00:00:14.8800000 Hello everyone and welcome to today's presentation. Today we're going to talk about AI at the deal table. Ethical and practical guidance for Real estate professionals and attorneys. So some of you may be realtors and attorneys and kind of just that interconnection of how maybe we can potentially use AI in one area but maybe not the other, or where those risks really come into play. And as anything with Genai or AI, it is an emerging area. 00:00:44.1600000 So it's always good to stay on top of things, but also know that things can change in this landscape pretty quickly. So we also need to be cognizant of that a few housekeeping items as you always get logged in and settled with your volume. If you want to know when you get that CLE code for your CLE credit, that's going to come at the end of today's seminar within the last 10 to 15 minutes of the seminar here today. LexVid will auto populate that into the chat feature where you can write down that code. There will also be instructions in there on what to do with the code to get your CLE credit. 00:01:21.9200000 Now you need to write down that code exit out of today's seminar once it has concluded, then you enter into your individual lexfed account, enter that code in where prompted under Live seminars or the seminars you are attending and then hit Enter and it submits it back to LexFid. LexFid then has 30 days to take your attendance and turn it around to your respective jurisdictions who then put it on your audit report or log it for you however your jurisdiction does it. Now again this process does take some time because LexFed has that 30 days to turn it around and turn into your jurisdiction. So if you don't see it on your audit report or on your account with your respective jurisdiction within the next 30 to 45 days maybe then reach out to Lexfed to see where it's at in the process or if there's anything that has caused a hiccup or an error along the way. Now don't worry too if for some reason there is an error, lexfed does record these and your attendance is recorded here today and how long you stay when you sign on and when you sign off so they can verify your attendance if something does happen and they need to double check that as well. 00:02:31.7500000 But again that CLE code comes into that chat feature at the end of today's seminar where you can write it down. From there I will also announce it verbally and and repost it myself in the chat feature just so you have Multiple opportunities and no one can say they did not get the code. Aside from that, if you have any questions throughout today's seminar, I do encourage you to ask those questions so we can hopefully get them answered. If you don't want to ask them live here today, then you can always email lexfid after the seminar, have them ask, ask them the question, and then they submit it to me for a response. But if you do have questions live, you can put those in the chat feature as well. 00:03:12.7800000 I will try to monitor the chat feature as I'm speaking, but I may not get to your question until the end of the seminar. And that's just for flow and consistency in giving the presentation itself. But please do ask those questions and put them at the chat feature if you do have them. With that, if you have not taken a seminar with me, my name is Kerri Sheehan. I'm assistant General counsel of Taft, Stentinius and Hollister law firm located in their Indianapolis, Indiana location. 00:03:41.0600000 In my role as assistant general counsel, I do all things ethics, compliance, anything that touches or concerns the rules of professional conduct or anything compliance internally with the firm, which includes generative AI and the ethics for lawyers and ethics in using it for our clients, especially if we do real estate or realtors. That's one area where, where this could come into play. So today we are going to explore some of those nuances if you are a real estate attorney or you're a Realtor or both, and how the implications may be a little different depending on what role you're doing or what hat you're wearing in your practice. And maybe how you advise your real estate clients too on their use of gen AI. Because I don't know if anyone's out buying a house nowadays or, you know, looking at real estate. 00:04:34.7500000 But realtors, it's not just attorneys, but realtors and other professionals are using generative AI. Now, the thing is, hopefully within those professions, they have their own seminars where they can learn about the risk and the benefits of generative AI. But as lawyers may be counseling them, we also need to be aware of of the risk and benefits and how our clients can use certain things. So we're going to talk about some of the risk associated with lawyers and then if those risks parlay over or go and flow into the risk associated with clients in this seminar, particularly real estate clients or real estate professionals. Now, just to refresh your memory, if you have not taken a generative AI seminar before, artificial intelligence is a thing that is, you know, been around for a long time it's the automation processes, those virtual assistants, speech recognition. 00:05:31.5600000 It's also the things we get so annoyed with, like say we call a doctor's office and it says press 1 for this or say this name and it never knows what you're saying. That's a form of that artificial intelligence. The kind worm we're talking about here today is the new generative AI that started emerging and followed 2022. Really robust with chat. GPT is the large language models, the things that kind of think and act or feel like they're thinking and acting like a human being even though they're still a pre trained transformer. 00:06:05.2300000 That's the type of generative AI that we're going to be talking about here today. Now also note too there's a third kind of AI. It's called agentic AI and it is really starting to emerge right now. I can see agentic AI really being more the forefront over the next year, two years, even five years because it's your automated workflow, your AI, agents and assistants. It can also be those self driving vehicles. 00:06:30.0100000 So there's three types of categories of AI. And what we're going to focus on here today for you as attorneys or your real estate clients is the generative AI portion of it. So just kind of keep that in mind. Some other definitions to keep in mind today too as well. If you hear me say the word hallucination, it just means something that generative AI is made up bias and discrimination in regards to AI. 00:06:56.1600000 Now it's not a human being, but it can give bias or discriminatory results because of how it pulls its information. So again we need to keep that in mind too when we're talking or having this conversation here, this lecture today about AI. We also need to remember confidentiality, breaches, obtaining that sensitive information and that we don't want to ever feed sensitive information into an AI unless certain parameters are there, which we'll talk about. We're also going to potentially talk about some IP or security risk, lack of explainability and then just too that self learning feature because a lot of your clients, especially potentially real estate clients or real estate professionals may not have what is known and that we'll talk about enterprise versions of these generative AI things. They may not have the subscription version of like a contract thing. 