E78: AI in Investing: Risks, Opportunities & Investor Judgment

AIME Episode 78 Transcript
Sarah Florer (00:02.939)
Hello everyone, welcome to Alt Investing Made Easy. Today, Roland and I are going to present to you our discussion about AI and how it's being used in investments and alternative assets. Roland?

Roland Wiederaenders (00:17.154)
Yeah, sir. This has just come up so frequently in our practice. you know, you and I, just to remind the audience, we're attorneys and we're coming at this whole topic of alternative assets from the standpoint of being attorneys. these investment opportunities that we talk about here all start with the creation of documents, whether it's a disclosure document or a pitch deck, marketing materials. These are written words, written information.

If one of our clients ever got sued for securities fraud, they would be judged based on the written information that they put in front of investors. Of course, whatever they said orally, not in writing, but for sure the written materials. And I think it's important for us to think about this and talk with our audience about it because increasingly these words that create investments and that we use to evaluate investments are

being written more and more by people other than human beings.

Sarah Florer (01:23.874)
Yeah, and I think it's been really interesting, isn't it? Because there are tools now that you can use to help you create a pitch deck. There are tools that we've seen clients use to help them digest and understand their PPMs. There are tools that can be used to verify or look at different kinds of issues. I'm sure there tools being used in evaluating the pipeline of investment opportunities for people that are working in different kinds of...

whatever it is, whether it's commercial real estate, whether it's a hedge fund. guess in a sense, hedge funds for a long time have used not AI specifically, but programming to determine algorithms, to determine trading strategies. So there it would be expected that AI would participate in probably algorithm development and everything else. it's just like with everything.

AI, the uses that we're all finding in our personal lives and our work lives for these tools is pretty cool in a lot of ways and useful. But then always there's this cautionary tale and something that we would always want to talk about, which is the documentation that people rely on for making the offerings and their alternative investments. These are legal documents and AI cannot create those for you on your own.

You want to show we talk about that a little more rolling. We were just talking about it before.

Roland Wiederaenders (02:49.934)
Well, yeah, and you know, it's always been an issue for us when clients show up with drafting their own documents and issue only in the sense that typically, you know, we say it'll be easiest if we draft the documents, but I've had clients before, you know, really believe that they may, hold on.

Roland Wiederaenders (03:13.954)
that they may be able to gain some efficiencies or really make sure that they're getting their point across by drafting their own documents. So that's always been a thing. And I think now it's just with AI, the documents have become more voluminous and it's a new tool for people to draft even more words and undertake analysis using the AI.

We're always presented with our clients coming to us with documents and just in our role as attorneys, we're responsible for 100 % of what's in that document. so that's just the thing. Client drafted documents have always been a thing, but now they're doing it with AI.

Sarah Florer (04:02.731)
Exactly. And the same cautionary tale is there that we have with AI in general, including in your personal life, right? Because I think everybody is using AI for recipes and for meal planning and for gem planning and for... Somebody I spoke to today uses Gemini to summarize her emails, her personal emails every day. all these excellent... You can draft lessons, you can learn a language. There's all these things that you can do, which is great. However...

when it comes to these legal documents, there are hallucinations. And in those personal contexts, a hallucination isn't such a high-stakes thing, is it? Because if you have a recipe and it makes a mistake and you cook something for eight minutes instead of seven minutes, it's probably not going to be overall detrimental to your life. But for lawyers, we have to spot hallucinations, which can be difficult. And it adds time to matters sometimes.

And also on top of that, there's some interesting trends that we're seeing with some of these AI tools where they get into a loop of what everybody's asking them to draft something, this, that, or some kind of agreement. And then there's this format that's coming out that actually is not necessarily 100 % aligned with what would be good standard established practice from a legal point of view, what that particular document should be called or what it should look like or.

how things should be ordered in it or how they should be described. the interesting thing is that I think AI is a great base resource for lots of reasons. And you can get really far with it and things that don't require a lot of expertise, but the hallucinations become a bigger and bigger issue. The more precise the expertise involved needs to be. And when it comes to legal practice, you know, we...

we would consider expertise being very important and that we are held accountable, like you said, Roland, for everything that the document says, not just the parts that we read or review.

