TrufflePig AI vs Excel for Finance Teams from Building Models to Real-Time DCFs with Ian Schnoor

Foreign. The Mod Squad featuring Ian Schnorr, Executive director of Financial Modeling Institute. Giles Mail, humble MVP and co founder.

Of Full Stack Modeler and Paul Barnhurst.

The FP and a guy. This will be the first tool we've tested that's 100% a new spreadsheet experience. They like to call themselves the AI First Spreadsheet.

I like that it built it and it linked it. It built the working capital schedule and it linked it in here. This is the correct flow. This is better than some other tools are done.

Something I thought was really unique. They say, hey look, you know some of the other tools demo real well, we may not demo as well but we believe we do a much better job of the real work.

Let's look at the appreciation. PPE Schedul beginning balance plus capex minus depreciation. Excellent job on that.

What we've done here is a little bit more of the real work and it has performed better. I think what they said is true.

Done a really, really nice job.

If you're wanting a third party spreadsheet AI tool, definitely one I'd consider, no question, 100%. Today we're gonna be testing kind of a unique name. Welcome back to another episode of the Financial Modeler's Corner, the Mod Squad. Unfortunately, once again, I think the third week in a row we're missing one of our three core members. Giles unfortunately wasn't able to join us this week. Last week in and Giles went without me and the week before it was Giles and I. So we're running at two thirds capacity today. But we're super excited to have you. I'll do a quick intro of myself and then I'll introduce my co host and we'll jump into things. So my name's Paul Barnhurst. Obviously this is Financial Modelers Corner. Thrilled to have you here on the show. We're testing AI tools so you don't have to and helping you give an understanding of what they're capable of doing. So why don't we take a minute and let you introduce yourself in.

Sure. Thanks Paul. I was going to say thanks Giles because he and I did it, as you know, last week when we tested another tool. But yeah, we had a great time last week and. And you and I are going to have a good time this time this week. I am Ian Schneuer, executive director of Financial Modeling Institute, fmi, the world's only financial modeling accreditation organization and community. And we are thrilled to be, you know, I'm thrilled to be working with you on testing all these different AI modeling tools. Modelers all over the world have been really curious to know what's going on in the world of modeling and AI these days. And my entire career has been focused on modeling, from my time as an investment banker to working to running, you know, a training firm, having the honor of building the financial modeling that's part of the CFA program. So I've been really focused on modeling for many years.

I know you have. You built many a model and we've tested. Now I think today is tool number seven. So I'm just going to run through what we've done. You can find all the episodes on YouTube under financial modelers Corner at the end. They're the last seven episodes, I think, that we've done so far seven or eight. But we've tested Excel, Agent, Rosie, AI Trace, Light, Tabs, AI Shortcut.

Yep. El Car, Elkhar.

And today we're going to be testing kind of a unique name, Truffle Pig. What do you think of that name for a spreadsheet?

Well, Truffle pig, you know, there is a meaning behind it for a spreadsheet. Gosh. I mean, naming anything, it's memorable. So I don't mind that I like it. It's certainly unique. It's easier for me to remember that than some of the other names that might have been made up names.

But yeah, for sure. I mean, I try and decide if I like it or not, but I totally agree as far as the marketing of being unique and remembering it. Seeing the big pig and truffle pig is really easy for me to remember versus some of these others that are words I've never heard of. Right. They've made up a word as their business, so they do have that going for.

And, you know, we'd have to ask them why they named it that way. Right. As you know, there must be some connection. We can figure it out. I mean, it comes from the origin, I guess. A truffle pig, which I remember reading about once is, is a type of pig that is used to dig up truffles. Right.

Yeah. And I know they're focused on trying to find insights, you know, or trying to get the work done so you can find the insight. So I don't know if the idea is, hey, let us take care of all the drudgery so you can focus like, you know, the pig on finding that insight or that thing of value.

Yeah, I. That's probably right. That's probably right. And it certainly is memorable. So we'll give it. We'll Give it a shot here today, won't we?

All right, so why don't we start. We'll show their website, we'll go through the little one pager we have with them and then we'll jump into testing. One of the first things is this is. Well, this will be the first tool we've tested that's 100% a new spreadsheet experience. When we tested Shortcut, we did test Shortcut for at least one problem in their spreadsheet. Then we went and did the rest in their Excel add in here, 100% spreadsheet. They like to call themselves the AI first spreadsheet. Initial thoughts on new spreadsheet versus Excel add in and I know you're a career long Excel user, so I'd imagine, like me, you probably have some strong, strong feelings here.