00:07:48.4000000 So they use just a large language model like ChatGPT to help draft their client's contract or to help draft that description for the house listing or to help draft something in real estate. Now the problem with that, we'll See is the confidentiality is the cause for errors in different areas. But then we're also going to talk about how can you, as a real estate professional or an attorney advising a real estate professional, advise them of how to use generative AI in a safe fashion for their clients as well? We do need, as lawyers and as non lawyers or the professionals that we counsel need to understand AI. It's affecting everything. 00:08:33.0400000 It's affecting every market, every profession, and it's not going away. It really is not. So we need to take our knowledge of how it affects us as lawyers and be able to translate that into how it affects our clients industries or our side hustles. Maybe you're both lawyer and real estate professional and know where the parameters are in each type of area. So again, we really need to think about it. 00:08:59.9900000 As a lawyer, some of the ethical implication comes from the rules of professional conduct. Now obviously, real estate professionals are not subject to the rules of professional conduct, but us as lawyers are. So when we are helping real estate professionals, us doing the work for them, we have to remember that they're our client and that we cannot breach any of the rules when doing that work for them. Some of the rules that could potentially be breached with the use of generative AI are the rules on the screen in front of you. The first one is competency rule 1.1. 00:09:34.8100000 Under competency, we have to remain competent in anything that touches or concerns our practice of law that affects our clients, including relevant technology. Relevant technology is generative AI. So we need to make sure we understand how it affects us and how we can use it as lawyers to help our client. But then also too, if our client is that real estate professional, since that's what we're focusing on here today, how they can use it competently and how we can advise them to use it competently. Now, if you think of some of those examples, maybe some of it is tell them, you know, that description when they're writing the MLS listing or the listing maybe plugging in there, you know, can you revise this description or edit it or make it sound better? 00:10:20.9000000 But then also too part of that competency is also telling them that they can't lie or be misleading in those descriptions. They have to follow their own ethics as Realtors and they have to, you know, disclose the certain things. And so just kind of coming, you know, into that area where you're having a collaboration on how you can use it safely, we also need to, to communicate with clients, us communicate with our clients, and then potentially your realtors or your real estate professionals you're working with communicate with their clients that they're going to use AI. And that's because there's some risk associated with it. So even though the Rule of Professional Conduct 1.4 only applies to US lawyers, it's also good to tell your real estate professionals you're working with that they should tell their clients if they're using AI to help generate the listings, help create that contract, help, you know, do the escrow or whatever it is if they're using it. 00:11:21.3000000 They also need to communicate. Now they may not be bound by Rule 1.4 like we are, but they still should have communication and best practices with their client. And that is because a lot of generative AI models don't keep confidences. So that goes into a 1.6 for lawyers and then also realtors. They don't want to be listing all their homeowners or their clients private information out on the website, like you know, bank account information or money things or different things. 00:11:53.2600000 But if they're putting it into a generative AI model, they could be doing exactly that. And we have to understand how the generative AI model works to understand if they're breaking confidentiality for their clients, just like we would break confidentiality for our clients. Now that assessment does come down to the certain search engine or the certain generative AI model you're using. For instance, Chat GPT is a self learning AI model, meaning anything you input into, keeps it, retains it, stores it, learns from it, makes new connections. Meaning if you put in something like the sky is blue and it didn't know that, it's now going to make that connection and then the next person that ask a question where that connection is relevant, maybe someone asks the same thing, what color is the sky? 00:12:48.1200000 And and if it took your input and then it could push out what you put in out to a third party. Same thing with client information on our clients as well as real estate professional clients. Anything we input into these large language models that are public facing, meaning we don't have an enterprise subscription. I don't mean the subscription that's $20 a month that any Joe Schmo on the street can get. You have to have a subscription that is for a business and a contract that specifically states it's not going to store and self learn on your information. 00:13:29.7600000 Just having the subscription that's available to the public does not guard against these confidentiality concerns because you're subject to all the same terms and conditions as that person who's using it for Free. You're just paying for an enhanced model, but not enhanced security features. Unless you get that enterprise version with the contract from a representative of the company and go through all those bells and whistles. And this is something your real estate clients who are using it for their clients need to understand. Because I've talked with some real estate professionals, some realtors and others who use it to draft or get forms to draft contracts like for real estate deals because they don't want to hire a lawyer anymore. 00:14:18.2400000 So then they go to chat GPT or they go to Perplexity, or they go to a large language model and says draft me a real estate contract, here's the address, here's the party's information, here's, you know, the personal bank accounts or whatever needs to put in there. And they feed it all to this large language model and not even realizing or not even comprehending because they didn't read the terms and conditions that that large language model is retaining that information, learning from it, storing it, and could potentially push it out to unknown third parties. Now that is obviously problematic for us who are bound by Rule 1.6 confidentiality, but it's also problematic for people in other professions like real estate who gets client confidential information like the addresses, the bank accounts, the escrow, all those things and then just decides to willy nilly put it into chatgpt or somewhere, not even thinking that it could go out to third parties, not even thinking that it's not just deleted right away because it's not, it's saved, it's stored, it's utilized to give other answers to unknown third parties. We have no control over the data that's put into these large language models that are public facing and no control over where it goes or how they use it. They could be selling it to the government, who knows. 