0:00 Who are you? And where are you going? What do you want? Hello and welcome to Trust and Estates Weekly. I'm your host, Kevin Bemel, and I'm here every Monday at 10:00 a.m. Pacific time to talk about uh estate planning and trust and probate administration. Uh primarily from the standpoint of uh professionals like
0:35 myself um who do this for a living. I guess that's the definition of a professional, right? but also for those of you who uh don't do this for a living, but perhaps you found yourself um appointed executive of your parents' uh estate or or you're a trustee and you're not really sure what to do. So hopefully this will give you information that will make your job uh in that regard easier and uh help you uh speak to your representatives uh more cogently. So today I'm I'm going to be
1:13 talking about a topic that seems to be addressed on pretty much every business podcast to which I listen and that is the topic of artificial intelligence or AIS. Um, for the for the purpose of of this show, when I use the term AI, I'm usually I I'm I'm almost always going to mean uh what are known as large language models. These are things like chat GPT and Perplexity and Claude um Gemini um
1:50 and and just you know so you know uh what what my you know sort of background in this is I've been using chat GPT as an intricral part of my business for um over a year now probably a good year and a half and I I also use uh perplexity quite a bit. Um although that's I'm typically using that for non business uh um issues. Um primarily because it's computer desktop app does not scroll
2:26 properly basically and so it's too frustrating to use it on anything other than my phone. I have occasionally used Claude and although it it supposedly is the best at least according to a friend of mine who is still in the United States Navy uh I I don't I don't particularly like Claude. Um I do use it for a couple of business tasks mostly because I tried it out using those tasks and and continue to do so because it has the memory of the previous times there. I haven't used Gemini. I I hear it's
3:03 great. I haven't used Grock. I hear it's also great, but haven't used them myself. So, um that but the the issues that we're going to be that I'm going to be dealing with today are not so much uh any they're not model specific. um they are the issues in which we for which we need to be concerned when we're engaged in the administration of a a trust or a probate. And so that that's that's going to be the topic. I I
3:39 have uh I believe six thoughts. Yes, six thoughts here in my notes uh and and I'm going to go through them um and and give you give hopefully give you some some food for thought. So, uh, let's start out with, um, understanding that whether you administrate professionally the way I do or whether you're doing this as maybe just a a one-off situation for your family, you still act as a fiduciary.
4:16 That is the central character of the role that we have. And I've talked about this many many times on on this show. So the essence of being a fiduciary is one has to set one's self-interest completely aside and whatever decisions actions are taken must be for the best interest of the trust or the probated estate. Okay. So, and and those things can be in conflict,
4:51 especially if you're a beneficiary as well as a trustee or or as well as an a personal representative. And so, that's where we really need to be careful because we need to be prudent in exercising that fiduciary duty. So I think the key thing to keep in mind is AI is a tool but it cannot substitute for making decisions on our part and for getting proper guidance from qualified
5:29 professionals. Maybe a maybe a lawyer, maybe more than one lawyer, maybe an accountant, real estate professional. Okay. It cannot be it cannot be used as a substitute for for proper input. Okay. I I shudder to think what would happen if I went into court to defend one of my actions and I said to the judge or my lawyer said to the judge on my behalf, "Well, this is what you know, Chat GPT said to do." I I'm pretty sure the judge would look at me and say, "Uh, you know, I don't have the authority to revoke
6:04 your license, Mr. be but uh I'm going to refer this to the bureau for consideration you know uh because that is not at all a proper way for me to uh handle the fulfillment of my fiduciary obligation. So the duty uh to take personal care is nondellegatable meaning I cannot transfer it. If you're if you're acting as a fiduciary you cannot transfer it to someone else.