Well, I mean, I've done some initial preliminary testing and you know, it's certainly been working well from our preliminary tests as a tool. Yeah, I don't love as much having to learn a new sort of quasi spreadsheet software as we'll demonstrate. It does some funny things. But listen, if it's the right, if it's ultimately the best and the right tool, then I guess that's, you know, people will be fine using it this way, but they've decided to have it in its own, you know, app as opposed to an Excel add in. And I'm not sure if you know why they did it that way.

My, my understanding from the con, I had a conversation with one of the founders back when and I believe it was this one. I've talked to a few, but usually what I hear when they build their own spreadsheet is they're not confined by Microsoft. Microsoft rules the older tech. It opens up opportunities that are easier for development. Right. That's generally the argument for its own spreadsheet, which makes total sense. And then the counter argument, everybody does Excel, everybody's so used to Excel, why do I want to fight against it?

Exactly. No, I could see that, you know, you have your own. You could export it into any spreadsheet tool. Right. And there are not hundreds, but there are a handful that are used, of course. And so allows you to work with anybody, I suppose.

Yep, that's kind of my thinking is really being able to control all the UI and UX and having ownership of everything. But that also creates more work versus just being the agents. So we'll see how it works. I mean, at the end of the day, like I said, I could tell you if it performs Way better than all the other tools. I'd be open to using it more.

Totally. Yeah. No, you get used to it.

I mean, a spreadsheet is a spreadsheet. At the end of the day, they're not that different.

Right? No, that's exactly right.

So, all right, so let's just jump through. We'll do a quick look at the website. You're pretty lightweight. They're still in an open beta, but they do charge during the beta, which I think is kind of interesting. Usually you see kind of a free period, which they did have. So, you know, we got the AI first spreadsheet. Finally a spreadsheet to speak your language, describe it and it's done. You know, and they show an example of a DCF model. You know, pretty typical. Skip the busy work, focus on the insight. So, you know, get great answers fast. Build presentation ready workbooks. Work in one contextual workspace. You'll understand your workflow. To me, what it feels like is, you know, their marketing here is. Here's what you can do. Like some of the sites, it's hype. We're going to save you a bunch of money. We're better than everybody else here. They're basically, it feels to me very a basic side of saying, look, we're here to help you.

Yeah.

I don't know what your thoughts are. You know, work at the speed of thought.

I like their vibe. I like their messaging and their articulation. I like the layout. I mean for somehow Giles and I were talking about this last week. A lot of the AI finance tools that we've looked at have a similar look and feel like if you scroll up, I mean, you know, a lot of nice white space and a bit of a demo and you know, there seems to be a more similar than not look and feel.

But agree. I agree with you. They're more of a similar look at Phil than not.

Yeah. And they. But I certainly like their messaging and their vibe so far. Totally.

One thing I think is interesting, this speaks to the AI advantage. You know, web search AI intelligently pulls facts into the grid as you work. Ask questions, compare sources without tab hopping. So, you know, a little easier to do the AI, pull it in, pull in from another source and compare it. That might be a little tougher to do in Excel. Probably could do it with Copilot and some of those things. But I'm going to guess just kind of looking at that, that to me strikes me as probably something that was easier to do, you know, and a dedicated the way they Built it, but don't know, you know, and then obviously, like everyone we have, we secure your data, it's encrypted, it's going to be safe. I don't see the sock compliance yet. That's probably why they're still in beta. I imagine we'll see all those logos like the others when they're done and then get started for free. So very, very lightweight pricing. Pretty similar to the others. They have a free version. If you do an annual, it's 21amonth. If you do monthly, it's 30 for the Pro and 200 for their Super. And, you know, if you want a team plan, can you look, can you.

Go back up the description? So, you know, Giles and I were unsure last week on the pricing whether, whether you could get, you know, what, you know, for this, the pro plan, it says you get unlimited default Mode Messages and 15 Max Mode Messages per month. Again, it's not totally clear what that distinguish. What, what determines and what differentiates between.

You select it doesn't show it here. I played with the tool enough. There's a place where you select default or max mode and then it. Right.

But I guess it might not be clear when you're doing that whether you need one.

Oh, yeah, I get it. That makes.

Right. And should I be using. For this particular request, should I be using the default or should I be using one of my 15 max credits kind of thing? So, right. You kind of.

No, I think that's a great point. It's hard to tell from there. You don't, you don't know for sure.

Anyway, let's take a look and I'll.