Roland Wiederaenders (06:01.646)
There's been a recent example of people getting bad medical advice through AI. I've got cancer and AI is telling me to stick garlic in my ears or some crazy thing like that. And people do it and wind up destroying their health. So just as much as we need to be circumspect in using AI about things like that, for sure the important thing, something like your personal health. What we're talking about are

Sarah Florer (06:07.845)
Hmm.

Sarah Florer (06:14.309)
Mm-hmm.

Roland Wiederaenders (06:30.19)
creating or using AI to create documents that you could be held legally liable for or creating something that you're trying to make a sale off of. You're trying to convince somebody to invest in your deal. you got to be really careful, I think is the main point.

Sarah Florer (06:40.613)
Mm-hmm.

Sarah Florer (06:49.691)
Well, in another observation, Roland, think people at the same time we're all using these tools, we're all getting better at recognizing when others are using these tools. And you have to, think when it comes to a pitch deck or a sales document that you're presenting to people that you want to trust you as a human so that they'll invest with you and you can have access to capital raising that way. To me, it strikes me as a judgment call and that...

showing that you're willing to invest the time and money in making something from scratch as opposed to what we all know Claude can do in 10 minutes with the right prompting, then when it comes to a pitch deck, say, I think that can be a differentiator for people. more and more, we're going to become better at spotting what is the common AI response on any topic versus what is actually a human putting together something.

I can't reiterate. Personally, I would invest in something that feels like there's a human behind it. We've talked about it at length on this channel about how in the end, in an alternative investment, it's the manager, it's the individual who's promoting that deal that investors are trusting. So the deal mechanics and the specifics of the location or the project or anything are

also important, but if that person is not striking confidence in those investors, they'll never get their money over. I think the efficiencies, perceived efficiencies, I would say from AI, because I don't think they're always actually resulting in real efficiencies, has to be balanced against what you're trying to convey to your audience.

Roland Wiederaenders (08:29.73)
Yeah, you know, and we've said it so frequently on here before that a great deal can be made horrible by a bad manager and a mediocre deal can be made great by a good manager. And, you know, I think defining what a good manager is, good deal sponsor, using all the tools at their disposal, but in the end, you know, what you're saying, we're really relying on this human element. I want a human being to be making these investment decisions.

Sarah Florer (08:39.781)
Mm-hmm.

Roland Wiederaenders (08:59.374)
For sure, let's use the computer aided tools that we have. man, I think the bottom line on all of this is what we've observed, and we deal with highly intelligent people. I've just observed a great overuse of AI generally.

Sarah Florer (09:19.877)
Yeah, I would agree. you know, it's interesting, shall we talk a little bit about the tools that we use for the audience? For lawyers, we've always had detailed databases of information that we use to rely on to understand the law, specific cases, different nuances in the law, history of different issues as they've been developed, standard documentation, established documentation, things that have changed through court challenge.

there is a huge, vast amount of information that's available in a legal database, the typical tools, the two companies being Westlaw and LexisNexis. And this all used to be in books that you would find in law libraries. now, then it went online and that was amazing. And a lot of people have never had the experience of doing legal research in a book, but only do it online. And now, of course, we have...

agentic AI tools that are built into our already pre-existing databases. And so for lawyers, think not much has changed actually. It's just that rather than Boolean search terms, we use more narrative search terms. But the same issues remain. And that is that the information you get from a vast database of data is only as good as the search commands you put into it to generate output. And then

know, our AI tools can help us get down into granular detail very quickly, but everything still has to be reviewed because of hallucinations, even with inside these legal tools, but also because, you know, template documentation and things that provide standard outlines for what might be common in certain agreement or whatnot. It's not necessarily exactly what you want. It's not exactly what lines up with your experience or the specific deal.

I would say that the AI tools are helping in terms of making search commands easier for this vast array of data that we lawyers always have had to use and be comfortable with, but it's not making it any less amount of work. I don't know, what do you think, Roland?