00:15:42.7000000 But again, we have to read those user terms and conditions, we have to know where the information goes. So while us as lawyers are bound by Rule 1.6 confidentiality if you're a real estate professional or advising a realtor or real estate professional if you're, they should consider confidentiality in terms of the client information they're getting from their clients, buying and selling properties or whatever they're doing to help them in that context. So we need to remember that for us lawyers there's also other implications that may not be relevant to real estate professionals because they're not in front of a court. We are, but we're also bound by candor to the tribunal, fairness to opposing parties, truthfulness to third Parties and supervision. We can't false advertise and we can't do anything under 8.4D that's prejudicial to the administration of justice or 8.4G, as in girl bias or discriminatory while we're practicing law. 00:16:40.1500000 Now you may think, how do all these rules at the bottom of the screen really come into effect? With generative AI hallucinations. If it gives us a hallucination going back to that definition page, it means it's something false. It it's telling us a lie, something that's not true. If we turn around and give that to a court opposing party or a third party, we violated these rules. 00:17:03.9600000 So we always have to know to double check the information so that we're not giving out false information to other parties. Now, while real estate professionals or your realtor clients are not bound by these rules, they still need to think of this in that concept as well. Because when you think about it as a realtor, if you ask ChatGPT for something or a contract or you know, kind of that purchase agreement and it gives you a form, but in the form it cites law or it cites to a statute and that's wrong, you have your client sign it, you send it to the other side to buy the property. But the terms and stuff in there are false, made up or hallucinated. That's a problem. 00:17:47.0800000 It can happen to anyone. It's not just lawyers where the hallucinations really take effect, it's also other professionals. And so we always have to double check anything that a generative AI model gives us because it could be wrong and we could violate these duties of truthfulness. Our real estate professionals could violate duties of truthfulness to their clients or other parties too and make the deal go south. So again, we have to double check everything it gives us to make sure it's using proper terminology, using proper cases, if that's relevant, using proper statute citations or proper law. 00:18:24.0600000 If you're doing a purchase agreement or some type of contract too and using it for that, and then on top of that we can't even put in there confidential information because it's self learned, self trains. And so the best so far in looking at all these, whether you're a lawyer or a real estate professional, is to use it for things that are generic. Meaning if you want a form, you say I want a form purchase agreement for a property in X state. Great. There's no client information in there for attorney or for real estate professional. 00:18:57.2200000 There's no confidential or private information. You're not asking necessarily for anything that would violate that. So it's going to give you a form. Now with that form though you would still need to double check it to make sure it's relevant to the state you're in and includes all the bells and whistles needed. So there still is oversight. 00:19:15.4400000 There's always human oversight, but we're not violating anything in the rules for attorneys or in the considerations for the real estate professional and doing things that are generic. If we go to more specifics, that's when we get into private confidential information or other areas like that. And we really need to not do those things for us lawyers too. Obviously if it gives us something false and we put it out there, it could violate the advertising rules. We're also responsible for our non attorney assistance and those we supervise. 00:19:49.1190000 That's rule 5.1 to 5.3 and also 5.2 in there and then that prejudicial to the administration of justice. Anti bias, anti discrimination under 8.4, generative AI can give you bias or discriminatory results. Meaning if it pulls from data that's discriminatory or one sided, you're going to get that result. Just like anything. If you only read books on one side of an argument, that's what you're going to believe because you don't have anything to counteract it. 00:20:17.2200000 Sometimes generative AI doesn't pull from everywhere it has like a swimming pool and inside that swimming pool is the information it can pull, pull results from or make connections from and then give those connections out to you, the user. Now the bigger the swimming pool, the more broader the connections, the more information it can pull from. The smaller the swimming pool, the smaller the data still going to make connections because it's programmed to give you an answer, but there's a higher likelihood that that answer is going to be false, misleading or discriminatory. Especially if it's one sided content it's pulling from say for instance, there's a generative AI out there that only pulls from say conservative viewpoints or conservative websites. Well, that's the responses you're going to get. 00:21:09.2300000 And so we do have to watch out for that too and learn whatever search engines, whatever, excuse me, large language models we're using, or whatever we want to use. Not only do we have to look for does it self learn, self train and where does it store the information, does it give it up to third parties, all those confidential things, but also too, where does it pull its data from? If it only pulls from California law and you're an Indiana lawyer or Indiana real estate professional, then you don't want to use it. And so you've got to read the user terms and conditions of these generative AI platforms to know what you're getting, where the information is coming from, how it uses the information you input into it, and be competent. That overarching competence that we need to think about too, this discrimination too. 00:21:54.2200000 And different things can come in with in real estate, especially if there's different fair housing things or different things that could come into play with that or it pulls data. If you're trying to do something in that area that's wrong or inaccurate, maybe you're trying to do percentages, maybe you're trying to fill out like a status sheet or something, you need to look where the information comes into play and think about those things. Now, with all these risks associated with the public facing generative AI, there is what is called enterprise, generative AI. Enterprise, like I said a little earlier, are the types of generative AI platforms that you pay for. You pay money, and I mean big money for you're paying per user. 00:22:39.7800000 It's not just $20 a month, you're talking like hundreds of dollars a month per user. You have a signed contract, not just what's on the website, but a signed contract with the provider with terms and conditions. And you talk through the security features. Does it store your information? Does it not? 00:22:58.8500000 Can you turn that off? You know, does it stay within the four walls of your company? Those are things you want to look for. But most enterprise versions, the data you put into it doesn't get stored, it doesn't self learn, self train and it doesn't send it out to unknown third parties. If you put something into it, it stays within the four walls of your company. 