Can AI Replace Tax Lawyers? (Part I)

In episode 33 of How Tax Works, Matt Foreman evaluates human and AI answers to a straightforward tax question, seeking to (somewhat) answer the question of whether AI can replace or augment tax lawyers.

How Tax Works, hosted by Falcon Rappaport & Berkman LLP Partner Matthew E. Foreman, Esq., LL.M., delves into the intricacies of taxation, breaking down complex concepts for a clearer understanding of how tax laws impact your financial decisions.

Follow us on Bluesky:
@howtaxworks.bsky.social

This podcast may be considered attorney advertising. This podcast is not presented for purposes of legal advice or for providing a legal opinion. Before any of the presenting attorneys can provide legal advice to any person or entity, and before an attorney-client relationship is formed, that attorney must have a signed fee agreement with a client setting forth the firm’s scope of representation and the fees that will be charged. 

This Podcast is Hosted by:
Falcon Rappaport & Berkman LLP
1185 Avenue of the Americas, Suite 1415
New York, NY 10036
(212) 203 -3255
info@frblaw.com

Transcript

Matthew Foreman [00:00:00]:
Welcome to the 33rd episode of How Tax Works. I'm Matt Foreman. In this episode, I will evaluate, or at least discuss, how close we are to having tax lawyers replaced, or at least significantly augmented, by artificial intelligence. How Tax Works is meant for informational and entertainment purposes only. This may be attorney advertising, and it is absolutely not legal advice.

Matthew Foreman [00:01:02]:
Please hire your own attorney and no, artificial intelligence is not a substitute for your own attorney. How Tax Works is intended to help listeners navigate the intricacies and complexities of tax law, regulations, case law, and guidance to demystify how taxes shape the financial and business decisions that we all make. Before we get started, a few administrative things. New episodes every 2 weeks. The next episode is going to continue my evaluation and discussion of how soon AI will allow me or force me, I guess, to retire. If any questions, comments, constructive criticism, you can email me at my FRB email address, which you can find via your favorite search engine. Upcoming webinars, again, I have some. They're going to be in November, December.

Matthew Foreman [00:01:42]:
Got 4 really good ones. If you haven't been to one before, send me an email. You can get continuing ed credit for lawyers, accountants, lawyers, CPAs, EAs, and CFPs. So, so excited about that. Otherwise, you can I'm going to post about it on LinkedIn or I'll email people who've been to one before. And I think that's a good one. All right. Let's get this party started.

Matthew Foreman [00:02:05]:
So I have wrestled with the idea of doing this episode for a while. It's actually one that was suggested really early on by one of my partners. And It's interesting, right? It's an interesting idea. And people ask me at parties, you know, they're like, you know, is AI taking over your job? Is it changing how you practice? And AI is certainly transformative, even in law. But it's been much less successful in replacing deep technical expertise, right? Analyzing a contract for a specific thing, maybe. Determining the correct entity type to hold a business or an investment, whether a trust is grantor or non-grantor, if I could talk, the effective tax rate for the sale of assets from a multi-member partnership that does business in 6 states. I'm just going to say maybe here too. Maybe is a strong, strong word, maybe a weak maybe, but you know, it's an idea, right? Will it do it? And I think that's where the question sort of comes in is how good it is, because even if it's fairly decent, which in some situations it is, I still have to check it.

Matthew Foreman [00:03:17]:
I still have to make sure because, and this is, you know, I see, I see this so often and you see the news stories, but like people are really submitting briefs and documents to courts that just have made up citations that are clearly written by AI. And it's just like, what are you doing? Like, what are you, what are you thinking? You know, and it kind of surprises me. And attorneys do this and accountants do this. It's like, come on. You know, the germination of this episode, really the very first one, was a question someone asked me at a wedding really about 2 years ago, in addition to a comment made by my partner. And they asked me if I thought I'd have a job in 5 years because AI's technical expertise was expanding so rapidly. And I kind of chuckled and I said, I figure I have 20, 25 20, 30 years. By then I'll probably want to retire anyway, so I'm good.

