossal experiment. And I'll give you
- 31:53some examples of prompts in just a minute. But
- 31:56even asking the model the same question, which
- 31:58we're going to do when we demo this, in different
- 32:00times. So even asking the same tool the same
- 32:03question, you can get different results. In fact,
- 32:06I just read this morning about Claude and the
- 32:10three main tools and how to use them. And we'll
- 32:15get to that in a minute. But use their Sonnet
- 32:19tool and their other tool. It gave very specific
- 32:23reasons and context for using those tools. We'll
- 32:27dig into that in a minute. But even asking the
- 32:30same question in the same tool can give you different
- 32:32results. Asking the same question in different
- 32:35tools, obviously, will give you different results,
- 32:38but wildly different results. And I'll show you
- 32:39that in a minute as well. Let's see. Let's talk
- 32:45about the things that I put here on the slide.
- 32:48So opposition. This is you can develop a persona
- 32:51to be the opposition. So if you're looking at
- 32:55redlining agreement, you can say poke holes in
- 32:57my arguments. So if you're trying out arguments
- 33:01with negotiation, you can have that ahead of
- 33:04time. It's just like trial prep. Okay. Overload.
- 33:08Tools can get overloaded very easily. So break
- 33:11down the asks in your prompts. Otherwise, it
- 33:14can lead to the overload. Instead of saying draft
- 33:16me an agreement, draft me the introductory provision.
- 33:21Or let's iterate on that and get that good before
- 33:24we go on to any other provisions. Missing. Okay.
- 33:28What am I missing? So this is one of the things
- 33:30I like to ask in my prompts. What am I missing?
- 33:34What did I not address in this whole thing? What's
- 33:36missing in the agreement that should be in there?
- 33:39The other thing in terms of prompting, I say,
- 33:42take your time, be slow about it, do three iterations,
- 33:45then give me your best suggestions, okay? And
- 33:48then this is my favorite. And this is like, I
- 33:50think Henry Kissinger, when he worked with people,
- 33:55he wouldn't even read the first drafts. He would
- 33:57just return it with a red pen around it and say,
- 33:59is this your best work? So you could do the same
- 34:02thing with the tools and ask it, is this your
- 34:03best work? And then you'll get different results.
- 34:08So these are the current kind of basics in terms
- 34:11of prompting. So it is. an art, not a science.
- 34:15You got to give it a context and role and purpose
- 34:18and all that stuff. And it's iterative, okay?
- 34:21It's definitely not a one and done. So let's
- 34:24look at some good or bad prompts. Let's go to
- 34:26the next slide. So on this one, I need to decide
- 34:34whether to settle this case or take it to trial,
- 34:37provide a decision -making framework based on
- 34:39potential outcomes and pros and cons. So is this
- 34:42a? Good prompt or a bad prompt? It depends on
- 34:54how well you know your work. If you've developed
- 34:55a relationship using the tool so it knows how
- 34:57you like to do, if it knows you're an IP lawyer,
- 34:59for example, you might get a completely different
- 35:02framework if you're focusing on, you know, personal
- 35:04injury crisis than if you're an IP lawyer. So
- 35:07this is not a good or bad prompt. It just depends.
- 35:10Okay. Let's go to use cases. Okay, this is where
- 35:16a lot of the questions over here that you guys
- 35:19have asked so far in terms of how do I use these
- 35:21tools? Do I need to be using these tools? All
- 35:24of those kind of things. I would say, and I want
- 35:28you to think about this. You don't have to share
- 35:31your answer unless you want to, but what are
- 35:33your pain points, okay? Where in your daily job
- 35:38Is it painful or you look at it in email and
- 35:41you're like, I just don't want to deal with this.
- 35:44What are those pain points? Is it client intake?
- 35:47Is it tracking the work that's come into your
- 35:49department? Is it drafting agreements? Is it,
- 35:52this is the bane of my existence for one job,
- 35:54working with the sales team because they wanted
- 35:57everything on a quarterly deadline when it was
- 35:59important for them to get their commission. you
- 36:02know, same thing at the end of December when
- 36:04everyone's like trying to take vacation and they're
- 36:06like, no, I need to close this deal. Is it so
- 36:10many NDAs? Is it a contract or terms of service?