00:23:21.8100000 Now it does have a swimming pool outside those four walls of your company, unless you choose otherwise, where it can pull data just like any public facing generative AI. But it doesn't send out your inputs, which is the biggest thing when it comes to that confidentiality, security and making sure we meet all those requirements for our clients or for the real estate professionals to meet them for their clients. Now a lot of law firms are starting to get these enterprise versions. You can get them from different platforms such as Perplexity, Claude Copilot, Westlaw AI. If you're a lawyer, LexisNexis protege, you can get ones that are safe to use from that confidentiality security standpoint when it comes to doing legal work. 00:24:10.5300000 Now you can also purchase these enterprise versions and different types of Generative AI platforms if you're a real estate professional. However, some of the ones that you may want to think about if you're a real estate professional are not necessarily your large language models like seen on your screen, although these are helpful. But there's one specifically designated for contracts, for comparing contracts, for writing contracts, for, you know, doing certain terms. One of them that comes to mind is Lagora is a platform that has like a contract business component of it. And it's not just more general. 00:24:48.0300000 So you want to look at the different platform that's going to best fit your needs. If you're a lawyer, it may be a more general one, unless you're just a contract lawyer and want more specific to that, or both. If you're a real estate professional and just dealing in land contracts, purchase agreements, sales agreements, those things, then maybe you just want to shift towards the ones that are specifically designated for that. But we want to make sure that we have those enterprise versions so that our information of our clients is not going out to unknown third parties. It's not storing our information and it's not using it to self learn, self train, but look and vet the programs you're using. 00:25:30.9900000 None of us like to read terms and conditions. None of us. They're boring, they put you to sleep. It's. You can even say, I'm going to sit down and do it while I'm on a plane, but then you're probably going to be snoozing after a few minutes. 00:25:42.5600000 But we need to read them to vet our products because we need to find the appropriate tool to fit the needs of whatever we're doing. If we're that lawyer, we need one to fit the lawyer needs and the rules of professional conduct. If we're a real estate professional or Realtor, we need one to fit what we're doing day to day, keeping in mind some of those ethics concerns that lawyers have, but in the context of real estate and a lot of Realtors don't think of the ethics. I know they go through classes, I know they do those things, but then they just don't think of, oh, I'm putting in my client's Social Security number into a large language model that self learns, self trains and stores that that could be a problem. And so we need to stop and think about what we're doing, whether you're a Realtor, attorney or whatever it is, and understand the platforms we're using. 00:26:32.4800000 Don't trust the public opinion, don't trust Reddit on what to use, Read the user Terms and conditions, understand it for yourself and have that competency. But in picking that program to purchase that enterprise version and picking that program to see which one best fits your needs, fits all the parameters of security, fits all the parameters of ethics, you need to first, as you can see kind of from the flowchart here, figure out what information you're going to process through it. Meaning, are you just putting in contracts, Are you wanting it to do generic forms or just description listings? You know, just that you're going to list on a website where no confidential information is going in. If you want something generic just to give you kind of that blurb that's going to go on the MLS or the Zillow or something like that, you don't really need an enterprise version for that. 00:27:25.3400000 So long as you don't put confidential information in and you just tell it to make, you know, more descriptive or make it more friendly, or make it more, you know, whatever, then you don't need the big enterprise version. But if you are doing more contract review, contract terms, contract negotiation, wanting it to review that information, wanting it to generate contracts for you, putting in the client's information into it, then you probably need an enterprise version focused on contracts because you're putting in sensitive personal information. You need to think about those security features and ethics, and that's the kind we need. So then once we figure out what information we're putting in, we can kind of start categorizing, do we need more public, do we need more private? And then evaluating too, which one fits best. 00:28:11.5900000 Reading those terms and conditions for any platform, just don't think because you pay for it, it automatically doesn't self learn, self train. That's not always the case. Generally it is, but not always. So we still need to read those terms and conditions to make sure it meets our expectations, meets the ethical risk that, you know, are relevant to us, and then we choose the platform. Some other considerations that could go into that choice are going to be obviously consent of your clients. 00:28:40.3400000 Sometimes you may run into clients in law or in real estate. If you're acting as that realtor, you that say, I don't want you using Generative on my files, I don't want you using Generative AI to produce my purchase agreement. And guess what? In the legal world, the client's the boss. I'm assuming in the real estate world too, the client's the boss because they're the one purchasing the property. 00:29:04.7100000 So we need to have a communication with the client to get them to understand if we're going to be using these products, how we're using them, any risk associated therewith, and, and then get their consent to do them in lawyer land. That's mandatory in Realtor land, not always as mandatory, more a gray area, but a great ethical practice to do. We also need to look at the cost and understanding the price structure. Like I said, the enterprise versions cost money. Do you want one license for one person to use it? 00:29:35.2700000 2, 3, 4? A lot of them pay per user, meaning it's $400 a person for the year, it's XYZ for the year or XYZ a month or $400 a month. So you've got to also look at not only the security features, what you're using it for, but then look at the cost. And if the cost is too high, you can't just drop down to less security features and say, well, I couldn't afford the big one, so I'm just not going to care about confidentiality anymore. It doesn't work like that. 00:30:04.3800000 If you have confidential information, you need to either put up the price or not use it. So you need to think about those things too. But these tools can help with internal management, making things more efficient. But they should never replace a lawyer, they should never replace a realtor. We don't want people just to think this is their get out of, you know, law firm life or get out of job life free and go play golf all day and let the generative AI do the work and get paid for it. 00:30:33.9600000 It's not that because they can still hallucinate, they can still produce those false results or the bias results. So they always have to have human oversight. No matter what area you're in, if you're in law, if you're in real estate, no matter what area you're in, we have to have that human oversight. If you have that enterprise version and you can do it safely, thinking about all those rules of professional conduct and security risk. You can use generative AI for all kinds of things, summarizing depositions. 