Matthew Foreman [00:04:09]:
And they seemed astonished. They were like, no, no way, there's no way you have that long, not possible, shocked, gobsmacked, whatever, whatever words you want to use. And, and I kind of pointed out, and I point this out a fair amount, is that tax law is code. It's written by humans and is not properly drafted because it's designed to do certain things and specifically prevent certain things. Also, you know, and then this is the example I always use, is Tax law uses and and or as synonyms and antonyms, and it isn't always clear which is one without reading the context and understanding larger things. So, you know, I just, I don't see it. You know, tax law is incredibly complex. Every day I struggle to answer what are really straightforward questions because the answers are not straightforward.

Matthew Foreman [00:04:55]:
And the balancing test is what's really hard. You know, I talk about it all the time. And I say this to clients, it's a business decision. There's no right answer here. And that's where it's a struggle. And then, you know, I get responses from clients that are clearly AI. Sometimes they're correct, but even if they're correct, the analysis can be unclear, incomplete, you know, and I have to gently but directly, as I do, respond to them and say, I don't know about that. And sometimes they're incorrect and I can't figure out why they went wrong, you know.

Matthew Foreman [00:05:25]:
It's fine, it's going along, it makes sense. And then all of a sudden, you know, it's like, hey, you know, you should form a Maltese trust. And it's just irrelevant and confusing if you're opening a taco restaurant in the Lower East Side, right? On the Lower East Side. I think it's in the Lower East Side. Anyway, so you get that problem, right? And sometimes I get what I call as a pejorative the Holy Trinity of AI answers, right? Which is a correct citation with a conclusion that is the opposite of the correct conclusion. And it's often of sub-regulatory guidance. So revenue ruling, revenue procedure, chief counsel advisory, stuff like that. That is, they're generally the easiest to understand and the most, you know, you can disagree with them because they're sub-regulatory guidance.

Matthew Foreman [00:06:14]:
So if you don't agree with them, you can take the opposite position. But they read it and it comes out entirely wrong. And that just, that fascinates me. And so, so you have the correct citation with the incorrect conclusion. You have a made-up citation that happened. I mean, I got one 2 weeks ago that was, that was like this. And the third is a citation to the code or regs, right? Which are harder to read but dispositive. And it's, you know, code regs don't really say it, but they basically rely on their incorrect conclusion to show how the code or the regs say, oh no, no, this is the answer.

Matthew Foreman [00:06:52]:
And it's just, it's frustrating, you know, is the right word. And I think that there's a reliance on AI that's just totally unnecessary and unhelpful that they say, oh, here's what I got, you know, from AI. And I'm not naming names, but here we are, right? And I know how simple queries, you know, turn out, right? Because I get that from clients. This is what I asked. This is the answer I got. 99% of the time that's true. I would say 50% of the time they tell me it's AI and 50% of the time I can figure it out. AI has a way of writing that humans don't write.

Matthew Foreman [00:07:26]:
If they do, it's because they desperately love Charles Dickens and Mark Twain. And their first language, you know, that they learned is Latin. And so they just have a really flourishy, fascinating, weird grammatical way of speaking. So I can tell, right? So I know how simple queries do it do because clients send me those. And out of curiosity, I've looked into it before, right? But what I wanted to know, and this is what I did in this episode, is what happens when I give a question that a tax professional has to sit there and balance? And answer the questions and give a technical answer. That's what I was trying to understand, right? Can AI synthesize a complex fact pattern? Can AI discern between the competing law and cases based on relevant factors? This is— that's what I wanted to know. That's what I wanted to understand. I didn't go maximum complexity.

Matthew Foreman [00:08:20]:
I joked with a client. They didn't find this funny, but I joked with a client. Maybe it was probably February or March. So, so, you know, 4 or 5 months ago. And I said, you know, your question is one that would be a fact pattern for a class, like a private equity tax class that had partnership and corporate and cross-border and multi-state, a lot of multi-state, a lot of state and local stuff. And they were not thrilled to hear that because they want the, this is a simple fact pattern. No, it's not. And I think that's what I wanted to know.