- 36:13So what are your pain points? And you can share
- 36:20it in the Q &A. You can just keep it to yourself.
- 36:22But I want you to think about this because where
- 36:26can these tools be used to help you Remember,
- 36:32I talked about the low level things so that you
- 36:34can be using your judgment and your best and
- 36:37high use. Where are those low level things? Where
- 36:39are those bottlenecks? Where are things that
- 36:42you happen frequently that could go more smoothly?
- 36:45Where are those bottlenecks? Where are people
- 36:47always waiting on you or your department to finish
- 36:51something? And I would say, really, take a close
- 36:54look at your processes and procedures. Let's
- 37:03go back to those pain points. And then the other
- 37:07one I have is forgetting to eat and move. Like
- 37:09I'll sit down and I'll be here for six hours.
- 37:11Then I'll forget to get up and walk around, take
- 37:13some water, whatever those. So those, I made
- 37:15a checklist. I made a punch list. So of all those
- 37:18typos, when I'm drafting an article or when I'm
- 37:21drafting a actually anything, I made a list,
- 37:26a punch list of typos to look out for. So one
- 37:29of my favorites is does not. I type fast, I type
- 37:34faster on one hand than the other, and it comes
- 37:36out as doe snot. So not a misspelling, right?
- 37:40It's just a typo in terms of that's not what
- 37:43I really meant. So those elements, those words
- 37:49that I know to look for in my documents, I just
- 37:51made a checklist for it. So I made a punch list.
- 37:53So it made it easier for me to edit. Another
- 37:58example. When I worked with that sales team at
- 38:02Catafora, I drafted a template agreement that
- 38:07had everything on the front page are things that
- 38:09they needed to come up with to help me get a
- 38:12deal out the door. Then I had the terms of service
- 38:15as an exhibit to everything else. And so I trained
- 38:19them. on the five things, it happened to be five,
- 38:22five things that they could change in that agreement
- 38:24without having to come back to me. And I gave
- 38:27them the range of things they could change it
- 38:28to, and I gave them the language that they contained
- 38:30it to. Otherwise they had to come back to me,
- 38:32which was going to take them longer to get that
- 38:34deal done. So where are those pain points for
- 38:38you in your role as in -house counsel? So the
- 38:41beauty of these tools is that you get to make
- 38:44your own roadmap and you get to design your own
- 38:46solution and it can fit what you're doing. Okay.
- 38:49So let's look at the tools. Okay let's stop sharing
- 38:56that and I am going to share my screen. I got
- 39:00to move all these things over give me one second.
- 39:07Okay Suzy I want to thank you for your help because
- 39:12this is really valuable in terms of being able
- 39:14to show you all this without having to have 15
- 39:19screens up. Okay. Now, let me share the screen
- 39:26with you. Okay. Hopefully you can see the demo
- 39:38space here. And what I've done here is I've put
- 39:40all four things up here. I put Claude. I've put
- 39:47I've got GBT. I've got Copilot, I think is the
- 39:52other one. Oh, Perplexity. Sorry. I'll do Copilot
- 39:55as well. So there's five, I guess, in here. And
- 39:57then I have GCAI up here. And what I want to
- 40:01do is show you, I just took a sample agreement
- 40:04here. I'm going to put this over here. So what
- 40:09I did was I took this speaker agreement. So just
- 40:18regular speaker agreement. I sanitized it so
- 40:20you don't see any of the information in here
- 40:24that was part of the signed agreement I had obtained
- 40:28a copy of. But there are two things that I wanted
- 40:32to look at this in all the tools. And then there's
- 40:35one thing I wanted you to be able to consider
- 40:36in terms of what do I actually want in a template
- 40:40agreement? So there's two parts of it. Should
- 40:42I even sign this? What should I be worried about?
- 40:45one aspect. And then if I were going to make
- 40:48a template out of this, how would I improve it?