00:31:05.6400000 You can upload the deposition, tell it to summarize it, but again, you need the enterprise or safe one for the rules of ethics. You can have it do legal research, but double check that the citations are right. Westlaw, LexisNexis have generative AI products that can do legal research for you within their platforms. That's one of the safer ones or the most safe ones, because it's only pulling data from those two particular websites. Their data pool or their swimming pool is not outside the Lexisnexis or Westlaw data. 00:31:37.3200000 So they're not pulling from the Internet. So you're getting the Westlaw LexisNexis content, but you have AI now in it to help you get that content in a digestible fashion. But again a purchase price for it. If you use public facing ones for legal research, you're probably going to get false results, false citations, false case names, false connections or quotes from cases. It's going to happen, but so you need to know what platform you're using, understanding where it pulls its data from, understanding those things and then having that human oversight. 00:32:11.2100000 You can also use it for contract review. As lawyers like we've been kind of discussing presentations and drafting. It can do PowerPoints, it can do all kinds of things. You can also use it for more matter, trend spotting or knowledge management. Meaning if you want to know, like an analysis across industries you can find generative AI platforms and some of the large language models, the more generic ones can do this. 00:32:35.0200000 You could say find me trends of verdicts across the 50 states for the last 10 years in this area. So it's going to start pulling data and give you like a trend. So you can pull data trends, you can pull different things. And that's what it's also good for when we get to more the common use cases for our real estate professionals and realtors, finding those trends across markets, finding you know what's selling. Maybe it too is that comparison. 00:33:02.9000000 When you're looking to price a house or your clients looking to price a house, it can go to a generative AI and say what are houses in this area been selling for? And then the AI will pull the different sales and give you something pretty readily. I know like mls, you can do that by hand. Different things too. But a lot of realtors now are turning to more generative AI to help do this in a more speedy fashion. 00:33:28.0300000 So it's something we can do as we've talked about already too. Generative AI can help list descriptions, help with your marketing, drafting different chatbots too. Maybe you have a lot of influx into your website as a realtor, you know, people looking to get your services or ask about houses you have on the market. You can have a generative AI available or program to respond to them like a human being. Nowadays I know if you go on like Amazon and ask to ask to a customer service representative, first you get an AI chatbot. 00:34:01.1200000 Then if you want to talk to a human after painstaking stuff, you get transferred to a human. A lot of companies are using these chatbots to respond to website inquiries and to respond to different things. And they're pretty good at it if they're trained properly because they're thinking and acting like human. It's not like that old AI from the 1950s, 60s. This AI is interacting more readily. 00:34:25.6400000 So having those chatbots on your website, having those things can help too. It can also help with those market summaries, knowledge management, contract drafting, legal questions is another area where I'm seeing non attorneys use it. This isn't always the best because they can give you those false results hallucinations and the best is always still you to consult an attorney for legal advice which US Attorneys we want a job for that. But you will see some of your clients coming to you and saying well I asked Chat GPT already and they told me this. I asked Chat GPT that legal question, they told me this. 00:35:00.6100000 And so it's become the more I used to call them Google lawyers where they google everything and think they're a lawyer from whatever answer Google gave them these I call genai lawyers meaning they're using generative AI to get legal advice so they don't have to pay an attorney. I would say 75 to 85 to 80% the gen is wrong because one it doesn't understand the nuances. It can never replace the critical thinking of a lawyer. It doesn't understand the industry practices or what's relevant in that local community. It's pulling from a big swimming pool trying to make connections based on your input. 00:35:37.8600000 When it makes connections it kind of lights up and sends them to you saying I think this is your answer. I think this is it. This is what I'm seeing, this is the trend I'm seeing in my data. But it can't be that human being who's in the courtroom every day. It's not that human being that lives it and breathes it out in the field every day. 00:35:55.0600000 And so that is where these pre trained transformers or these generative AI models are really falling short is that it's exactly what they are. They're pre trained transformers. They're not humans. They don't have that critical thinking, they don't have that analysis. But they can help streamline processes, automate certain things like the summaries and the drafting. 00:36:15.8300000 But we still need that human oversight. We still need professionals in every area to give the expertise so that we know that we're getting the right thing. Jenny I like I was kind of saying there it's good, the bad, the ugly, it's good for drafting Summarizing tone. I like to use it for generic stuff saying tone this letter down. Give me some harsh language regarding this subject. 00:36:39.0300000 Give me some nice language regarding this subject. It's great for that kind of stuff. It's also great if you have a writer's block and say write me a thank you note to send to my grandmother, didn't tell them. What about just a general thank you note, Then I take it and edit it. It's great for that stuff. 00:36:53.9100000 It's bad for judgment, compliance, verification. And it is ugly too if you don't verify. A lot of times when lawyers have gotten in trouble or other people in other professions have gotten in trouble, it's because they've taken the generative AI output at face value, not double checked it, not had that human oversight, then flipped it around and sent it out. That's a problem. So we always need to think, you know, we need to verify everything. 00:37:22.6500000 We need to make sure we doing all those things because that is going to be very important in it. And again too, especially if you're a lawyer, you going in front of a disciplinary commission or court, you know, when opposing priorities requesting sanctions because you mis cited something and you're saying, well it was Jenny, I did it. It's a software usage thing, not me. I can't get in trouble. I'm telling you now, that's not a defense. 