Show one other thing real quick. Let me just pull up here. So, you know, we've done some of these on the others beta, it's a new product. Launch date was April. So they're, you know, eight, not even eight months in yet. They kind of focus saying, hey, we're focusing on doing the, the spreadsheet. Busy work. So not specifically modeling. They're broader than some of the other tools. They say their ideal customers, finance, operations and sales team. So they're just going after the hardcore spreadsheet or so to speak, you know. And then something I thought was really unique, this is their claim is they say, hey, look, you know, some of these other tools demo real well. We may not demo as well, but we believe we do a much better job of the real work. You know, couldn't validate the claim. That's what they say. It makes them unique. We'll put it to the test here in a minute. So. All right, let's jump into the tool. I think we'll. What I'll do here is I'm going to log in and just show the basic spreadsheet. So let's see. Can I get a blank sheet? I should be able to. We go here new and so this is what a blank spreadsheet looks like for them. Maybe I'll make this a little larger. You can import your, you know, Excel or whatever you want here, up to 30 megabytes. So they're not going to let you import that 50 megabyte Excel file?

No, no.

You can download them as an XLs or a CSV duplicate something. And then they have just a very basic spreadsheet. They have their chart thing, you know, some filtering grid lines.

I don't even.

I'm not sure if I see a pivot table here. So it looks pretty lightweight, but I think the real point is to see how it does with its AI tool. So any. Any other thoughts on the. On this before we jump in? No, no.

It looks. It's pretty accessible for anyone looking at it.

It seems pretty usable. So I have the first case here, and what I want to do, just so people can see, is I'm going to pull up the actual case in Excel real quick. I'll do that while we test it, and I'll come back to that because it'll take me a minute. I don't want to spend the time. But notice the images aren't here. Anyone who's followed the episode, you can see the name is kind of when it was pulled in.

So, yeah, when you open the file, when you opened it, it sort of deleted all the. All the.

Stripped out a lot of the images. Yeah.

Yeah.

And so we'll show you what it looked like in Excel. So that was one of the first things that jumped out to me. But what I've done is I put the same prompt that Giles always puts in here. You'll go ahead and analyze this case. This is a basic Esports case. Notice that it's going to give you an example at the start of each of them, and you need to solve the problem. There's five bonus questions, several seven levels. I've said please solve level one first before completing the other steps so I can review it. And what I'm going to do is I'm going to go ahead and use the max, their most powerful model. I figure since we have access, we might as well.

Sure.

And let's go ahead and kick it off and see how it's doing.

Yeah, let's give it a second here before we pause, because let's see what it's doing and what it's thinking about. We, you know, last week when Giles and I were on our own, we were just about to pause recording and it was done, it finished. So let's see how quickly this takes. But It's.

So level one's done. That looks right. Updated 10 cells. It's thinking. So pretty quick. It's only do level one perfect. The formulas have been applied.

Results.

It's giving me the solutions. It tells me the formula it used.

Can you click on the. The green cells in?

Yes. Do you want to keep these changes? So what it does first is notice. I like this. It puts a box around it saying, hey, here's all the area we're working on. Do you want to keep these changes? So I can click in and see. Oh, it did left. That looks right. I 53. What surprises me is I would expect this to be red over here, kind of like Excel. You can see the blue around it. And you see that. Oh, there we go. If I click in up here, it gives me all the columns. Okay, so that all makes sense. I'm going to go ahead and say yes. And then what I'll do is I'll ask it to solve the rest of the. Please solve the remain levels and bonus questions. And while it's doing that, let's keep talking and I'm going to pull up just so people can see what the case looks like in Excel so they can kind of see how it's different.

Sure, sure.

So I'm doing that here now as it looks like it's going pretty quick. It's doing level two. It's analyzed that. So we got. All right, have the case here and we'll bring that up in Excel while it's thinking. So it's done. Level two, that's done. It's on the level three. It looks like.

Yeah, we'll have to see if it's done correctly and we'll have to take a look at the nature of the functions. But there you go.

So here's what that case looked like. You could see quite different. All the color and stuff was really stripped out. So you could see, compare. If I split my screen here, let's come up to the top. Well, it's working. And we'll minimize this for a minute. So if I just scroll up to the top right, all blank there. You can't even see. So you can see the wording here, but even that, like the spacing and so definitely pulls it in differently. That's something that I imagine will get better with time.

Yeah.

To be more on parity with what you'd see in Excel. All right, why don't we pause it for a minute? It looks like it's still working.

Let's see how is it working or is it done? Let's see. Well, why don't you click before you pause? Just click on level three. Let's just see what the formula looks like in that.

All right, let's take a look at level two first.

Let's see what the formula.

I just hard coded the numbers well.

Oh, wow. Okay. So it's not. Can you click on the. You did the formula here.

So if I click here. It did a hard code. Level three. It did what you would expect to call an indirect. That's fine. Put the column around it. I think all those answers are right.

You have the answers. Do you know if those are the correct answers?

It does concern with the previous I do. I. Yeah, those answers are all right. From what I can see, they all look correct.

Okay.