Roland Wiederaenders (11:29.826)
Yeah, yeah. I just, you brought up the term hallucinations. I think that this is the biggest risk. you know, also the distinguishing the Boolean search versus the natural language, you know, and I can ask a question. The other day, this came up and a client said, I need a deed of trust for a condominium in Texas. And that's real estate. And they said, but

we don't need all these land provisions in the deed of trust because it's a condominium. And I said, well, no, it's still real estate. But I wanted to make the client meet their question. It's a legitimate question. I'm not an expert as much with real estate law as I am in securities law. I know a lot about real estate law for sure, so I'm not going to sell myself short. But I do use these research tools.

I went to Protege, which is Lexus's AI tool, and I asked it in a natural language query, is a deed of trust in Texas different for a condominium versus a, I decided to contrast that with a single family home, single family residence. And it came back and said, yes, there is a different form for that. It didn't provide me with a form, but.

In the end, I just had to use my judgment because it seemed like, you know, just the answer is how to perfect a security interest in real property is through a deed of trust, period. That's the answer in Texas. But yet it was giving me this sort of, I don't know, prevaricating answer. And honestly, you know, I think it tells us what, you know, we're speaking about it as a...

has a mind of its own and it really does to some degree, but I think it gives us the answer that it thinks we want to hear. And so to close the point or to finish the story, I went back to the client and I said, listen, I've got given you a document that I know will perfect the security interests in the condo that you're trying to achieve. And the AI tool told me that there may be a form for condominiums, but I don't really trust it.

Roland Wiederaenders (13:48.964)
I don't trust its answer. If you want a more detailed answer, I'll call up our buddies up in Plano, the real estate guys, and ask them about it. But that's going to take more time. it's going to, you know, more time. What does that mean? More money, right? Make this deal more expensive. And so I told that to the client. And once I said that I've given them something that achieves the objectives without spending any more time, you know, they go off and...

Sarah Florer (14:05.105)
Yeah.

Roland Wiederaenders (14:17.453)
Maybe one of our listeners will correct me and please do. We always invite and want audience participation. So if anybody knows about this and has a different form, deed of trust for condominiums versus ordinary, regular, non-condo real property, please correct me. But that was just a really good example. And I think the clients come up with these questions and maybe using AI on their own, or they come up with these questions that...

You know, we're sort of new and I don't know, just they have a different quality than questions that clients have come up with in the past. I don't know if that makes sense.

Sarah Florer (14:56.849)
I think it makes a lot of sense because I've had everything from points and questions about documentation that are drafted to look like their natural language from the client but include some technical queries that also are not relevant but make the client then believe they're relevant, right? I think ultimately some of these AI tools, they don't have the ability to really weigh what's important.

in a practical sense. And so I got one document back from a client that had all these red flags about these big issues. And when I went through the entire document to try to understand what was going on, which of course is increasing the time, there was one typo that had changed something that on a plain reading of the language would not have changed the meaning, but because it's an AI tool and it's picking up, you know,

patterns that are inconsistent, it made this seem like there's a huge change in the meaning of a clause, which there wasn't. There was enough other evidence in that clause to demonstrate exactly what the intention was behind the clause. There was one small typo. And I'm not advocating that typos are okay, but I've never read a legal document prepared by another lawyer myself that didn't have one typo in it at least. That's the nature of anybody's work actually.

And typos are different from substantive errors for that reason. But it was just interesting how this AI tool had created a paragraph of information that we had to process with the client to come to the conclusion that there was no issue. And that's the other side of efficiency that one thinks one's getting is that also if it has to be dealt with, then it's going to raise the legal fees. I'm just thinking of this. I remember a long time ago.

talking to a doctor and I guess it was a time when technology for scanning a full body was becoming available. And I asked my doctor, should I do this? Should I do a full body scan to see every little thing that might be there? And she was like, no, you will drive yourself mad because there'll be things, there'll be little anomalies in your bone structure versus a typical bone structure. There'll be little benign tumors or little things, little funny things that are not important, but once you know about them, you're going to freak out.