00:37:48.8500000 There's over 600 cases where judges have said that's not a defense for lawyers and it's probably not going to be a defense for other professionals in business either because why? Rule 1.1 Competency. We have to be competent as lawyers. Other professionals have to be competent in their profession to the standards of their profession. Competency. 00:38:11.9900000 Competency includes being competent in relevant technology that affects our practice and affects our clients. We have to be competent. So if you think you can just go blame AI, like maybe you blame a law clerk or an assistant for something wrong in a brief, it's not going to fly with a court. There are again over 600 cases that tell you it's not going to fly with a court. These are real time sanctions of attorneys coming down in real life cases. 00:38:38.7200000 And there is actually a website, I don't know if I have it in the PowerPoints here today, a website that tracks when lawyers are getting in trouble across the country, usually for giving false information to the court, that truthfulness because they relied on the outputs of generative AI and didn't double check it. So we always need to double check it now in order to effectively Use generative AI, we need to think about how we talk to it. Now that may sound weird because again, it's a pre trained transformer. But talking to a generative AI model is kind of like learning a new language. It's also kind of like talking or learning to communicate with a human being. 00:39:20.9700000 We all know sometimes talking to a spouse, significant other or family member, the ways we communicate with them may be different than the ways we communicate with a colleague. We also know that some people may need a little more in depth communication. Like step by step details of how to do something. Where some people may need just hey, I need this done. And they say okay, I got it. 00:39:41.5500000 And you don't have to give them step by step details. Gen AI is no different if you're talking to a generative AI large language model, we have to learn how to communicate with it so that we get the results we want from it. It's not going to adapt to us, we have to adapt to it. Because it is a pre trained transformer, we're not going to change its programming. Only the programmer can change that. 00:40:07.6900000 We have to change how we communicate with it to get the results we want. And then it can be even more effective in those areas where lawyers or real estate lawyers can use it, particularly if you have those large enterprise versions. So for my lawyers, you'll understand this reference. When you are talking to a large language model, think of it like a law clerk or a first year associate. Meaning you're not just going to say I need a motion for summary judgment on this issue. 00:40:35.2900000 You're going to give them more details. You're also going to go back and clarify, double check it, send it back, double check it. You're not going to take it at face value. So think of that generative AI like a first year law associate or law clerk. None of us would turn in a brief straight from a law clerk without double checking it, reading it and going through and editing it. 00:40:58.0400000 We should think of generative AI the same way. We need to be conversational with it. Not barking orders and commands, but training it and teaching it what we want and how we want it. We need to be detailed in our request, not vague. We can't just say draft me a motion, eliminate on what topic? 00:41:15.9400000 How many pages do you want? Do you want certain venue? Like we have to be detailed, you have to provide context, clarity, expectations, review the drafts, refine your prompts, get a new draft, refine your prompts, get a new draft. It's a back and forth with the generative AI. It's not just a hit it and quit it or enter it and be done. 00:41:34.9800000 And it's a back and forth conversation with the generative AI large language model. So we have to remember that. And it takes work, it takes work to understand how to communicate with the generative AI to get the results you want. So we can't just set it and forget it, meaning enter it and say up, we're done. This is the product it gave me. 00:41:55.1300000 No. If it's not what we want or we need something else, we tell it to do something else, we get another one, we review it, we tell it to do something else, we, we get another one, we review it until we get to that final product. Just like a first year associate, just like a law clerk on how we would interact with them. So we need to engage, refine, repeat. Keep that in your head every time you talk to a generative AI. 00:42:17.5700000 Engage, refine, repeat. Detailed prompts, review the output, follow up, make changes, continue this process, reiterating it till you get to that tone, format and context you want from the generative AI. Now obviously when I'm talking about this, how we talk to generative AI, remember we still have to keep in mind those ethical considerations and security. But just imagine you're in the enterprise version where you can do whatever you want, enter the client information, it's all going to be confidential. All those things are met. 00:42:49.2200000 This is how we talk to it. Detailed prompts, refine, repeat. We, we need to think about that way. You can do this process and the one that doesn't keep confidence is the public facing one. But it has to be more in generic format. 00:43:01.6300000 But if you want the more detailed one, this is the process. Engage, refine, repeat. You're not accepting the first draft usually ever. You're collaborating with a pre trained transformer just like you collaborate with an associate, a law clerk or a junior attorney to get to an end product. The only difference is, and you may find this funny, you may find this not For a generative AI model, it's available 24 7. 00:43:26.5100000 It doesn't talk back to you when you send back drafts in red line or tell it to do it differently. And it doesn't get annoyed when you ask it to do it differently or scrap it and start all over. A first year associate, a law clerk, may roll their eyes at you, may cuss at you under their breath. Generative AI is not going to do that. So just think of it as that back and forth with a pre trained transfer your computer, that's 247 with no back talk. 00:43:52.3700000 Utopia. Right. So again, but we have to think about it in those ways. Now. There are five different or there's multiple different prompting techniques or acronyms you can use to understand how to prompt a generative AI model efficiently to get better outputs each time. 00:44:10.2900000 Some of the ones I like to use, these are the acronyms on the screen in front of you. There's one I like to call Clear Legal, Prompt and Safety. Each one of these, the letters in each one stands for something different. They all end up the same place. It's just whichever word you like and you can make up your own too. 00:44:26.6300000 These are just ones that I'm giving as a demonstration. They're not the end all, be all. It's not the only ones out there. And you'll see that when we go through the details. For instance, the clear framework, you need to tell the generative I its concise role. 00:44:40.7900000 Who is it? What is it doing? Is it mom yelling at kid? Or is it a lawyer responding to a judge? Is it a lawyer responding or writing an email to colleagues? 