6:40 I can't even say well this is what my lawyer recommended for me to do. Okay that may be true. Okay. But the decision still rests with me. If I got advice from my lawyer and that that advice turned out to be wrong, okay, that doesn't excuse the fact that I made a bad decision. Okay? Or I may have breached my fiduciary duty. Okay? If I got advice, bad advice from my lawyer, that may give me an action
7:15 against my lawyer for malpractice or something like that, but it doesn't it doesn't uh excuse me from breaching my fiduciary duty. Okay? And and and and it won't you either. Okay? So, uh, let's say you're dealing with, uh, some sort of tax issue and you go to chat GPT and you find out that, uh, um, okay, let's say ChatGpt tells you that, uh, um, you know, you don't have to file
7:52 an estate tax return, okay? Because let's say the gross value of the estate is $5 million and the current exemption is uh from estate taxes is $15 million. Okay. And so therefore since the estate won't owe any estate taxes, it doesn't need to file um a 706 an estate tax return. Sounds perfectly logical, doesn't it? Okay. However, there could be some weird aspect of that
8:27 of that estate that requires the filing of a return and because chat GPT didn't know about that weird issue, it didn't give you proper advice. Okay. So, uh, when when it's a situation like that, the proper thing to do would be to reach out to a CPA. And by the way, a CPA who is knowledgeable about estate and and uh trust uh rules of accounting,
9:02 okay? Especially tax law. And check with them. Let them ask you a few questions, okay? Most accountants, especially the the good ones, they're not going to charge you for for, you know, a 15 or 20 minute conversation where you say to them, "Hey, here's what's going on. Do you think I need to file um do you think a 706 needs to be filed on this estate or do you think anformational return needs to be filed for this trust?" Okay. Now recognize the caveat. It's not just
9:40 any accountant, but an accountant who's familiar with these issues. That's who you need to. But you can't rely on an AI to answer that question. Okay? uh and and and if if you do initially ask the question um to to a to an AI, I tend to I tend to um say chat GPT because that's the one I use most often, but but if you if you ask that question of an AI, okay, you you definitely need to get some kind of independent verification unless you feel that you're
10:18 experienced enough on your phone to to be confident that the the information you got from chat GPT is the correct answer. Now, notice there's a a a distinction there. It's not chat GPT that's making the decision. You are based on the information that you got back from chat GPT. Okay? Now, one should keep in mind that you know they we talk about AI hallucinations and such those are real
10:55 and depending on how you ask your question. Okay, chat the some of these AI models will seek to give you the answer it thinks you want as opposed to the the so-called correct answer. So we need to be careful when we put our queries in to be uh neutral in terms of leading the witness so to speak. Okay. All right. Where so so if if AI can't if AIs can't
11:30 make decisions for us where can they be used? Well, so most of of where I use AIS is in kind of administrative in not in well that's a bad word to use if we're talking about administrating the estate in sort of like clerical duties if you will. Okay. So if I need to um uh review a lot of information, okay, and I and I want a summary of it, I may use an AI to do that. So let's say
12:06 I've got a very long uh you know trust document, okay, I can upload that trust document and ask the AI to give me a summary of the key points. Or let's say that I want to know uh what does the trust document have to say about compensation to the trustee? Well, I can upload that document. Maybe it's a 30 40page, you know, trust document and the AI can go through it for me and it can find where in the document it discusses uh trustee
12:43 compensation and summarize that. Okay. Now again we need to be careful. We can't assume just because it gives us a summary that that that summary is comprehensive. So we need to ask it from where did you draw from where in the trust document did you draw the information for this summary right and then just doublech checkck it make sure it didn't miss anywhere or mis misinterpret uh what those sections say. Okay. So, uh, those those are, you know,
13:18 that's sort of a, if you will, almost a clerical chore, okay? But but an AI can be much more efficient at it. Okay? It can also be used for assistance in drafting a document, but it should be subject to review by the the proper advisor. So there may be you may need to make routine notices to beneficiaries. You may have you may be writing status updates or or um things like that. So um
13:54 you know if it's within your purview let's say it's an a status update. So I know for my for in my practice I do a status update uh at least monthly in some cases weekly if the matter is uh complex andor there's there's a lot of activity um over short periods of time. So I don't do this but in theory I could I could use an AI to help me draft those those reports. Okay. And if since they're they're an a status update on the work I'm doing. So