Level. Level four looks all right. And they did a formula again, which makes sense. So I don't know why they didn't do one for level two. We'll have to ask it. That all looks good. Level 5, let's see how it did. Those all look right. It did a formula again. Yeah.

It's just fascinating to me. The solutions that the different AI tools use are pretty significantly different from one another. And I would expect that because different people would solve the questions differently as well. But they're solving them in very different ways. Some of them are much longer, more complex, others are much more manageable. Like, this is actually looks like a reasonable solution.

This one. If I look at all these, they. They look short so far, man. Right. I'm not seeing anything long any let. Yeah, these all seem pretty reasonable. I mean, I might solve them differently. This one I would do that way, you know, this one here seems pretty straightforward. There's level six that hardcoded again. Interesting. I don't know the reasoning. It's the right answer. No, some are right, some are wrong. First two are right.

You want to ask it why it chose to do that or if you want to.

I think we'll do that here in a minute.

I mean, it's done though, right? It's obviously done. And what about it looks like it's done.

It missed some of level seven. So some is right, some is wrong. There. First one's right. And then it looks like it got off on most the rest. So interesting. So not quite as good performance there.

Can you go to the very bottom?

Do the bonus questions. Let's see if it's as it's done.

Is it done?

Oh, it's still solving the bonus question.

There you go. Bald's doing the bone. And there it goes. It's putting them in. So it saved them.

The first two are right. Get the formula. I don't know. I don't know why it didn't upper concatenate. Oh, because you're looking for a certain letter. It assumed capitalized because it put the E. So that makes sense why it would do that. Okay. Matchmax. Fairly simple form. I'm not sure where they hard coded a plus 10. Oh. Regarding questions 11 to 30. So maybe that's how they started it. I don't know. The108 is right. This is wrong here. Slightly off. And then this one is way off. But given. I mean, the map looks like it came through right. So I don't know why I got the colors so wrong. It's supposed to be 434 and it got 31. So we'll ask it here. We'll let it. It's going to give me its summary probably another 20 seconds. And then we'll ask it why it did not use formulas on every level. So I'm just going to say yes for now. Why did you hard code some of the answers? You were supposed to include the formula for every level that you use.

You're really scolding it now.

You're really.

You getting to scold it and aren't you? Let's see what I said. No.

So you'll laugh at this and maybe our audience will. We did an interview on one of my other podcasts and we had some guys from Netsuite and one of the guys is like, you got to treat it like a junior intern. You just gaslight it and just talk. It's like you'll get better performance from it. You told us we gaslight our AI. You'll get better performance. Performance. So I need to be meaner actually is. Whatever.

Okay. Okay.

So it said, you're absolutely right. I should have used formulas for all levels. Let me fix the levels where hard coded values with proper formulas saying the formula for level two, which. So let's see what happens now. Got an error. It's thinking.

Oh, I see. It's. It's thinking about the levels that it put in dead numbers.

I will Say this one ran faster than most of the Excel ones. It feels like not a lot. I mean, not El Car, which I tested last week, but I think probably on par. Maybe a little faster.

Maybe. Yeah. I mean, other episodes we've definitely had to pause before we could even see anything.

But I. What I do like in this one is I do like. You can clearly see where it's working and kind of that color box around it.

It's nice. Yes, Yes, I would agree with that.

Unfortunately, parsing the delimited text in level two with a singular formula is challenging without modern Excel functions like text split, the formula I attempted aren't supported in this environment.

Oh, that's fascinating. It's fascinating. It's claiming that it understands how to do it, but the software doesn't have the capacity to use. You know what? We may have just answered our previous question when we were. When we were applauding the simplicity of some of the earlier formulas, which is good. It might have been limited in its. Based on the. It hadn't occurred to me. Did it occur to you they hadn't?

Like if I tried typing. Not all textbook. Right. Here is a formula.

What's that?

Why is it interesting? Because. Right. It's showing and what I do like is it gives. It gives the column to it. That's okay.

It is there. It is there.

It's right.

Yeah, it is there.

Sure. Go ahead and make the changes. I'm going to ask it. Text split is available as a formula in this spreadsheet. Please use it and then we'll move on. We'll take one last look and move on to the next task in the interest of time for level two, I'd say for answering. Sure.

Let's see what happens.

Thank you. See, I'm being nice now.

Now you're being nice to it. Yeah. Let's see if that changes. It's interesting.

It told us it can't, but then it. Well, let's see what it does. Let's look at the others and see now.

So.

Yep, that one was good. Like we expected that one. This one has it. This one now has the formula used an index row direct. Let's see if we got them all. Right, now that's level six. Still got it wrong, though. The number.

Are they all wrong?