Sarah Florer (17:20.439)
So don't do it. And sometimes I think these AI tools do that. Like they raise, and yes, what you will end up with once you work through all of that is a document that can pass through an AI tool, which I guess we're somehow agreeing is a good thing. But at the same time, I think what you really have is a lack of that practical judgment. And this is something that we attorneys offer you. We can spend a lot of time.

making documents extremely perfect. when you go to law firms that are much bigger hourly rates than ours and they have teams of associates and senior associates and a junior partner and a partner on every deal, that's exactly what happens. It gets filtered, filtered, filtered. So by the time your product is done, there's been so many eyes on it that everything is caught. Now, you pay for that quite a lot. And so that is something that...

there's a practical approach, which is you don't necessarily always need that for every single deal. And this is a judgment call and a fine-tuning that a human can make. It's not necessarily something that a machine can make. Now, I'm sure a machine will start to be able to attempt to make these and become quite good at them, but they still have to be verified and checked. And then there's the other question of everybody knowing that this is all machine-generated and all of us probably wondering underneath it, like, is anything real anymore?

Roland Wiederaenders (18:23.502)
Yeah, that's

Sarah Florer (18:43.931)
Like where is the human contact in any of this? So I think it's a question. I think AI is very useful. I'm glad clients can get good guidance on kinds of questions to ask if they feel like they don't need, they don't understand something. But I also think that you can get really caught up in details that are too, that are quite frankly wrong or come from completely different industries. That's another thing I've seen coming.

with work in the alternative investments is that some of the typical nomenclature and structures and matters that get raised in big deals led by investment banks are coming up in these small deals. Well, you're not going to an investment bank because you can't afford it probably if you're on a smaller deal. And you don't need to...

necessarily get into all of the vast number of things that get attended to when you're paying a top-shelf investment banker and his team of many, many people to raise your private offering for $10 billion of whatever it is, private company that's being sold, or whatever private commercial real estate fund is being marketed. So this is the scale that I think we're not yet to in a reliable way.

Roland Wiederaenders (20:00.715)
I really like your analogy about the full body scan and the notion of how we exercise judgment as attorneys. I hate to say it, but we did go to law school and there was a reason why we went there and they teach you some really specific things. Part of it is using good judgment and that's one of the things that

Sarah Florer (20:03.876)
Hmm.

Roland Wiederaenders (20:26.817)
that really I've observed has grown in me, my nearly 30 years of practice in your 20 years. That's just something that all the experience we have, we've seen these things and we're really good at pattern recognition as well. And we're maybe even better at assessing out what truly is the practicality of any particular issue that we're dealing with.

Sarah Florer (20:34.555)
Hmm.

Sarah Florer (20:52.635)
Mm-hmm.

Roland Wiederaenders (20:55.481)
You know, guess we'll probably flesh this out as much as we could today, but I will tell clients that I'm sort of glad whenever I see that they give me this AI analysis or all these questions, because honestly, it takes me more time. And what does more time translate to but more billable hours? So if we're going to take the full body scan and look at the

problem with the toe to fix your earache, then so be it. We have to bill for it, right? I hate to be so cavalier, but if we get this analysis, if we get anything from a client, we're responsible for it. We're responsible to read, digest, and respond to them and say, no, this is relevant. This is not relevant.

Sarah Florer (21:28.119)
Yeah, exactly.

Sarah Florer (21:38.523)
Yeah.

Sarah Florer (21:46.775)
Exactly. And let's just carry that through to we have clear guidance from courts, not just in the United States or in specific courts of Texas, from really many courts have now grappled with this issue around the world. And that is attorneys are still responsible for what they submit in a legal context. It doesn't matter what AI told you. It's become a problem with pro se defendants using AI. And there might be a tiny bit of grace, but otherwise you're held in contempt.

produced documentation that's objectively wrong, that could have been easily checked or easily verified to be incorrect. And the doctrines behind all of that are, you know, for another day, but it makes sense, it? That if, you know, a court is trying to establish fact or the legal process, sometimes it is a court, sometimes it or the judge, and other times it's the parties and the jury's the final decider of fact.