14:30 then if I review them and I think they reflect properly the status of the matter, I'm good. But if there's some kind of legal issue or sometime some type of accounting issue on which I'm updating the status if an AI writes it certainly minimally I have to review it and I may need to have it reviewed by my lawyer or my accountant as well or the accountant for the estate as well. Okay. So nothing should come out of an AI and be used without it being reviewed minimally
15:06 by you as the personal representative and potentially one of your advisors. Okay. Another uh a proper use of AI um would be in analyzing data. Okay. So, um, let's say that you've got a property, there's an incomeroucing property as an asset to the estate and you you get, you know, monthly reports from a property management company about them. So, you might upload those reports, maybe you've got, you know, six
15:41 months of reports. You upload them and you ask the AI, uh, identify trends or areas of concern. um reviewing these reports, okay? And it can help you maybe it'll identify uh some trends that you wouldn't notice just by looking at the individual reports. Or you might say to it, put together a monthby-month spreadsheet of the data so I can compare it on a month-by-month basis. Okay? Typically, a management company isn't going to do that. A property management
16:17 company is going to report on a monthly basis and they're going to give you the monthly results and the year-to-date results, but but they're not going to break it out, you know, each month for a six-month period. So, if you want to see what's happening in terms of expense trends or or income trends for that matter, an AI would be would be very efficient in compiling that data. And then, you know, depending on which model you're using, it will give you a link to potentially a um, you know, an Excel
16:52 spreadsheet or or some sort of PDF document or something like that. Okay. And finally, an AI may be useful in terms of task management, workflow, giving you workflow efficiencies. Now, I use Apple reminders. for me. I haven't seen where an AI would be useful for that. But um um you know perhaps um I I I do use it uh when I'm not clear um you know what a timeline may be. Like for example, uh earlier this morning
17:29 I had it in the back of my head that um one of the estates I administrate has been sued and and Discovery was served and and there's a certain deadline uh to respond to discovery. So uh you know I I did it quickly in my head because the the deadline is is pretty short. But if if it had been a longer deadline, I would have I could have asked the AI, okay, um I have, you know, 90 days to respond to, you know, this particular request for discovery that was served on, you know, March 9,
18:05 2026. What is my deadline for responding? Okay. Um uh I I can tell you for a fact that that AI is very good at calculating those dates um because I I I do that periodically for a different project on which I'm working. Okay, in this case I use perplexity to do that. So for those kind of workflow efficiencies and making sure you're staying on top of deadlines, uh definitely AI can be very good. Okay. Uh let's see next um where let's let's take a look at
18:40 where AI could potentially create legal risk. Well, as I mentioned a little while ago, um an AI does not substitute it cannot substitute for proper legal counsel. Okay, which is you know an attorney admitted to the bar who is experienced in the area of trust and estate law. Okay. Uh uh you know an AI may give you you know some ideas uh to uh you know to discuss with your lawyer but it cannot uh
19:15 it cannot substitute for that lawyer. Okay. Um, I don't see a day in the near future when, you know, I'm gonna show up with, you know, uh, your honor, I'm being represented by, I don't know, chat GPT, right? A little chat GPT robot, you know, approaches the judge or is there, you know, in Microsoft Teams, which is the uh, virtual platform that LA Superior Court uses. I I, you know, not not going to happen uh, here in the near future. It's going to be a live person, right? Um, and and so,
19:52 uh, you know, an AI can be used, again, as I just mentioned a minute ago, to, uh, help you come up with questions to ask your lawyer, but it's not going to give you the advice that your lawyer would give. That that's number one. Um, it can also be problematic if you're using it to interpret documents. Okay. While it can help you pull out pieces of the documents, find them more quickly in terms of interpreting, for example, how um how a a court might rule
20:31 on a document or how a particular uh um uh paragraph in a document should be um should be applied in the actual administration of the estate. eight that that could be problematic. Okay. Now, let's let's say um you know you you have a a clause um in a trust that requires you to report to the beneficiaries. Okay. And but it doesn't specifies any
21:07 periodicity. Right? So you don't know every week, every month, every six months, every quarter. What do I do? Right? So one of the things you might do is you might query uh chat GPT or an AI and say typically how often do trustees under the particular type of trust that that you you know for which you're the trustee how often do they report and what do they report and and and ask it that sort of marketbased question right so that you