Matthew Foreman [00:08:55]:
I didn't want to go like total max complexity, but I did fact patterns of the nature of questions, the types of questions I get a fair amount, and I suspect many of my listeners do too. And, you know, they're the ones I can answer, provide guidance without sitting down for 4 hours and researching. Right. So before I get into that, let's take a quick break and I'll be back. We'll talk about the first one. Hope you— hope you're enjoying the temporary new music. Okay, so Before I get into the fact patterns, which I'm going to, I'm going to read, they're also available on, on the FRB website that has the episodes for How Tax Works. The fact pattern, I just, there's going to be like a PDF or something where you can just read them yourself.

Matthew Foreman [00:10:08]:
But before I get into them, I tried to evaluate them somewhat objectively, the answers, but also I didn't want to just pair AI against me because that's, that's not fair to AI for two reasons. One, I hope, gosh, I really hope that I'm better than AI. I freaking should be, candidly. I charge enough an hour and I've been doing this for not a trivial period of time, so I should be better than AI. So I also got some of our summer associates. I asked who would like to do it and two volunteered very nicely. I asked the associates, I voluntold who I wanted to do it because I decided that's who did it. And then I fed it into AI.

Matthew Foreman [00:10:57]:
The summer associates were allowed to work on each fact pattern for 1 hour and then they spent an hour together working, you know, answering it. So total, so they spent a total of 3, well, 3 hours, right? They had the ability to research through using our research tools, which did not include AI. In fact, they were forbidden from using AI. But as anyone will tell you, you're using AI even when you don't want to. I told them they can use Google, but not like Gemini or any of the search engines, AIs and stuff. AI tools. AI is a plural. Weird.

Matthew Foreman [00:11:33]:
They verb it. I don't know. But, you know, they could Google. and then they could use our, our research that we spend thousands and thousands and thousands of dollars a year on. Um, so, you know, BNA, Tax Notes, Westlaw, Lexis, stuff like that. So they used real ones. Um, the associates had 1 hour per fact pattern, so 2 hours total. Um, and they did not discuss it together.

Matthew Foreman [00:11:55]:
Um, 'cause I wanted to see what they had separately. Um, and then AI just kind of fed it in. I'm going to go through some stuff where I was curious about their answers. And so I went out and I, I sort of augmented it with additional research and additional questions. But I, you know, that's it. And these fact patterns, oh boy. Oh, and there was no clarification whatsoever. You'll notice in one of them, I think the second one, there's actually a fact that's missing.

Matthew Foreman [00:12:24]:
I thought I included it from a fair reading of it. I definitely did not. So I thought that was interesting. And, you know, I went into it. So the fact patterns I designed, it's a combination of what you would get in like a law school exam and how clients write emails, right? So there's irrelevant facts, it's narrative-based, you know, they talk about things. It's not totally in logical order, but how people think, you know, so it's in generally logical order, but not completely. And again, fact patterns are online, so I'm going to read them. Because I think they're important to do, but I'm not going to really go into them.

Matthew Foreman [00:13:01]:
I'm not going to, you know, I'll probably read pretty quickly. So, all right. The first one, question 1, redemption of a partnership interest, right? An Italian company, King Information Data S.p.A., KID, or KID, is a diversified holding company that owns numerous businesses across various lines of business throughout the world, with many businesses crossing borders. Kid owns 40% of GKG LLC, GKG, a Delaware limited liability company that does business in nearly every state in the United States and a handful of countries throughout the world. GKG is a partnership for US federal, state, and local income tax purposes, and it operates through the GKG entity only. The other 60% of the owners of GKG are a variety of individuals and Delaware corporations, none of which would be considered related persons on the Internal Revenue code. All the owners of GKG, including Kidd, have held their interests from the time that GKG was formed. GKG is a pro rata partnership, and the allocations and distributions have no preferential items or special allocations.