00:44:51.4000000 The concise role of the generative AI is going to be imperative. Then you need to give it any legal or contextual background about what you're wanting. So the details that go into it, but again, make sure you're in that enterprise version. But the details, if there's any law you want it to consider, if there's any contextual background, you need to give it those details. You also need to tell it your expectations, boundaries and limitations. 00:45:17.7200000 Expectations, you want a 10 page brief. Boundaries, you don't want it to do XYZ limitations. Maybe it can only be 5 paragraphs or 3 sentences, or you know, not cite to New York law, or you know, not include xyz. If there's parameters and boundaries and expectations, you have to give it to it. One of the biggest things I found is if you don't give it these expectations and boundaries, it's going to do what it wants. 00:45:44.7300000 I've prompted it many different ways to see outputs. I've said create me a PowerPoint on this topic. It gives me six slides. I say create me a PowerPoint that's going to last an hour on this topic. It gives me 10 slides. 00:45:56.8900000 I finally say give me a PowerPoint with 30 PowerPoint slides, each one talking about a different topic that would last an hour in talking, presentation with talk, and guess what? It gives me those more details. The more detailed you are with it, the better results you get. And so you've got to Think in these, the A stands for actionable tasks instructions. This is saying, create me an email, Create me a memorandum. 00:46:20.4300000 Create me a PowerPoint. Tell it what you want. Don't just say, here's the information, what would you do? Tell it the context you want it in. A lot of times it gives me bullet points and I say, no, I want this in paragraph format. 00:46:33.8300000 So we'll change the bullet points into paragraph format. You, you have to tell it how you want those results, and that goes to that R. Result definitions, how you want those results. So you're going through concise rule, role, background, expectations, boundaries, limitations, actionable task, what you wanted to do, and then how you want your results to come out clear. See, it stands for that. When we go into those other ones too, like legal and prompt, and the other ones on the screen, you're going to see the same concepts just with different words. 00:47:04.0300000 So whichever one sticks best in your mind, that's the one. I go with lawn or legal. Launch your role. What role are they? Kind of like that last one, concise role you're going to see, I guess, similarities here. 00:47:15.7600000 Explain the background, guide them in the task you want, add in your constraints, you know your limitations, and then list the deliverable, or list out how you want it delivered. A little more generic on the terms, but just to get you a way to think of a framework, we need our role, background, task, container, constraints, how we want it provided every single time. When we prompt a generative AI prompt, similar, just with a little more oomph or a little more detail in it, you have the purpose, the role, the orientation, the method, the parameters and the template. Templates, how you want it, parameters, those limitations, the method. If you have a method of how you want to get there, like show your work, give me paragraphs, orientation, kind of be that background, the role they are, is it the attorney, is it the realtor, is it the client? 00:48:01.8400000 Who are they? And then the purpose. Like we're writing this to explain XYZ so they all start to go together. It's just again, whatever you want to use. And you don't have to use any of these to think of the detail you need. 00:48:15.4400000 And prompting a generative AI safety is another one. Scope the who, what, when, where, why, accuracy, like what, limit, like what, how you want it done, or do you want them to use law? Do you want them to use more generic things? Filters, limitations, exclusions, the ethics that come into play, tone, target, audience, yield, deliverable. This one's a little more nuanced in the areas but again, all of these clear safety, prompt, legal or anyone you can think up up on your own gets you to better results. 00:48:46.2600000 In a generative AI, we just have to be detailed. That's the point I'm trying to get across. You have to talk to them like that first year associate or, or new realtor or whoever it is, you need the back and forth with them. So drafting prompts is going to be an area where you're going to practice and fail and practice and fail. So here's just some prompts that I've kind of drafted in different ways just to kind of give you an example of these acronyms coming into play. 00:49:14.5700000 This one for my lawyers out there. Deposition summary. Bad prompt, summarizes deposition. It doesn't go through the scope, the accuracy, the filters limitations, the audience, the output detail I want or the method or anything like that. I just said summarize. 00:49:29.8600000 So I'm going to get that. Just a blah summary that's not going to have any parameters, any limitations. I haven't told it to cite to certain pages. I haven't told it to, you know, do certain things. So what I'll get is a generic chronology summary, missing key admissions, missing page number citations, failing to identify contradictions and essentially useless in my trial prep. 00:49:51.9200000 If I prompt the AI and upload the deposition, just say summarize this, it's not going to be useful. I'm going to have to go back and talk to it again. When I talk to it again, I'm going to be more in tuned with the acronyms. So I'm going to use safety here just because it's a little more nuanced in areas you can kind of see how it plays out. I'm going to tell it the scope. 00:50:12.5600000 You're an experienced in house litigation attorney working on a commercial dispute involving a breach of fiduciary duty. I, I give it different things. I also say you have a 274 page transcript obtained on here. Create a summary organized by role knowledge, personal financial interest, communications, document awareness. So I'm even telling it how I want the summary, you know, categorized or what categories I want in the summary. 00:50:36.9100000 I'm also telling it for accuracy to include those page and line limitations, direct quotes to key admissions, flag testimony, ensure summary. Then if I want ethics to come into play or anything, I tell them to cite or discuss any rules of law that are relevant. I tell the tone. The target audience is going to, they want them to write like a professional. I want it litigation ready, meaning for a summary for My team of the key testimony and then I tell it the yield I want or the output is a structured evidence based summary that supports the theory using headings, subheadings within the summary. 00:51:11.8300000 You see this compared to summarize this deposition. This is going to get me that better result. This is going to get what I want. Or more closely, I've engaged it. I'm going to get an output. 00:51:24.3100000 Then I'm going to look at that output and say okay, well I need to refine it a little more. Maybe I don't want those paragraphs I said maybe I want more detail here and not as much here. So I'm going to repeat the process. I'm going to start having that back and forth conversation. Maybe now I want a timeline creation. 00:51:41.1800000 Maybe I want a comparison analysis. It is a conversation back and forth. So we need to think about that. We need to. This helps. 