21:43 have an idea of what maybe the standard is and then you can you can doublech checkck that standard. So maybe hopefully if you're a professional then presumably you do u but if you're not a professional maybe you know contact a professional and say hey uh you know here's the situation what do you think about you know it appears that uh trustees are basically in this situation reporting every quarter or every or every you know year or whatever say what do you think about that and and get some input from some
22:20 advisor ers, ask your lawyer, you know, I'm I'm I'm trying to figure out what is my reporting periodicity obligation. Okay? And and your lawyer should be able to help you with that. Okay? But at least um but getting getting some picture of that from an AI would be fine. um using the AI conclusively to interpret that clause would potentially create a le a legal risk. Okay. Uh another issue so often times a a a trust
22:57 will carry a provision that says um distributions need to be made for the health, education, maintenance and support of the beneficiary. Right? And so while you may query the AI and say typically what kind of expenses fall under health, education, maintenance and support. Okay. And that they you know that again that would be sort of you know going out to the market and and getting a bunch of opinions about that. Um that might be fine. uh you could not
23:36 rely on the what the AI tells you in making your own decisions. Those decisions have to be yours based on the specific facts of the particular situation. So, uh, let's say for example that, um, you know, a trip to Europe, okay, um, maybe the AI tells you that a trip to Europe, generally speaking, is not included in health, education, maintenance, and support. Okay? And the beneficiary has
24:12 asked for a distribution to take a trip to Europe. And you and you think, well, the AI says no, so my answer is no. That's going to create a legal risk for you. Okay. One of the parts that's missing in that is a ask them why do you want to go to Europe? Oh, well, you know, my my child has this uh god forbid, you know, rare disease and I want to take my child there uh to meet with a specialist. Oh, okay. So, so you know that travel
24:47 has a has potentially you know potentially a nexus to health maintenance and support. Okay. Or even education, right? So, or maybe maybe it's something not quite so cataclysmic. Maybe um the person is a chef and they want to go to Italy to study uh under you, you know, to study Italian cooking. Well, that that could very well fall under the category of education, right? And so, we need to make our own broader investigation,
25:22 okay? And make sure that we're exercising our human judgment, not defaulting uh to uh to what the the what the AI says. Okay. And then finally, I spoke about taxes. So, um, estate and income taxes and income tax returns, they need to be minimally reviewed, I would argue, prepared by a qualified accountant. Okay? So, you know, using an AI to make uh tax determinations would create big-time legal risk. Okay? Let's take a quick
25:58 break. When life brings unexpected changes, managing inherited property shouldn't add to your stress. Magnum Properties specializes in probate and trust real estate management. We handle the complexities while you focus on what matters most. From property maintenance to tenant relations, from legal compliance to maximizing value, we're your trusted partner through every step. Magnum Properties, Inc. turning property challenges into peace of mind. Call us
26:35 today at 213-264-4396 or visit magnumprops.com. Okay. Next, let's talk about confidentiality, privacy and data security concerns. So uh one of the things about which you need to be careful is if you are uploading documents into the AI system, what does that mean? Okay. Now, we uh you know, if you've been following AI at
27:10 all, so you know that there's lots of discussion about AI's, you know, taking at least potentially copyrighted materials in order to train their LLMs and and and there's been quite a bit of, you know, pulling of hair and nashing of teeth over that. Um those are complex legal issues. I'm not here to talk about those. uh but if we upload documents into an AI, what is their uh what is their policy on document retention and
27:46 the use of uploaded documents to train models? Okay. Um and and to the extent that they are retaining them, what does that mean? So that is an area where uh we need to be careful especially if you're using the free versions. Okay. So I right now I do use just the free versions. I don't I don't upload any proprietary documents like you know a trust agreement or something like that um because I don't know enough about how the various models
28:21 handle those issues and I don't want to create a a a confidentiality or a breach of privacy um issue by having those you know those documents um by ha by having uploaded those documents. So another issue could be um if the document is uploaded have I waved some kind of privilege right one of the benefits of uh someone having a trust is it's not public uh information okay so when an estate is administrated via a trust
28:58 there is no court record because there's no probate filed and so the administration proceedings ings stay essentially confidential as opposed to a probate which is a court proceeding and for which a lot not necessarily all but a lot of the information may be open to the public. Okay. And so by uploading a trust document is confidentiality essentially being waved or at least breached uh such that uh people who