- 40:50What should be included? Okay. So that is the
- 40:52agreement. And you know what, Susie, I didn't
- 40:57put this in the materials, but I can add it to
- 40:59the material so that people can just see that
- 41:01better in our example. So let me move this out
- 41:04of the way. Okay. Let me go into the speaker
- 41:08agreement. When I asked Claude, what can be improved
- 41:12in the speaker agreement template? What's unclear?
- 41:15What's missing? What are the greatest risks?
- 41:18Okay. First of all, I asked four questions in
- 41:20the same prompt. Probably should have just done
- 41:22one at a time, but I wanted to just have it go
- 41:25through, give me the first thought here. So,
- 41:29biggest risks, ongoing. It's an ongoing agreement.
- 41:34It didn't even have any termination things in
- 41:36it. The IP clause gives the speaker the sole
- 41:38ownership, but the company helped create it as
- 41:41paying for it. Where does that go in between
- 41:45these two things? Liability is asymmetrical,
- 41:48and it cuts against the company. So if you're
- 41:53the company, you want a better liability. provisions
- 41:57in there. Choice of loss follows the speaker's
- 41:59address, not the company. So you might even not
- 42:01know what it is. Hopefully there's address in
- 42:03the agreement. No force majeure. Okay, then it
- 42:07says the language is unclear. And so it gives
- 42:09me five areas in here where it's unclear. So
- 42:13you can see right here. the liquidated damages,
- 42:17it's one night of lodging, you know, what happens
- 42:19if it's more than that. The mileage reimbursement
- 42:22rate was hard -coded as opposed to being a reference
- 42:24to the IRS standard rate. So there was something
- 42:27in there that was going to need to be changed
- 42:29every single time. Things that are missing. So
- 42:33insurance specifics. It talked about dispute
- 42:35resolution mechanisms. It talked about exclusivity
- 42:38in terms of the content. a lot of problems or
- 42:43a lot of things to add if I were going to be
- 42:45in the template. So it says it's useful. I can
- 42:49turn this into a red line version. So if I had
- 42:51a playbook or I had a template already and I
- 42:54was getting paper from the other side and I wanted
- 42:56to compare it against my agreement instead of
- 42:58forcing my agreement on them, this is a good
- 43:01red line tool. Okay. So I want to go back to
- 43:05this agreement because I want to showcase the
- 43:07next two tools. Let me move this one back over.
- 43:12Okay, let me open this in a different way. Let's
- 43:17cover that. Here's what I want to show you. Co
- 43:20-pilot. Too many things on the screen. Co -pilot.
- 43:26Okay, so this is a Microsoft 365 tool. So if
- 43:31you've got those tools in your office, you've
- 43:32got this at your fingertips already. So let's
- 43:35do the same thing in here. Describe what you
- 43:37want to edit. Let's... Let's edit your document.
- 43:40I'm like, oh, let's see. I want to go back up
- 43:44to the same prompt here. Copy that. And now I
- 43:51want to go back to the word. OK, so let's type
- 43:54the same thing in, OK? Same thing in Copilot.
- 44:02And it's not, it's not going to take things.
- 44:05The thing about these tools are so powerful,
- 44:06they don't take very long at all. Okay. Lining
- 44:08things up. I'm going to look for that. And then
- 44:11while it's doing its thing, I'm going to tell
- 44:13you about the GCAI. So right up here, you see
- 44:16where I can, I've got a extension. Oops. I can't
- 44:26restart it. Okay. I had all these set up, of
- 44:30course. And then when I go to use it, it doesn't
- 44:31want to do it. So now you see two little tabs
- 44:33here, GCAI for Word and Copilot. These are two
- 44:37choices in Word right now, in Word that I can
- 44:41use to edit my documents or help me. And these
- 44:44tools are great. OK, it's going to ask me. I'm
- 44:48using my company name. This is the GCAI. Sorry,
- 44:52I'm being back and forth because this was not
- 44:53done yet. So let me go back over to GCAI. It
- 44:57gives you some document information for who I
- 44:59am, who my company is, company profile. I've
- 45:03got PractiGal in here. I have not attached this
- 45:06to a project. So let's just continue. And we'll
- 45:09ask it the same thing. How can I help you today?