No, on level six, they. It's hard to tell if all of them wrong. Most of them are wrong for sure. Looks like almost all of them are slightly off by a. Trying to see if they're all off by the same number. No, they're not. So I don't know what it did. And then level seven, they're mostly wrong. So it struggled with the mapping and I don't know if it brought something in wrong or similar formula to the others. It could be the way it brought stuff inside, but. All right, so let's see if it fixed number two and then we'll move on here. So it's still thinking on level two. Ah, interesting. It tried a sum product and now it cleared it. I don't know why it's not using a text split. Interesting. You're right. Of course. Right. AI always defers to you that you're right whether you are or not.

That's right. That's right.

And calculate the sum of the two largest values. So interesting. Now I want to see what this one fishes. We'll let it work for just one more second and then we'll move on. Yeah, thoughts so far as you watch this. What's kind of your take here?

I never want to lose sight of the fact that it's impressive that it can do anything that you can, you know, enter an English language prompt and get an answer and it knows how to read it, find where it needs to go and populate some sort of a formula. I mean that's still mind boggling. Yeah, I don't love that it, you know, hard coded data that it seems to have gotten a bunch of things wrong from what you are telling me, right from looking at your answer queue. So I don't know.

How about you?

I sort of mixed on it. What do you think?

Yeah, that's kind of my view. As well as mixed. It didn't seem to perform quite as well in giving the answers. The one area I was impressed on, the answers it did give that were formulas, all of them were simple. If part of that is how the AI is thinking about the spreadsheet, that could be why. Yeah, but I always appreciate when you give me a simpler answer over a more complex answer.

Absolutely.

So I do appreciate that part. We got that answer right there. Now gave me a formula. We'll give it one minute here. Okay. Complete it. See, it just won't complete level 2. Interesting.

While textsplit is available, it says using it with large function encounters compatibility issues. In this Excel environment it's saying this.

Would be an answer, but it didn't put that answer in. All right, interesting. So clearly a little bit of compatibility issue, I think we call it. Good. I'm going to say yes, we're going to close that one. Let's just download it. So I have it. Ah, you can download as Google sheets. That is nice. I. Oh, I guess they optimize it.

Yep.

Totally.

Totally.

I do. I do like that. All right, so now here's the more complex one. I'm going to shrink this down a little bit. So this is the Giles case.

Every get. It's. It's kind of made a bit of came through though.

Right.

Is that the image?

Look how the image came through to what it normally looks like. Again, really struggles with the pictures and stuff right now.

Right, right, right. Rife. Yes.

And just for our audience, I'll show you what it looks like in Excel real quick so you can see the differences. It can't bring in the pictures. It kind of struggled with getting the stage quite right. If I compare the two there, you can see that the audience pit, the orchestra looks fine but without the rest of it it looks really weird. But we don't really care so much about the look for this. We want to see how it does on the case. So let's go ahead and do the max again. I would expect it to really struggle with this case given it struggled with some of the compatibility and formulas for a simple case. Yeah. Let's go ahead and put it in. Please complete Level 1 before proceeding to the other levels. Thank you. See, I'm being nice again.

You're always so polite.

You made me a conscientious of whether I'm being nice or not in.

Yeah. No, you have nothing to worry about now.

I'm going to be self conscious. No, it's all mid. I'm satisfied. All right. Why don't we pause it for a minute and we'll come back I imagine for a minute or two here. So we're going to pause.

Great idea. Back in a moment.

We'll be back with you here in a moment. All right, we're back. We gave it about. About 12 minutes or so to run solve level one. Did that beautifully. Got them. All right. We can insert those answers if we want. It did level 2, hard coded them and then it struggled.

Well, that's not great though. Right. Level. Level two.

That's.

That's not ideal if it's hard coded.

No, hard coded. It got them all right. But it didn't give us the formula like we would have liked Level three. It's hung up on what's interesting. What I did notice, and I know we've seen it with others, is the answers are in this file and it was able to look at those answers and say, let me redo it let me redo it, which I appreciate. You could see the learning and figuring it out. Also. It's a little like cheating, right. If I was in esports and I had the answers, makes it a lot easier. Yeah.

And it's obviously it knows what it needs to work towards, so. But it's. It's chugging here. It's chugging, right? It. It's gotten through a few level levels. It's really chugging to work through the rest of them.

It definitely is. You know, it's on a little bit of the longer side for other tools. Generally been a little quicker. I think the longest we've gone is about 14 minutes and we're right about 14 minutes now. So I'm going to go ahead and kill this one and we'll move on. I mean my take is maybe it would have got there on all of them but took a little longer, didn't struggle a little bit. But not bad. I mean we can see similar to all the other models. It's getting some of the answers right. It's sometimes following instructions. So, you know, similar. The things I like is you get to always say yes to apply it. You can see that I think it's thinking is good and it shows us that. But let's move on to the next one. This is one of the areas I've seen an analysis of them before and when they analyze some data and put together kind of report, it seemed to do really well. So I'm hopeful here we get a kind of a good result. So we're going to take the same prompt that Giles has used every time. We're again going to use the MAX model. Let it think a little longer and we'll go ahead and kick that off.