29:35 ordinarily would not have access to that information now might have access to it. Now, especially in a case that has some level of um uh publicity, if you will, maybe it's a celebrity u uh you know, the administration of a celebrities's estate, right? And and people tend to be curious about this stuff, okay? uh for for you know for good or for bad. I mean uh you know there there's a lot of interest in things that uh you know I I would argue people have no business
30:12 knowing about. Okay. Um, but you know, by putting them by uploading them into the AI, have you now created a path whereby I don't know a a reporter or somebody else would have access to it that they otherwise would not have? And and what kind of liability um would that mean for you? Okay. So before you would uh you know use an AI um so let me before I get to that one um another issue is um issues like like
30:48 uh healthc care documents. So now you've got HIPPA the I forget what HIPPA stands for. It's basically a confidentiality law with regard to information about someone's health. So it it it narrowly constrains when that information uh can be revealed. Okay. And uh the fact of the matter is uh uploading health data into an AI could potentially mean that there's been a HIPPA violation. Okay. Um
31:23 the same may be true of financial documents. There's no HIPPA for financial documents per se, but there are sensitive financial records. Certainly things that have personally identifiable information, so-called PII, things like social security numbers, bank accounts. Okay. Potentially very problematic to upload uh uh documents that that have that information in them or to input it directly into an AI. Okay? And so before
31:59 taking any of those steps and by the way lawyers do in in some cases uh upload that kind of information. This is why lawyers have special uh additions, if you will, of AI models that have the kind of excuse me, confidentiality, privileged uh access uh um protocols that the publicly available models do not.
32:38 Okay. So before uploading any of that type of information uh you need to do some due diligence on the vendor whichever model it is that you're using. So what are the vendor's data retention policies? Is the data encrypted? Is it anatomized or you know does it essentially remain in the system complete so that someone who works for the company could potentially have uh somebody who works for that vendor might have access to that data.
33:14 Okay. In general, it would be, I think, grossly imprudent to rely on the, you know, the free models, the publicly available models from a confidentiality and privacy and data security perspective. Okay. Uh, you're probably going to need at least some kind of enterprise level AI um that has contractual protections, encryption, etc. if you want to use it uh if you want to use it with the kind
33:49 of confidential information or privileged information that oftent times is in the possession of someone who is a fiduciary. Okay. So, we've talked about where um legal risks could be could be incurred. We've tal I've talked about uh privacy and confidentiality issues and and the you know the risks um if if uh they were potentially breached. So um let's talk now about the litigation
34:25 implications. Okay. So first of all is the and and these are speculative questions. So to be honest with you, I don't I don't know um the answer to all of these questions. And the first one is a is a big one. So what is this discoverability of of responses and outputs that one gets from an AI? Okay. Is that something that if you're required to respond to discovery needs to be included in the
35:01 discovery? Okay. So, could usage logs be evidence? Um um and and could responses that you've gotten from AI um be evidence. Okay. And this is why it's so important that if you are going to use an AI in your fiduciary activities that you be very careful about how you use it and make sure you haven't substituted it for proper advice for making for using your own judgment in making decisions that you haven't in essence
35:36 essentially let it supplant you as the fiduciary. Okay. And if the prompts coming out of an AI are discoverable and and let's say someone suing you sees there's a pattern of this is what the AI says this is what you did right so you you know and and they and they can get a hold of those prompts so now you may be in you know you may have caused yourself a big problem uh because you know number
36:11 one uh you didn't you didn't determine in advance whether the uh responses you got from the AI were discoverable number one and number two you didn't uh perhaps you didn't uh you know do your your due diligence beyond the AI okay I'm not I'm not saying that you didn't just it certainly it might appear so and so then you're going to have to be able to show okay well yes this is the response I got from the AI and then after that here's what I did uh to to investigate further before I made my decision and in the end
36:48 it is not coincidence per se that what the AI recommended and what I decided are the same thing but you know I I I think I queried the AI pretty well and then when I checked with my advisors um they you know they advised me in such a way that it became apparent that this indeed was uh the best decision and so that's why I made that decision but the AI did not make the decision. Okay. Um I mentioned this at the beginning and that is I relied on an AI is not a