- 45:13I'm going to cut and paste and put the same thing
- 45:15in there and see what it gives me. Okay, now
- 45:18you see up here in this tool, I'm in the chat.
- 45:20It also has playbooks and then it also has settings.
- 45:23So all of you guys that asked where my privacy
- 45:25is and security and I am a closed system, this
- 45:29is where you can set that and I'll go to that
- 45:31in just a second. So right now it's generating
- 45:34a response. So let me go back over to Copilot
- 45:36and see what it says. Let's make this bigger
- 45:39and we'll see how. What's it telling me? Okay,
- 45:44so this triangulates, right? Add mutual cancellation
- 45:47rights because that's the same thing Claude told
- 45:49us, right? Company name. This is what I love.
- 45:52It gives me language that I can consider. So
- 45:55think about this for a minute. If I'm a new legal
- 45:59intern or I am a first year associate, I don't
- 46:03know what I don't know. So I could look at this
- 46:06red line and say, oh, that looks good. It sounds
- 46:08better than, you know, what I would have drafted
- 46:09or then what's in here. Apply. Let's just apply
- 46:13it. It's going to line things up and it's going
- 46:16to red line this and it's going to give me the
- 46:17other language. If I didn't know if I weren't
- 46:23a lawyer for so long, I might just look at that
- 46:26language and I'd say, oh, that's great and use
- 46:27it. So remember, I talked about judgment being
- 46:30important and talking about all of our experience
- 46:33is what's irreplaceable. This is where we come
- 46:37in and use our judgment and experience to say.
- 46:40Okay, well, I don't know why you put it here
- 46:43as opposed to the very beginning, like up here,
- 46:46but we'll fix that. We need our judgment to look
- 46:53at this language that the tool suggests to see
- 46:57if it makes any sense for us to use. So that
- 47:02judgment and that experience that you have is
- 47:07invaluable and coupled with a tool that can help
- 47:10do this quickly or measure it against one of
- 47:13our playbooks, super important. Okay, so we can
- 47:16go down and we can look at all of these. What
- 47:20do we have left? We have 10 minutes left. We
- 47:22can look at all of these paragraphs one by one,
- 47:27which is what I recommend. I recommend the overview
- 47:29and then I recommend going for the ones that
- 47:31you're gonna change. Let's go back over to GCAI
- 47:33and see what it turned out. Okay. IP ownership,
- 47:37right? Governing law, no remedy if the speaker
- 47:39cancels, ongoing. So very similar results to
- 47:42what Claude gave us and what Copilot is giving
- 47:45us. So I want to say these give us a good checklist
- 47:49to go back in and fix this agreement, right?
- 47:53So I want you to see those tools, these two,
- 47:56Copilot and GCAI, because they're right inside
- 48:00as an added app to Word, okay? Very powerful.
- 48:06Now I want to go into GCAI and I want to see
- 48:11the settings because I want to answer those questions
- 48:13you guys had. Not that it's going to be the all
- 48:16in all of all this, but if you want to auto apply
- 48:20changes, you can turn that on. If you want to
- 48:22auto jump to the changes, you can do that. And
- 48:24it's put the context of who I am in here. Okay.
- 48:29That's all the settings I'm going to get for
- 48:30this particular iteration of the tool. And then
- 48:36if you have plan B, one of the pain points of,
- 48:38like, those come in, why should I have to spend
- 48:39more time on this? Need to train the paralegal,
- 48:41need to train the business people to use these.
- 48:43This is one of the ways to do it, where you have
- 48:45a template that you've approved and that they
- 48:49can use. And it would run like if they got one
- 48:54from the other side that's not yours and they
- 48:56got one that they want to sign, then you can
- 48:58compare it to your playbook and say, but you
- 49:00got to fix these three things in there before
- 49:01you can use their agreement. Okay, so this is
- 49:04one way to use that tool. So let's go back to
- 49:10perplexity. Okay, perplexity came up with very
- 49:14similar things. Payment cancellation is the number
- 49:17one problem. Recording rights conflict with the
- 49:20IP ownership. So you see how these tools are
- 49:23giving us similar results, but not the same thing.