Matthew Foreman [00:14:02]:
I said that because, gosh, that could get complex, right? Um, both Kidd and GKG are calendar year, and they both use a cash method of accounting. Simplify, simplify, simplify, right? I don't know if that corporation really could use cash. We're going to pretend this is, this is how tax works, Lane. So we're going to pretend, uh, the revenue expenses are 80% from the US, 20%, um, from outside the US. After a series of disputes regarding the direction of GKG, right, managerial disputes, owner disputes happens, specifically whether it should move its headquarters from Seattle, Washington to Cincinnati, Ohio, as well as the business, as whether the business should take on new investors to accelerate growth, KID has requested that GKG redeem KID's ownership interest in GKG. After the management and other partners held a discussion in accordance with their internal governance documents and the DGCL— that's Delaware General Corporate Law— an offer was made by GKG to redeem KID, which was approved by the relevant stockholders of KID. I'm just saying there's no like partner or shareholder whatever disputes here. As such, KID's interest in GKG were redeemed for $100.

Matthew Foreman [00:15:06]:
KID's basis in the units in GKG is $0, probably not plausible, but we're going to roll with it. GKG has a zero basis in its assets. Again, not plausible, but we're rolling with it. And this for this transaction, assume tax rates for the 2025 year. For those who don't know and didn't pick this up, you're not a sports fan. This is a reference to Ken Griffey Jr., of course, whose first name is George. George Kenneth Griffey Jr. And King information data comes from King County.

Matthew Foreman [00:15:33]:
Which includes Seattle. And at one point he wanted to get a large new contract. The Seattle Mariners, as they were, were disinterested in paying market value for players. That's why, that's why they had 3 Hall of Famers leave in a matter of 3 seasons. And yes, I referred to Alex Rodriguez as a Hall of Famer because he should be one. We're going to get into that another day. You can email me if you want and I'll delete it. I don't care.

Matthew Foreman [00:15:56]:
And then there's a number of questions which I'm going to go into in time. Also, and I'm not going to bury the lead, this is more or less Grecian Magnesite. Grecian Magnesite. I did this on purpose. I wanted to see how it came up with this. Grecian Magnesite was a case, and I'll explain it. If you don't know it, you can Google it. It's fine.

Matthew Foreman [00:16:14]:
There was a tax court case in 2017. TCJA, or what's commonly referred to as TCJA, changed the law so that the Supreme Court— or excuse me, the tax court ruled against the Internal Revenue Service. And then the TCJA changed it starting in 2018, and then the DC Circuit confirmed in 2019, and it overruled both of them, overruled Ruling 9132, which is the IRS's long-held position. But IRC 864(c)(8) is the same as the revenue ruling. And what it states is really, really that if you have a US partnership that has part foreign source and US source income, and a foreign partner of the US partnership sells the partnership, do you look through? Right? Because the C corp rule is if it's a US C corp and a foreign shareholder, there's no tax, which I'll get into in a second. That's a question in here. But if you have a US partnership, do you look through? And the answer according to the IRS was yes. And that's what Revenue Ruling 9132 said.

Matthew Foreman [00:17:17]:
The Tax Court and DC Circuit in confirming said no, no, no, no, no. That's not how it works. You don't look through. So Congress, as Congress is permitted to do, change the law 2018 going forward. This is obviously in 2025, although I didn't actually directly say this is in 2025, but I think that's a fair assumption using 2025 tax rates. None— no one really pointed that out that I didn't directly say that. I'm only thinking about it now. But this again shows that I did not share this with anyone.

Matthew Foreman [00:17:44]:
I didn't actually have anyone read it. That's not true. I had one person do, and I didn't set a You know, you can read it, but I don't really want, so I don't really want comments, uh, to be candid. So, okay. So summer associates, um, again, neither of them just for context have taken international tax. I think one took partnership tax, but not both. Uh, to be honest, I didn't expect any volunteers and they volunteered, so they should be commended for their bravery. I, I will tell you that.