37:24 defense. Okay. I I just I cannot imagine any judge no matter how competent or not thereof right accepting that as any justification for anything that that you have done. Okay. I wouldn't even uh I wouldn't expect a let's say that um you you you need to sell a property. And so you go to an AI and you say, you know, what is the house at at 123 North Main Street worth? And it spits out the number. Okay? And you decide, okay, I've
38:01 got someone who wants to buy the house for that number. I'm going to go ahead and sell it for that number. and then it turns out that she sold the house for too low a price. I cannot imagine the judge saying, "Uh, yeah, that was fine." You know, um I I think the judge is far more likely to say, "Are are are you a qualified real estate professional?" No. Oh, well then why didn't you get one involved in this sale in the sale of this property? at least getting uh you know getting a proper valuation of the
38:37 property etc etc. Okay. So it it you know um and and by the way AI's potentially could be very good at valuing properties and in fact automated valuation models have been around now for gosh at least two maybe three decades or more. When I say around, I mean in, you know, in in general use for at least two, maybe three decades or more. Okay? But just because that is the case doesn't mean that you can
39:12 substitute what the AI tells you for your own judgment. Okay. Another problem with with AIS that I've mentioned a little bit is uh hallucinations um that that to which AIs are sometimes subject. And so if you've let let's say you're you're asking it a legal question, okay, and it and it gives you the the the answer, okay? Uh and and the answer looks like the one you want, okay? you have to be careful because some of the AI models
39:49 they're they're developed sort of to please us um which can get in the way of their object objectivity to the extent that they even have objectivity. Okay. So, uh, you know, they might tell you, oh, you know, don't worry about, you know, this issue, uh, you know, and and and may have picked up on it that you are really worried about it and don't want to be and so it's telling you, oh, don't worry about it and the next thing you know, it's actually very important and and now you're in trouble. Okay. Um
40:25 so all of these um all of these mishaps they do have implications for litigation either by making opening you up to potential litigation or by uh putting you in a position where if you're in litigation uh your position is not as strong as it might have been had you used AIS more prudently. Okay. Um so you know where where you might use it um uh well is so
41:05 I've got an one of in one of the estates that I'm administrating. I have a problem with a lender who has delayed months and months and months in recognizing my authority. So, I may have to go to court to get an order forcing that lender to do certain things or to refrain from doing certain things. And I am I'm I've been thinking about how I can use an AI uh to uh uh be more efficient in putting together the documentation that I may need uh if I
41:40 have to pursue litigation. And one of the things that I'm thinking of doing is using it to go through my personal cell phone records to find the exact dates that I called the lender to, you know, their customer service to, you know, deal with to deal with the matter at hand. Right? So instead of having to go number by number through three or four months or actually more like five months this has been going on five months worth of phone records you know dozen well not dozens of calls
42:15 a day but you know many many calls a day uh and potentially missing the number and AI could be very efficient in going through and finding th the numbers the the dates that that you know the particular number for the this particular l uh this particular financial institution was contacted. Okay. So something like that could be very good. Now I would supplement it from my own recollection of a pro how many times I spoke with the lender to make sure that the AI identified all of
42:52 the instances. Uh but in the end the AI is probably going to be more thorough in identifying the number of times the the the times that I called I would be going through the the phone records manually. Okay. So in terms of litigation support that would be an example of where I think it would be very effective. Okay. Again potentially more effective than than than a human being um without creating any kind of liability. So, I would be very comfortable saying to to it, you know, if I were
43:28 challenged on that, well, how did you know you called on all these dates, I would say to the judge, you know, your honor, I uploaded my phone records into chat GPT and chat GPT identified every time there was a call to or from the lender's number. and and your honor, uh, you know, artificial intelligence, it is going to be much more thorough and accurate for a a task like that than I would be myself going through them manually. And so that's how I did it. I think that argument would be compelling
44:03 for the judge. Okay, let's take another quick break. Probate and Trust TV, your curated source for the most useful and insightful content on estate planning and trust and probate administration, featuring standout professionals like Bill Gross, Paul Bralet, James Cunningham, and yours truly. Whether you're a lawyer, accountant, real estate professional, or working on your own succession plan, this channel helps you quickly find the clearest, most