And what are you asking it to do?

So what I'm asking it to do. Let me read through it. So we have great point here. So if someone's coming in for the first time, this is just a ledger data. So what we have here is we have a monthly P and L summary for ourselves so we can see it. We have a chart of accounts and then we have a trial balance. So you know, different accounts, product sales, revenue sales department and the. The credit and debit amounts. And we're asking it to analyze the data and I would like you to produce a comprehensive financial summary of the data, the performance of the company on a separate tab. Make the summary professional. Include visuals with a mixture of KPIs and graph charts. Review your own work before completing this task. The final summary tab must Be top quality for a senior board meeting presentation. So we're basically saying, give me your presentation.

Let's kick it off, let's give it a moment to see how it performs here.

So I can, you know, if we go through the thinking here and I like that it leaves it and then minimizes it. So the user has. Here's the data it has. The user wants me to do the following. It breaks it into steps and then it says I should do this. It analyzes it and creates the report of the 11 steps and we'll check them off so we can see it's thinking. It's done. You know, it's read the data now, it's working on the exercise. We're going to pause. We'll give it a few minutes to run here and we'll come back. Welcome back. I wanted to let everybody know we have a third co host with us today. We've replaced Giles for a few minutes. Do you want to introduce our third co host in.

Yeah, this is my. This is actually my dog, Giles. So we've got another Giles. No, he's. This is. This is Benny. Benny is a cockapoo and he's a very good financial modeler, actually. So he was curious to see what we're doing and so I thought I would bring him along for this part of the demo.

Alrighty. So we can see here. It thought for a few minutes. It gave us a summary of what it did. You can see all the key highlights of the complete summary. It had 11 steps. It did here. Although it's interesting, it didn't check all of them off, even though it says it's done. So that's a little weird. Visualization created and it says yes. So we can see where it worked. Let me pull this back. You can see all the areas that worked. I highlighted everything. We're going to go ahead and hit yes and you know, decent formatting. The one thing it didn't do a good job of is I would have liked it to fix the column width. You can kind of see not great there. But as far as I can tell, I did some checking. All the answers appear right. You can see the formulas that use the sumifs. It used a sumif here. It did hard code the values which I would have liked to see it link here. But, you know, not. Not the worst. It did give us. Let's look at. It did a P and L trend analysis. Then it did a revenue breakdown by category which totals to the revenue up here, which is the right number. And then you can see it showed what percentage of them, all of them were what the average monthly was. Did something similar for expense analysis. Looked at both cogs and operating expenses.

It's doing a good job.

Yeah. Key insights. And then it even did a year end financial position. What's your current assets, what's your fixed assets, what's your liabilities? I don't know if I've seen another tool get into that, but one of the better analysis and the area I think it did the best job on is these graphs for the most part are really easy to read. The only thing is I have to move this out of the way to, to see all of them. I would have liked them to be over a little further so that that's not in the way. But the graphs are relatively minor.

I would say that minor. Obviously it's not a. Right.

Yeah, right. If it does all this right, I could move. I can move a graph, not.

You can move them. Yeah, no, I'm, I'm impressed.

So yeah, I mean I noticed this got cut off a little bit the gross profit. But again on the ledger, I don't know why if I enlarge it but on the whole one of the better I've seen. I'm impressed. I think it did a really good job here. Any thoughts before we move on to modeling?

No, I mean it. Again, I agree with you. I think it's understood the ask and put together at least a very, very strong starting point with some visual aids.

Yeah, this is something. If you gave me this, I would ask it to get rid of the zeros, but I could either work with it or clean it up in. In fairly short order. I think I could have something that's presentation ready and that's what we want.

Yep.

All right, why don't I pause here. We'll get you set up and then we'll be back in just a minute with you folks. Right, we're back. We've completed the first three exercises from Giles and struggled a little bit with some of the Excel stuff. Did pretty good on level one. Definitely struggled a little with level two. Took longer than plan and then did a very good job on the executive summary. We asked for one of the best we've seen. Now we're going to put it through some of the financial modeling pieces. So in. Why don't you take it away?