Matthew Foreman [00:18:09]:
Um, and they're also both rising third years. Um, so they're entering their last year of law school and they should be commended for being willing to do it. I will not be naming names. If you know them, you can ask them, but they don't have to tell anyone anything. They are brave enough to do this. So for the first question, right, the first question is, is the redemption subject to tax in the US? Provide citations. They cited Revenue Ruling 9132 and that a sale of a partnership interest is taxable to the extent it's true with the ECI, which is what this is, right? Effectively connected income. and therefore 80%, you know, is taxable because it's 80% from the US and 20% foreign.

Matthew Foreman [00:18:51]:
I mean, they got the answer right, but they cited something that was— Revenue Ruling 9132 is overruled. So overruled by Grecian Magnesite, and then which was then overruled by 864(c)(8). Someone asked me if 9132 is good law again, and I said it's actually not. I actually would not they cite it even though it is the correct statement. It's actually relying on long-held IRS principles which are ruled to be incorrect by a competent authority, so, or competent court. So no. Question 2, what if the partnership where GKG was a C corporation? What if it's an S corporation, right? Yeah, they got this one wrong. They said it's taxable under 301, 302 redemption of stock, a C corp is taxable, which is true, but not if the shareholder is a US person.

Matthew Foreman [00:19:36]:
And then next question, what if it's an S corp? And they said, no, no, no, there's a foreign corporation shareholder, it can't be an S corp, you'd blow the asset election. So they got that part right. So that was obviously, you know, sort of a false answer. Obviously my love for S corporations goes very deep into the fact that they should, you know, never be used. But I think more importantly, possibly you throw it in here to show where they can't be used, right? The next question, what if the LLC were a general partnership? And they said no, it doesn't change anything. And they said, and this is actually really interesting, they said it could change it for foreign tax purposes, but I was only asked about US, so it's not relevant. So I actually got that part right. They took the next little note and thing today.

Matthew Foreman [00:20:19]:
Well, no, but maybe because if it's an LLC, there could be, you know, issues there. So I thought that was interesting that they got that part right. I don't know if it changes it for foreign tax purposes. I'm a US tax lawyer, and if we were doing this, I would Absolutely get Italian counsel involved. Absolutely. 100%. Because that's how I roll. I like to get competent advisors in whenever possible.

Matthew Foreman [00:20:39]:
Do you have any suggestions for ways to, for the continuing partners to lower their taxes? This was it. You know, they suggested a whole bunch of things. It's something that's really inapplicable. So I'm not going to embarrass them by saying what it is, but it's something entirely inapplicable. I'm just not going to say it. Not relevant. And then, you know, any other things they ask if there's real property for withholding tax purposes. I'm not totally sure how they know about withholding taxes because withholding taxes, if you never take international tax or not, maybe they've heard it.

Matthew Foreman [00:21:10]:
We've discussed it a few times on various meetings that some of them may have been on. I don't know. I don't remember who's on what calls and some of them were, you know, a month ago and, you know, my brain's a giant pile of mush these days. So here we are. But that was, you know, they spotted the issue. It's a good one. They also cited Grecian Magnesite here, and they cited how 864(c)(8) overruled Grecian Magnesite. So they clearly identified it as an issue, but they weren't totally sure what to do with it.

Matthew Foreman [00:21:35]:
So I thought that was really interesting. You'll notice they didn't mention 754. That last, you know, question 4 about suggestions, I want that 754 election. You need it there. So that's what was going on there. The next one's associates. I'm going to try to talk about these and I'll reference that they're different associates, but it's not going to be totally clear on who's who. I suspect if they listen to this that they'll know who's who without even trying.

Matthew Foreman [00:22:00]:
They're different ears. They have different expertise. They do different things. I think very highly of both of them, which is why I volunteered them to do it. All right. So question 1, is the redemption subject to US tax? One got 864(c)(8) and one did not. The one who did not did a full sub-K analysis. I don't think that one does any cross-border stuff.

Matthew Foreman [00:22:22]:
And the other one has done a couple of cross-border clients with me. So I'm glad that the one who— I'm okay with how it turned out. The one who did the full sub-K analysis did a full 741, 751 hot asset analysis. I mean, it was all of 2 sentences. Again, I gave them a total of an hour. So reading and stuff like that, you know, you have to think about it. So, you know, pretty, pretty good. Question 2, would your answer change if it's a C corp? Both said their answer would not change.