44:38 practical content YouTube has to offer. Subscribe now to Probate and Trust TV. Spend more time learning, less time searching. Okay. So, I've gone through uh um some of the concerns a bit about AI, how it can be used properly in administration, where there's legal and confidentiality and security risks, what its implications are for for litigation. So, I want to wrap things up on this topic by talking about uh how if if you're not
45:15 already using AI or maybe even if you are uh how can you do so prudently or more prudently and that really comes down to having a policy for yourself for your firm if if you have more than one person on how AI should be used. Okay? And especially if you have a multi-person firm, uh always best to have policies in writing. So, uh if you're so so think about, you
45:50 know, what is it? What are you or your people permitted to you for for what are your people and you permitted to use AI? Okay. And and that would get into the issues of privacy, data security, etc. Okay. Uh what supervisory approval may be necessary? So for example, let's say you've got a you know a lot of people in your firm. You have several layers. Okay. And by the way, maybe you're not a professional fiduciary, but uh you know, you have a company um that and and you
46:27 have other people working there and they're helping you with the administration of your parents' estate. There's nothing wrong with that per se if if you're using the staff of your business to help you, okay? But especially if they're not familiar with the issues of estate administration. So then they may just go ahead and use an AI to help them with some kind of task and then you don't realize until it's too late that there's been some sort of confidentiality breach. Okay? And so if
47:03 you are going to use it and it doesn't have to be anything, you know, long and drawn out and particularly uh, you know, particularly fancy, but just put put down some some basic rules. So, you know, if any documents are going to be uploaded to an AI, you know, it needs my approval before you do that so that you as the person knowledgeable having watched this show, right? uh you can decide whether this is something that it's okay to upload um due to privacy concerns or data retention or whatever.
47:38 Okay? Uh or maybe you have an enterprise level AI system at your company already. Okay? And so in that case it it's not going to matter. Or maybe you can maybe it has the ability to sort of segregate out what you're doing. And so if anybody at your firm is going to be helping you with the administration, they have to know to use that segregated portion of the AI or however that may work. Okay? So set the basic use policy and then make sure that you you train your staff.
48:13 Uh again, I I don't think this really differs whether you're a professional or whether you're doing this one off. If you're going to have people helping you, they need to know what are the ground rules for using AI. And again, it doesn't have to be a, you know, dayong seminar. It could be just, you know, you could take just get together for 30 minutes and say, okay, look, you know, I need because I'm I'm administering this estate, I need to be concerned about, you know, issues A, B, C, and D. uh you know this is these are the uh guidelines for how you can use AI as
48:51 you're helping me uh with my administration duties. Okay. And and make sure they understand the limitations, the risks. Okay. So I I think it's it's silly to set constraints without without explaining to people why those constraints exist. Okay? because oftentimes constraints seem just they're just there to be constraining when in fact there are good reasons for them being there and and people need to know that. Okay. Finally, uh what disclosure
49:27 needs to be made? So, for example, do the beneficiaries need to be informed um that you're using AI to help you with the administration? Okay. Uh and and does the let's say it's a trust that that's an easy one because this trusts often uh often um u put down in writing what the the various uh obligations and duties and and authorities of a trustee are. So uh
50:02 does the trust restrict uh the trustes ability to delegate? Okay. And if so, does using an AI constitute a delegation of of duty or a delegation of authority? Okay. So, uh that's something also to consider uh and and put into your your policy. Okay. Maybe you need to talk to your lawyer and say, you know, within the verbiage of this trust agreement, uh, you know, how would it how would it view my use of AI? Would that be an a
50:38 permitted delegation of authority or would it not? Okay. And then finally, uh, keep in mind that that an AI is a tool. Okay? And this is we're coming back to where I started out. Fundamentally, a AI is a right now at least is a productivity enhancing tool. So whatever policy you're going to establish, it should be most likely within that keeping AI within that realm of enhancing productivity and it should
51:16 most likely restrict the use of AI uh and and so that it's not used to make what otherwise should be human decisions, human judgment calls. Okay. So in the policy it's as important to say for what issues it may be used as it is to say for those for which it may not be used. Say so be clear in both of those. Okay. And that way the people who are working on it they can you know they they can they can do so in the best way.