We are, we are. Although the first thing I want to do is something a little different. I'm going to throw a bit of a curveball. When I was loading up this tool When I was getting Trufflepig set up, it said, you know, it told me what I can ask for. It said, you can ask for a lot of different things and say you can ask it to pull some public company data and build a dcf. And it literally said, it gave me an example. It said, you can tell it to build me a comprehensive DCF model for Nvidia using real financial data. So I thought I would take its exact prompt and see what happens. So I'm going to toss in its prompt that it told me I could ask for and I'm going to put it in there and see what happens. I've got nothing on my screen. I just said, build me a comprehensive model for Nvidia using real.

We want to try, let's see what happens with the default and then if it doesn't, we can try the max real quick and see if we get a different answer.

I use the default. Yeah, I use the default. And let's see what it does now. It has, it has stopped here. It said, thinking I'll build a comprehensive ECF model for. Indeed. Before I start, let me clarify a few key parameters. So this is great. And you know what, I love this. And you know, this is a theme that I keep talking about in all of these tools. For anyone who's of the view that AI tools are going to replace you or that you might want them to replace you, I would say that's probably not the right way to think about it because it hasn't done anything yet. It says it wants me to please provide your preferences. So it needs, it's looking for my guidance on the projection Horizon. You want 5 years or 10 years revenue growth? Should I use analyst consensus historical average whack components? Should I calculate current market based whack? How should I calculate the. The wac? How do you want me to do the terminal growth rate? What kind of long term growth rate? How much level of detail do you want? Quarterly or annual? Right. A sensitive. So it's asking a lot of. You need to know how to answer those questions and please, please provide your preference. Or can I proceed with the standard defaults? So standard defaults, five year annual projection market based WAC, 2.5% trim growth. Okay. And a sensitivity analysis. So market. Let's say, let's just say for now please proceed with standard defaults. Let's see what happens here. Please proceed with standard defaults and come back reasonable.

I mean, right. You could question some of them, but there's nothing there that I look at those defaul and say, what were you thinking?

No, but what I'm curious to know is I was doing. Right. There's a lot. I mean I used to, you know, in my training days I used to love teaching valuation and DCF valuation. There's a lot of nuance to doing a dcf. There's a lot to understand. Are you using unlevered cash flows or levered cash flows? Are you using a. How are you thinking about the terminal year, the terminal period? How are you taking the terminal value? Are you using some sort of, you know, a perpetuity calculation or some, some sort of a multiple. How are you thinking about the, the long term change of working capital? It's. So there's a lot to think about. What is it?

These images, I'm guessing these are links to the different websites it's gone to.

What?

Sorry? You see all the images after each of the things?

Yeah, yeah, I can see that. I don't, I don't. I can click on one. I don't know what.

Yeah, click on one. Let's see what happens.

Well, it's taking me to, it's taking me to something called finbox.com and it's wanting to verify that.

So whatever thinking it does, it allows you to see all the websites it searched.

Okay, so it's searching helpful. Yeah, yeah, yeah, it's definitely looking at.

Although that's a lot of websites, it makes you a little nervous. Why is it looking at so many for each question?

Yeah, yeah, it's going to a whole bunch of different sites. This is the Financial Times. I can see this here. Like that's Investopedia while it's still thinking. And again, this is interesting. I tried this again earlier and it provided me a response very, very quickly. In this instance, it's this particular one it's looking for what are Nvidia's current total debt, cash and cash equivalents, shares outstanding. Here we go. We're finally getting something here. Beta and betas as of fiscal 2025. So it went to a whole bunch of sites to kind of, you know, arrive at. I mean that's fascinating to watch that, what it did here. So what did it, what did it give me? Says it's an Nvidia DC evaluation model. All figures in dollars as of November 21st as today. Key assumptions. It's using, you know, revenue growth, operating margin, tax rate. It's got a bunch of assumptions here. It's got the WAC calculation. So arrived at a risk free rate. It says 10 year treasuries it's got an equity beta 2.1 which seems like a reasonable beta for a stock like this type of volatility. I'm not sure how it derived a wacc a beta. I could ask it. It's using a 5 1/2% market risk premium. Come up with a cost of equity of 15% and a cost. Oh, here we go. Oh, look at that formatting on the fly. Real time. It tells me that the pre tax cost of debt is only 3%. That's a little interesting. So again, as a, as a modeler, as a valuator, I need to think about whether this makes sense. The first thing jumping out at me is it's wanting to use a cost of debt, a long term cost of debt at 3%. Paul, do you think it's reasonable to.

Assume a long term cost of debt? So that's not. That's before tax benefit which means that you'd be like 2% when you figure the tax. Yeah.

Do you think it's reasonable to assume a long term cost of debt pre tax? It's lower than the risk free rate.

That's what I was just saying. I'm like okay, if that was post tax, even then it's low.

Yeah.

But at least.