But in the end, the process is about establishing fact, making a determination, and having a consequence that's relevant. this is something that has to be done involving human judgment in the end. And this is why we attorneys can't say, your AI document looks OK, without actually understanding what it says.

Roland Wiederaenders (23:12.111)
And this is the one thing that encourages me, Sarah, when I think about AI, or am I going to be replaced by a robot? I went to court unusually last Monday with a probate matter, but there was a human judge sitting behind that bench, and there were human beings that show up in that courtroom. There were human attorneys that had to stand up in front of the judge and make arguments and ask the judge to sign orders and do things. And as long as you have that,

As long as you have human judges, they're going to require human beings to stand up before them. I suspect, I don't know, I guess this could change in the future. It'd be interesting just as a thought experiment. How could it be different from how it is now? But we do have these human judges and if they... They're real sensitive to, like you said, the things that AI might just make up to... You submit...

some made-up case to the judge. I mean, could be disbarred as an attorney for doing that. anyway, I think that there's always going to be a case for human attorneys as long as we have human judges.

Sarah Florer (24:12.027)
Mmm.

Sarah Florer (24:21.391)
Yeah, well, and even then, I don't know if it's a world we want to live in where we think things can be automated to the point of having non-human judgment involved. mean, but that's exactly what a judge is,

Roland Wiederaenders (24:30.736)
We wind up with Skynet, right? I mean, that's what Terminator was all about. So we're fighting the battle every day against AI. I'm just joking. No, we use AI, but we use it very responsibly and very circumspectly and carefully. And we would counsel all our clients to do the same thing.

Sarah Florer (24:37.041)
Exactly.

Yeah.

Yeah.

Sarah Florer (24:48.933)
Mm-hmm.

Sarah Florer (24:55.057)
to do the same. Absolutely. Yeah. I think that wraps it up, huh? Yeah, great. Probably.

Roland Wiederaenders (24:58.736)
Well, what do you think? Did we miss out on anything? Oh, cross-reference checks. I was going to mention that specifically. That is a really good use of AI. You know, we have these 50-page documents that say, accept as provided in section 5.1c, you know, blah, blah, blah, whatever. And is there really a 5.1c in the document? And then when you look back at it, does it actually say the thing that it's supposed to?

Sarah Florer (25:15.727)
Yeah.

Roland Wiederaenders (25:24.954)
And I've had clients give me a really good cross-reference check that I know that they did with AI. I think they told me that they did it with AI and it was great. It was good for that purpose. So there's some really good applications for AI.

Sarah Florer (25:33.841)
Yeah. No, you know, following that, I discovered a use recently which was setting out your survival clause. Your survival clause at the end of a contract is where you gather the certain clauses that you want to survive the termination of the agreement. And it's always changing the way they're numbered or identified. And if you can draft it at the beginning, it's going to change. And so then that's quite a useful tool. It's kind of like a little summary of certain things at the end.

Roland Wiederaenders (25:47.367)
there you go.

Sarah Florer (25:59.825)
just one place where you put all of that. So I would say that's a pretty useful.

Roland Wiederaenders (26:07.332)
Well, we want our clients to know that we're trying to be as efficient as possible for them and also using all the tools at our disposal. If we insisted on handwriting agreements or typing them on a typewriter, nobody would hire us. But it's the same thing with this. So the intelligent use of this and let's hear maybe we don't get too many comments, but let's hear from our audience members how they use.

Sarah Florer (26:24.016)
Yeah.

Roland Wiederaenders (26:35.554)
AI and let us know in the comments.

Sarah Florer (26:36.497)
That would be great. That would be great. All right, everyone. Thanks for joining us today. We hope you enjoyed this episode. Please like and subscribe to our channel, and we'll see you next time on Alt Investing Made Easy.

Roland Wiederaenders (26:48.868)
And remember, take aim with your alternative investing strategies.

Sarah Florer (26:53.211)
See you next time.