51:53 Okay. So so those are my thoughts on using AI for in a fiduciary capacity. I I I'm sure that over time my ideas about this will evolve and I I I think you know at some point I'll do another show on this topic as I integrate AI into my practice more. I'll I'll I'll share that with you. Um you know including any any pitfalls I find along the way. But I do encourage you to think about it and experiment with it. AIs are an
52:28 enormously powerful tool and you you've if you listen to discussions about AI um you hear all kinds of things they're going to wipe out all jobs or this or that. I I I I don't I don't buy they're going to wipe out all jobs. What I think they're going to do is they're going to bifurcate. So those of us who are who figure out how to use AIS responsibly, productively, okay, they're going to make us more valuable and those people
53:04 who shun them are more like are going to be more likely to get passed by. Okay? So just like we had to learn how to use the internet, we need to learn how to use AIS responsibly and productively. And so I encourage you if you're not already experimenting with it with respect to your professional life to do so. Um I I did I'll I'll refer this wrap up with this and this may be useful. So on one of my other shows I think it was business lessons from the military. I
53:40 had a a a viewer who asked me a question about uh getting he he was he said he was in blue collar work. He didn't specify what and he said, "What was my advice for getting into white collar work?" Okay, maybe it was on people making money. But in any event, he asked me that question. And my response to him was, I'm not sure that getting out of blue collar work today is a good decision. It is the case that white collar work is more vulnerable to AI. Okay? especially the anything of a
54:16 clerical nature or an administrative nature in the broader work world not in the trust and estate world uh is more vulnerable to uh to AI. What I recommended to him was he figure out how can AI be used to make his blue collar work more productive. Okay. And I don't know, I don't know if he was a tradesman, electrician, plumber, HVIC guy, or if he was um in uh
54:51 you know, some kind of manufacturing maybe. But somewhere in I think virtually every kind of bluecollar work out there, there is a way that AI can be a can be a productivity enhancer and that but the people who are going to know that are going to be going to be the people who are already doing the work and see oh how can I use this tool to make myself more efficient? How can I how can I work more smartly using AI? All right. And
55:29 and there's two opportunities there. One is to be someone who is more more efficient, more productive, and that's always an enhancement. But two, it gives you the opportunity to train others on how to be uh more efficient, more productive. Okay? So you've got your opportunity in your primary work, but you also have an opportunity doing some consulting to other people maybe that who who are in your industry. So that's my approach to AI. It's one of the reasons why I did this show and
56:05 hopefully you found this not just interesting, but you found it you you find it useful uh if you're involved in trust and probate administration and in any other way that that you know in any other areas of your professional lives. So, thank you very much for joining me. I appreciate it. Uh tomorrow at 10 a.m. Pacific time is my show, Business Lessons from the Military. On Wednesday at 10:00 a.m. Pacific time is my show People Making Money. And then every other Thursday at 5:00 PM Pacific time is my show Jiggers's Place where we have
56:40 a cocktail in a schmoo and we talk about non-finances related topics. So I hope you'll join me again later in the week. And that just leaves me to thank you once again. Oh, by the way, if you have questions, comments, criticisms, uh links will be in the show notes um to social media as well as my email address and and that's on all the the platforms uh um YouTube, Spotify, Apple Podcast. So, feel free to reach out to me uh if if you have if you would like to. And
57:17 got to wrap things up now. And so, thank you very much again for watching and signing off as I always do. Marie darling, you are my bell. Courage at all times, my friends.