Matthew Foreman [00:22:48]:
So that's, that's wrong. What if it's an S corp? One just said it can't be an S corp and the other one discussed EMP. Obviously can't be an S corp. So that's, that's fine. You know, what if it were a general partnership? Both said it would not change their answer. They were correct that being a general partnership won't change the answer for US tax purposes. They both got it wrong. Here we are.

Matthew Foreman [00:23:08]:
You know, one got it wrong and kind of got it right. So, you know, such is life. Anyways, so for continuing partnership to lower tax, both said 754 election. So, you know, obviously we do a fair amount of that kind of stuff, right? Whether in estate planning or buyouts, redemptions, just general, you know, every orgs, do we need to do a 754 election? We're doing this, you know, getting, taking on investors, people selling. So they know it. I was really happy they brought that up here because that was really where the answer, that's what I wanted. Any concerns or other info that would be helpful? You know, one again talked about hot assets, which I thought was good. That person also discussed mixing bowl transactions.

Matthew Foreman [00:23:45]:
So they wanted to know— I didn't even think about this. I did not think about mixing bowls. So I thought that was actually really, really, really helpful because, you know, what if they'd only put the money in, you know, 2 years earlier, 8 months earlier, and there was a fight? And what did you put in? You put in appreciated assets and now you're getting cash, right? That could be mixing bowl. Could be a disguised sale, right? Especially, oh boy, under the new, you know, the new bill, OB Thrice, as I'm calling it. You know, there's not even a presumption anymore. You have to basically show why it's not, you know, so that's a tough one. And I thought that was good. I thought they did okay on the whole.

Matthew Foreman [00:24:23]:
I'm going quickly. I know I went slower with the summer associates, but that's because I was talking about the things more generally. But yeah, I mean, I definitely want to talk about the whole, how the US doesn't really tax C-corps for foreign shareholders. But, you know, it's worth a shot. And now for the AI models. So we actually did two AI models, which I'm going to take a quick break right now. We'll be back in just a moment, get some more music in, and then we're going to talk about the AI models because I think that's a really, really important thing. And we're back.

Matthew Foreman [00:25:05]:
So for this, what I wanted to do was I thought about, you know, what AI model to do, right? Should I just use the Google Gemini for free? Do I use a large pay one? Do we use a legal one? And I came out that I'm not going to use a free one. I'm just not. I'm going to use one of the large pay ones that's well known that I will not name. And I use the legal one. If you're thinking I'm going to send this person an email, I want to show them what mine can do. You can send that email and I will delete it. I'm not interested in evaluating AI models. And if you'd like me to help you improve your AI model, I will do that.

Matthew Foreman [00:25:44]:
But I'm going to charge you my hourly rate, which is payable in United States dollars. And I appreciate that people say that, you know, that's that like, you know, I should help it and this and this and this and that. And I'll explain it to you this way is that I get paid for providing legal advice. And you, you know, in this situation, people get paid to create an AI model. And I always tell people that I don't get paid on your AI model and you don't get paid when I provide legal advice, you know, so, so pay me, you know, and I'll help you. That's fine. But I'm not trying out new ones. And if you want me to help you, you know, pay me.

Matthew Foreman [00:26:19]:
If that's harsh, I'm cool with that. I'll also justify it, right? So AI model number 1, a large general model, surprisingly slow with answers. You know, we actually had to go kind of let it ponder for a while, went over to the other one and then came back, which I thought was interesting. So it's sort of like dealing with me, whereas I sit there and go, oh, that's an interesting question as a stall mechanism, right? So question 1, US tax, nailed it. You know, it cited 864(c)(8). It did not mention Grecian Magnesite. It got the 10% withholding under 1446(f). You know, it was an interesting question I had.