00:51:50.6600000 And not just deposition summaries, but think of this in summary of anything large. Maybe you have a large contract, maybe you have a large something you want to summarize down. It can summarize and make hundreds of pages searchable with strategic summaries throughout. You just have to prompt it, right? It can help process things in minutes. 00:52:06.8900000 It can clarify and go through. Now you obviously need to verify accuracy before you rely on it 100%. But one way it can summarize volumes of information. Legal research too, or just research in general if you're more in that real estate professional category, not just an attorney. What is the law and the non competes in Ohio. 00:52:28.0500000 That's very generic prompt. I'm going to get a surface level general overview, possibly outdated information. Because I didn't say if I wanted current information, old information or both. I didn't ask for any case citations or anything like that. So it's just going to give me a blob paragraph. 00:52:44.9200000 I didn't have any facts in there and I didn't ask it for any specific output. Do I want it in a memo? Do I want an email? What do I want it? So I'm going to get a bad result which I'm going to have to go back to with my legal framework or my clear, my legal, my prompt, my safety or whatever acronym you're using with more details. 00:53:04.7300000 I'm going to tell it, you know, you're a senior employment attorney advising on these. Here's what we want. These five key legal areas, I wanted them to include links to sources. I want citations and I want links to them. I'm telling it, don't produce me False results or anything without a citation. 00:53:20.5700000 I want the citations in blue back, blue book format. I want them to cite to ethical rules as well as cases. I want this in a memo blue book format, provide comprehensive, authoritative legal analysis. So I'm giving it more detail, then I'm going to get a better result. So then they're going to give me something. 00:53:39.2300000 I can refine it. I can, you know, ask for more site verification. It doesn't get it. I can double check the blue book sites. I can ask for now adverse authority or motion strategy, or I can take what is given me and prompt off of it or refine it. 00:53:53.7300000 I mean gauging again, gauge, refine, repeat every single time to get to an end product or multiple end products that could possibly help. This process can be redone over time in multiple different areas. With legal research though too. Remember I talked about using Westlaw LexisNexis more than others because there's over hundreds of falsely cited cases. This right here is actually a hyperlink to the website. 00:54:21.1700000 The website's on the next screen. At the time I pulled this down, which was probably about six months ago now, there was only 340 reported cases of hallucinations where attorneys have been sanctioned. Now that number's over 600. If you go to this website, that website is www.Damien D A M I E N C H A R A R L O-T-I N.com so damien charlton.com and he tracks in real time as cases are reported and coming down, all the attorneys who've been sanctioned for hallucinations and false reporting and lying to the court and then also their penalties, jurisdictions and what they falsely cited. So great way to do it. 00:55:05.6500000 Like I said at the time I pulled this down, there was 340, but that was about six months ago or so. There's over 600 now when you go look at them. But that again, is that website there to get that information? Some of the most notable ones are on your screen. We're not going to go over them in detail because we're running short on time. 00:55:22.8300000 But just make sure you check your legal citations because they can be wrong in the name, they can be wrong in the citation itself. You can double check them through Westlaw or LexisNexis. But like here, it's the same citation but attached to two different names. And sometimes it does that. So it contains a fake case and name. 00:55:43.0700000 So Finders versus Keepers was a fake case name, but that's a real citation. It's a real citation. But to the case of Johnson v. Caldwell. So it can mix up stuff. It's not always right if it gives you a citation. 00:55:55.2500000 So always double check that. But it can do research, it can do those. But it does have a higher instance of hallucinations when it does legal research. Another area we talked about is contract review. Obviously. 00:56:08.3700000 Is there anything risky about this contract? Bad prompt, it's generic surface level. We're going to miss key things, not have what we want, not tailored to a certain client. Good prompt, go through your things, go through you know what you want, who you are, your page line, citations. You want a redline version, you want a new draft, what do you want? 00:56:28.3800000 Go through them to find those different areas. Then again, engage, refine, repeat. Maybe now wanted to do a negotiation strategy, a president search client memo. Use it for those different things. Review contracts for the risk, compare them and there is contract specific generative AI software. 00:56:47.9100000 So also look for that when you're choosing your program. Presentation drafting. If you're like me and do a lot of presentations, it can help you with those. I don't ever use it as an ending point, but sometimes a starting point just to get a brief outline and then I add in my details. So here, make a presentation about a hospital client and the DOJ investigation that's very unclear. 00:57:08.2300000 We're missing critical nuances, we don't have slide specificity. It's just going to be very generic. It's not going to be great. So then we have to go through the details. So even on things like this, the generic things we, or you know, more presentation, we have to still do the good prompting. 00:57:25.1100000 We have to still use our acronyms, engage it the way we want to and then we can engage, refine, repeat again, develop the content more, include case studies, give me visual suggestions to make the slides more interesting. Give me an executive summary, give me bullet points, give me a summary that can turn in for CLE credit. Whatever it is, we just need to use it in that smart way. So again, presentation drafting and one of the last examples here before our time is up that matter transponding, analyze real estate cases data and tell me what you find. It's going to be very generic, no specific jurisdiction, no settlement patterns, no nothing specified. 00:58:00.8200000 So we need to have that good prompting again when it comes to that, we can ask it then to go deep dive comparative focus. This is great too for strategic applications like risk assessment, resource optimization, budget forecasting, business development and all the above. So when using generative AI, just be specific, treat it like that first year associate or someone you're training and you should be good to go with that. We are at the conclusion, so I want to make sure you get that code. It's E as an elephant, the number five, E is an elephant again, the number nine and the letter K as in kite. 00:58:37.8600000 E is an elephant, the number 5e is an elephant again, the number 9 and k as in kite. It's retyped in the chat box for you. Thank you for attending today and have great prompting out there. Use it wisely and don't be my next test case on a seminar. Have a wonderful afternoon. 00:58:57.3900000 It.