- 49:26That's because the tools actually operate differently.
- 49:30Let's see, what else do I want to show you in
- 49:32here? Okay, this did not give me a red line,
- 49:38but I can ask it for one, right? Does it want
- 49:41me to draft a revised agreement? Do you want
- 49:44to create one with a writer? So I tend to put
- 49:48something like that as an ongoing agreement.
- 49:50I would put a statement of work in there that
- 49:52could just say, This event is specific, here
- 49:55are the details for that. This other event, all
- 49:59the details for that. So that's good if you've
- 50:01got ongoing obligations. But similar results.
- 50:05So I gave it, you know, what's missing. I gave
- 50:07it, you know, what's problematic from an enforcement
- 50:10standpoint. I got smarter as I went from Claude
- 50:12to chat to GCAI to co -pilot. And so I asked
- 50:16a deeper question when I went in here to use
- 50:20perplexity. So let's do one last thing in GCAI.
- 50:26I want to go to, let's see where it is. Let's
- 50:30go down here. So GCAI is a paid tool. It's about
- 50:34$500 a month or $5 ,000 a year. So it's not cheap,
- 50:37but it is worth the money in my opinion. And
- 50:40I'm just doing a trial run here, which is they
- 50:43will let you test out the tools, which I highly
- 50:46recommend. So let me take a minute to answer
- 50:48some of the questions where you guys said, I
- 50:50can't get answers out of vendors or I don't really
- 50:52understand how the tool works. I don't understand.
- 50:54Is it a closed system or open system? I don't
- 50:56understand who can train the tool. Who can? What
- 50:58information are they using? Some of the tools
- 51:01that you use, the temporary chat feature, which
- 51:03is kind of like a Snapchat, it disappears after
- 51:05you close that. That's one way to do it. So the
- 51:08information is not saved and not used across
- 51:10the back end. But let me talk about, I wanted
- 51:14to go to one thing, company profiles. Let's see.
- 51:19I probably should have clicked on that in the
- 51:21middle of this. Oh yeah, it goes to that. Okay.
- 51:24Let me go back to the chat and the project we
- 51:28were working on. okay risk assessment guide this
- 51:32is great this is keeps keeps my stuff in here
- 51:34and it can assess an agreement and you can keep
- 51:38it and you can go back to it let's say i i didn't
- 51:40actually make the changes to the agreement now
- 51:42i really need to make the changes to the template
- 51:44it gives me all of my history and what i want
- 51:46to find out in here about what i did before so
- 51:48now i can look at it and say oh i need to actually
- 51:51uh give the business people a better checklist
- 51:54to fill these these fields in as they go through
- 51:57before it comes to my desk. So I know this was
- 52:01a quick overview, but I wanted to just show you
- 52:03how the tools can work and how you can get different
- 52:05results in them. So let me stop sharing and we'll
- 52:09go back to the slides. Gosh, too many windows.
- 52:19Okay, so let's go to The cautionary notes. Okay,
- 52:26so what are these tools good for? What are they
- 52:27not good for? You know, I showed you all of them.
- 52:30I talked about the temporary chat. You can turn
- 52:32off the learning mode in your data control. Sorry,
- 52:36I couldn't go through all of those little nuances,
- 52:39but I just want to say the variations and prompting
- 52:41can change your results and we can get smarter
- 52:44as we do more research and do more iterations.
- 52:47You can break it down, give it all the context
- 52:49you want and use the legal specific tools for
- 52:52legal work. Remember the crucial side of that
- 52:54spectrum. Use something like GCEI. Okay. Like
- 52:58I just used it. I'm going to Portugal at the
- 53:00beginning of September. So I used it because
- 53:01I needed some activities on two of the days.