Matthew Foreman [00:27:02]:
I'd have to research this, but is it 10% withholding on the total amount or the 80% amount? I don't know. I don't know the answer to that question, so I'd have to look at it. What I focus on withholding is know when you need withholding or when I need to research whether there's withholding. I don't worry so much about whether you actually, you know, what the actual amount withheld is, because I can research that. I can get that. Plus we have, you know, one of our counsels at our firm, you know, the only one in tax, spent more than two decades at a large multinational corporation and doing cross-border tax and subsea work. So anytime I have a question, I go to him because he's really good. He's really good.

Matthew Foreman [00:27:38]:
So that's helpful. What if a C corp, they got that right, no US tax, US cannot tax, and they use cannot as one word. It's actually false. US can space not tax, who's able to not tax. They actually could tax, they could, they could tax them. The US has made a choice not to tax. Most other countries don't tax it in the reverse situation. So the US is doing it as a way to create an incentive through taxation, which we clearly do a lot of, and every country does a lot of it, for foreign investors to invest in public companies, which are the predominant area of C-corps.

Matthew Foreman [00:28:14]:
So that's why. So cannot, it said, is actually wrong. And I'm nitpicking, but not at the same time, because the US is allowed to, it just doesn't. What if an S-corp? Nailed it. Can't be an S-corp. What if a general partnership? Got it, got it right. You know, general partnership doesn't make a difference. Planning ideas mentioned 754 election, mentioned a C corp blocker, you know, interposing a C corp between the corp and the partnership and the foreign corp.

Matthew Foreman [00:28:42]:
I thought that was interesting. Maybe, you know, you'd have to sell the C corp to lower the taxes. It's not really sure anyone would buy it. You lose the 754 step up. So that's kind of an interesting one. I'm not really sure it works. Notes 199 cap A, which it seems to think applies to every business. Which it may not.

Matthew Foreman [00:28:59]:
I don't know. I'd have to see what the businesses actually do. But it's a good issue spot to think about. Although, you know, I don't know if you actually— I guess because it's ECI, you might still get it because it's as if you sold the underlying assets and it passes through. So maybe I have to look at the mechanism for it. And then it mentioned a cost segregation study, which just like, no, I guess, you know, you get 754 election. Oh, you want to depreciate it faster, I suppose. Maybe.

Matthew Foreman [00:29:27]:
Sure. It's kind of a, you know, man, people just talk about, oh, just do a cost seg on it. And you're like, I'm— it's land. You can't do cost seg on land. You know, you can't depreciate— well, you can depreciate land in really narrow circumstances, but you can't— but generally you can't. Someone was asking where you can depreciate land. And I said that, you know, golf courses, if you like create a green, you can depreciate the green, which is just weird, but whatever, you're allowed to. So here we are rolling with it.

Matthew Foreman [00:29:53]:
So you're depreciating land, right? Any additional information? It referenced the tax treaty. It talked about FIRPTA. I was kind of interested that no one really talked about FIRPTA here because I don't talk about what this business does. And, you know, what if it owns real estate? It's a partnership that owns real estate, FIRPTA, 15% withholding, bam, you know, or 10% through a partnership withholding partnership. So I thought that was kind of an interesting comment that no one touched it until then. You know, got a little withholding stuff, but not really there. Talked about state tax withholding, which is good, especially if there's real estate. And it, you know, said, hey, we, you know, definitely think about the 754 election, which is dead on.

Matthew Foreman [00:30:30]:
You know, 754 is really the thing I wanted. The AI model number 2, this is a legal AI model. It is targeted at law firms. Okay. So subject to U.S. tax. Yes, 100% taxable, which is wrong. 80% taxable, my friends.

Matthew Foreman [00:30:46]:
80%. It noted the 1446 F withholding and the 751 hot assets. So it got that right, but then kind of, you know, got 864(c)(8) wrong, which is interesting. And this is a pause for a second, but so I asked it to check in light of Grecian Magnesite and it still sort of got it wrong. It mentioned FIRPTA if there's asset, you know, this and I'm like, well, FIRPTA is kind of binary, but more or less admitted 864(c)(8) exists and it changes answer. But I always find it really fascinating. It's like when you catch an 8-year-old doing something they're not supposed to do. And I don't have kids, but I think we've all been around a cousin or that we've been that child who— did you take them? No.

Matthew Foreman [00:31:32]:
What about this video of you taking cookies in the cookie jar? And you're like, ah. So anyway, about Grecian magnesite, if I were dealing with an 8-year-old I caught doing something wrong and they wanted to discuss Grecian magnesite, man, they could have thrown a child down the well and I'd be like, that's cool, man. Have at it. Let's talk Grecian magnesite, you know? But it's troubling that it gets, you know, one thing that irks me about AI is it gets it wrong. You ask it a question, it's like, oh, about that, sorry, my bad. And I'm like, well, shouldn't you have gotten that right? You know, what are we doing here? And I guess that's what humans do. So maybe not. I don't know.

Matthew Foreman [00:32:02]:
Question 2: What if it's a C corp? No taxes, C corp and foreign shareholder? No, it noted FIRPTA here, which is correct. and it notes the US withholding, which is good. What if it asked about, you know, the S corp question? What if it's an S corp? It got it kind of backward. It said that Kidd couldn't be a shareholder, which is in an S corp, which is correct, but it's not so much the fact that Kidd couldn't be a shareholder, it's that Kidd being a shareholder in GKG blows the S election. So it kind of got it backward. It's sort of right. It'd be enough to trigger you. But again, this is where, you know, sort of my conclusion comes in that They still kind of need us.

Matthew Foreman [00:32:37]:
What if it's general partnerships? It's no change, which is correct. Any suggestions? 754 election. You know, it's sort of amazing to me that suggestion 754 election. The thing about that that amazed me is that both AI models got this right and both of our associates got it right. And I see it missed in practice. I see, I see CPAs with 20, 30, 40 years experience. I see attorneys. Um, not so much tax attorneys, you know, which isn't like a brag, but like tax attorneys are really up on 704 elections.

Matthew Foreman [00:33:11]:
We're not so much worried about like, um, whether something needs to be a cash or accrual method based on gross receipts, just not something we focus on. So because of that, we focus on things like 704 elections a lot. Um, we sort of deal with situations where they come up a lot and I'm sure there's some that miss it, but I see it missed by return preparers who are licensed and have very relevant experience. So that always surprises me. It then said, you know, consider being a C corp instead of partnership. I'm sure the US owners would hate that. So talk about reviewing state apportionment. Thought that was a really good, really good thought.

Matthew Foreman [00:33:42]:
The other one, you know, didn't really talk about that, but I thought that was a good one. You know, make sure you have it right. You know, are there state taxes? How does that go in? You know, states have very different rates. There's UBTs, there's PTAs, which I'll talk about later a lot. So I thought was interesting. It talked to— look, look at, you know, foreign tax credits and withholding obligations, you know, just things to look at. And I thought that was good. Any other info? It went through a list of what was just a lot of what it already discussed.

Matthew Foreman [00:34:10]:
It said the check basis, it's kind of weird that it's zero, which is correct. It talked about the installment agreement, you know, whether there's an installment method under 453. I thought I was like, where did that come from? And I looked at the citation. I'm like, I don't, I don't think this is relevant, but It could be something, can lower your taxes, I guess. Non-US tax concerns. So it said try to plan now for non-US. I thought that the only one that talked about like, hey, talk to a tax advisor. I guess the other one's a tax treaty.

Matthew Foreman [00:34:36]:
So look into that. So what's the conclusion on this one? Right. You know, I did pretty well. I gave it a narrow focused fact pattern that had a lot of relatively recent writing that gave it the answer. You know, and then I gave it narrow and focused questions. So I thought that was an interesting one. You know, the next question is going to go into the next episode, which I am going to record right after. So you're going to be able to listen to it right after if you want.

Matthew Foreman [00:35:07]:
I know that's a lot of me, but, you know, what can I say? Like, what could be better than listening to me talk for another half an hour? With a really weird fact pattern. So we're going to end it here, going to a little, little cool hip podcast music, and we'll be back with episode 34. You know, I hope you enjoyed 33. The next one's 34, basically on the same topic. Be back in however long it is until you want to listen.