- 53:03And I said, I have time here. I'm going to be
- 53:05in Braga. What should I do? I'm going to be in
- 53:07Porto. What should I do? What should I not miss
- 53:09that's unique to this thing? And I like good
- 53:11food and I like museums. And it gave me a bunch
- 53:14of stuff. So now I got to go wade through and
- 53:15do the research. But these tools can be used
- 53:17for something like that, which is the minimal
- 53:21impact all the way up to documents that you would
- 53:24have to file with the court. Let's talk about
- 53:27what it's good for and not good for. It's good.
- 53:39Go one before that. There you go. So now that
- 53:42you've seen a few tools, what do you think AI
- 53:46tools are going to be good for in -house now
- 53:50that you see some of these things? I personally
- 53:52think it's good to try them out and see how they
- 53:56work across the board. You definitely have to
- 53:58sanitize your agreements, which is what I did
- 54:01in that speaker template. You definitely need
- 54:05to, if you don't want it saved, use the temporary
- 54:08chats and then also go into the settings and
- 54:12just look at them of what you can do and how
- 54:15you can control it. But these things on the left
- 54:19there, are good to start with. So automate repetitive
- 54:23tasks. So if you do summaries, outlines, case
- 54:27status, any of those kind of things, totally
- 54:30great to use it on. Repetitive tasks like scheduling,
- 54:34any of those kind of things, good to use it on.
- 54:36Really good if you're leveraging templates and
- 54:39playbooks. So you've commonly used agreements
- 54:41in your company specific to you or your internal
- 54:44clients and the groups you support. I would make
- 54:46sure those templates are all in here and that
- 54:48you've trained the system to be you and how you
- 54:50think about evaluating those agreements when
- 54:53they come in across your desk. Streamline communications.
- 54:58Okay, it's really good to use to manage expectations.
- 55:01It's, like I said, to make Sonia friendlier.
- 55:03Use plain English, especially if you're communicating
- 55:06with clients instead of the legalese that you
- 55:08might get back from outside counsel. Now you
- 55:09got to interpret that and send it on to the business
- 55:11people. Those are good places to start. Okay.
- 55:16Where is it not good? Okay. It lacks judgment.
- 55:19It lacks empathy. It lacks any strategic thinking.
- 55:23So any of those things where those are important,
- 55:26I would not recommend using it. Or it's going
- 55:28to require just enough time for you to review
- 55:30all the stuff. Hallucinations. Remember I said
- 55:33generative AI tools? Generate the information
- 55:37as opposed to looking at what you've got and
- 55:39then try to categorize it and make sense of it.
- 55:41These hallucinate. They make up stuff. They want
- 55:44to give you a response, even though there's nothing
- 55:47in the database that relates to them to be able
- 55:49to give you good information. Hallucinations.
- 55:51That's why I always ask for for sites and references
- 55:56so that you can go look at the primary information
- 55:59as opposed to AI information. Second, it's got
- 56:03a bias built into the training tools, okay? And
- 56:06then ethical concerns, you guys all asked about
- 56:08that. Those are things to consider. We could
- 56:13do a whole other thing on how do I use it from
- 56:15a privacy IT standpoint. And I say work closely
- 56:17with your IT department to make sure they're
- 56:18vetting them. However, they want to vet the tools.
- 56:21Okay. So please use these tools responsibly.
- 56:24I'm not going to say like use, you know, operate
- 56:26heavy machinery on drugs, but it's that same
- 56:28kind of an admonition of use these tools responsibly.
- 56:32I don't want you to end up as a case, you know,
- 56:36that, you know, ends up on the front page of
- 56:38the Wall Street Journal. Okay. Let's go to the
- 56:40next slide. And then I'll have, we'll get exactly
- 56:44through everything we planned on. So what should
- 56:48you be concerned about using these tools? Whether
- 56:50it's chat, GPT or GCAI. So we talked about privacy,
- 56:55security, accuracy, ethics, all of that. And
- 56:58also prompts, I'm gonna reiterate. Context is
- 57:01everything. So garbage in or garbage out. You
- 57:04have to give it the context or you're going to
- 57:06get garbage. Okay. I always offer quotes or paraphrasing.
- 57:10So I've put my own books in here. And when I
- 57:13write an article, I'll just turn to a page and
- 57:15say, let's write an article today on this topic.
- 57:17And then I'll give it some information and I'll
- 57:19say, write me a, you know, 250, 500 word article.
- 57:22And then it'll give me some quotes. And I'm like,
- 57:24I don't remember saying that. Is that from my
- 57:26book? And it'll say, oh no, it's, it's similar
- 57:29to something. And you would say, and I'm like,
- 57:32OK, so you have to watch that if it gives you
- 57:35quotes, is it paraphrasing or is it actually
- 57:37can you go to the primary source and find that
- 57:40quote? OK, reset. Sometimes the tools get out
- 57:43of whack. You've done so many reactions. Now
- 57:45you're getting what I call gobbledygook earlier.
- 57:49Sometimes you have to just do a hard reset and
- 57:51start over if you're not getting good results
- 57:53or your desired results. Just say, let's just
- 57:55start a new start a new chat or let's start a
- 57:58new project. I cannot say this enough, use legal
- 58:02tools if you're doing legal work, okay? So that
- 58:05GCAI tool or other ones, use that when you're
- 58:08doing more things higher up on the crucial impact
- 58:12scale. Okay, last, judgment, trust but verify.
- 58:15You know that military saying trust but verify?
- 58:17I'm gonna say it three times, verify, verify,
- 58:20verify, okay? The information if you're relying
- 58:23at. You will be replaced by the... These tools
- 58:27anytime soon, but you may be replaced by someone
- 58:30who knows how to use these tools to their to
- 58:32their advantage. So let me just chat about next.
- 58:35The key takeaways I want you to remember know
- 58:37your why why remember I said why are you using
- 58:40this tool evaluate the tools and. It is fundamentally
- 58:44a colossal experiment and iteration. It is not
- 58:47a one and done. It is not an easy button, but
- 58:49you have to use them, practice them, try them
- 58:51out. Is it giving you good results? Is it not
- 58:53giving you good results? If you're going to adopt
- 58:55a tool in your practice, I would say know your
- 58:58why is important because that will dictate everything
- 59:00else that comes from it. Okay. I think I've answered
- 59:04questions. We've run out of time. I want to give
- 59:07you, could you go to the slide with the QR codes?
- 59:12Okay. Now, if you need help for these, I've given
- 59:16you a couple options here. One is an AI audit.
- 59:19I offer that to webinar attendees. So three sessions,
- 59:21evaluate your practice, how AI can help you or
- 59:23not. And then if you need help personally, like
- 59:26you want more coaching, consulting, whatever
- 59:28that is, the Let's Talk will get you time. You
- 59:31can schedule a meeting with me. I do 30 -minute
- 59:34sessions for free a couple times a week. And
- 59:38both of these QR codes will get you time with
- 59:41me. Go to the last slide, please. And then we'll
- 59:46wrap it up. Here is my contact information. I
- 59:51want to encourage you to use these tools and
- 59:54see what's out there. I want to thank the Beverly
- 59:56Hills Bar Association. I want to thank Belinda
- 59:58McCauley for inviting me and Susie for your invaluable
- 1:00:01help behind the scenes. You guys are so lucky
- 1:00:04to have these amazing resources and the Bar Association
- 1:00:08with all the education they make available. This
- 1:00:10is being recorded, so it'll be available later.
- 1:00:13All the materials I talked about are available
- 1:00:15with it as well. Please connect with me on LinkedIn.
- 1:00:18and let me know you saw this presentation, so
- 1:00:19how you found me. And here's my contact information
- 1:00:22if you need more help with any of this. And let
- 1:00:25me see. I think, please post your email. OK,
- 1:00:29that was the last one. OK, emails there. You
- 1:00:33can email me. You can look at my website. You
- 1:00:35can look at Linktree. It's got all links to everything
- 1:00:37that LinkedIn will be the best way to find me
- 1:00:39as well. And you have my phone number and my
- 1:00:42email. Thank you so much. This concludes today's
- 1:00:48presentation. Thank you so much, Sonia, for an
- 1:00:50excellent presentation today. And thank you so
- 1:00:52much to our audience for joining us. Have a great
- 1:00:54afternoon, everyone.