Yeah, let's see what it. So it arrived at a. It arrived. And again the keyboard shortcuts are different. I discovered pressing. Most people know F2. F2 doesn't seem to work, but F2 and then the up arrow allows me to see what's going on. It's got a long formula in here. It's got. Oh, what is it doing for the weighted average cost of a capital. So it is. Oh, it's calculating. That's interesting. It's doing it differently. And a couple of funny things going on here. So it's showing me the total value of the company on a percentage basis. It's adding up.

It looks like that's a mistake that it shouldn't be percentage, it should be dot.

Why it's percentage in dollars. Now this company has very, very little debt. This company has very little debt. So of course the vast majority of total value will be the bottom line.

It's really almost immaterial to the WAC because that's so little debt. But 3% is still the wrong number.

But what. Let's see here. Now it's done. Do I want to keep these changes? Let's see what it's done. It tells us. It says complete. I built a comprehensive DCF model for Nvidia using The fair value per share is $48.55 and equity value. Now it tells me where is the current. The stock price is way above that, I believe. And again, I just find it interesting that yesterday it did a number of things differently and it, and it showed me some different information. So as an example, and I'm not going to spend too much more time on it, it's doing something here. You better believe that I'm going to have to dive into understanding everything that it's doing. Because first of all, yesterday it was telling me what was the. If I want to keep the changes. Yes, let's ask it. So it tells me that the fair value per share is 4855. Let's just ask it. What's the current stock price of this company? Because yesterday it was telling me the premium or discount to the stock price. And it's. I don't know why it needs searched.

It's 181.77. Let's see what it comes back with.

Yeah, I, I wanted it to tell me because.

Yeah. So roughly. Close that 180. Okay.

So it's telling me, it's telling me this means the stock is trading at a significant premium, 272% above the DCF model's fair value of 4,855. This is what I wanted to get to. It's generated DCF for Nvidia telling me that the fair value per share is $48. I hope to heck that no investment banking analyst or any other analyst goes out there, runs a DCF on Nvidia using a tool like this and gets a value of $48 and comes back and allows the MD to say, what are you talking about? The stock price is $180 per share. So it is not reasonable to derive a DCF that generates a stock price that's one quarter of the current market price. You know, I mean, as I said, I've been teaching and I used to teach DCF valuation all the time. There are only two reasons. There are only two reasons why a DCF model might generate an answer that's different than the current stock price of, of a public company. Right. There's only two reasons why the DCF value might be different. Number one is because your dcf, you and your DCF are making significantly different assumptions than the market's assumptions. So you. So either our assumptions are much, much more. Significantly more conservative and conservative and cautious, which is a possibility. So that's the first possible reason or the second reason. Is because you made a mistake, because there's an error in the calculations. Now the default ECF is so significantly below the current fair market price of the company that you better believe you're going to have to understand exactly what's going on. And is there a. Is it possible. So it's very possible that your assumptions are just a lot more conservative. But it's also very possible that there is a calculation error somewhere. And in fact, in fact, what I would think, and I noticed this the other day, so it's derived a weighted average cost of capital here. It derived a weighted average cost of capital. Again, I want to crawl through that and see if that is doing that correctly here. It's taking the, the equity value divided it by the total, which is like 99% multiplied by the cost of equity. Okay. So it's basically, it's basically there's no debt. So this part may as well get. It may as well get.

You can pretty much delete that out and your answer changes I think by like nothing.

But the point is this, the point is, is. Yeah, I think is this, is that, is it reasonable? Right? Is it reasonable to assume, I mean again, in the world of a DCF when we want to value a company, is Nvidia going to have exponential extraordinary growth forever? No. Is it going to stay in all equity finance business with no debt forever? No. And if you are building a dcf, the purpose is to think about a long term sustainable value of a company. So is it reasonable to assume almost no debt in the long term? Right? It's assuming that there is not going to be any debt in the long term of this business. Now I'd have to crawl through to understand what is. But I need to understand, you know, thinking. Yes, it's reasonable for ids. Here's why. Net debt is already. Well, that's just telling me what it is. Tech sector norm. Yeah, I mean, it's true. So it's telling me in the tech sector companies do have minimal debt when they get large and maybe that is a possibility. That's fine. So why do you think, I mean, I could ask it. Why does it think, why do you think think the DCF value is so much lower than the current stock price? Right. I'm curious to know why, why does it think the DCF value is so much lower too?

Because if you look at it, it assumed a 25% revenue growth rate. It assumed the same operating income. The WAC doesn't look totally crazy. The terminal doesn't. So I would expect it to be quite a bit closer could still be off. Right.

So it thinks, well, it thinks that the whack is too high. Now let's see here. That's what. So it's interesting to me that it actually has an idea. So if the whack is too high. Well, it's the one that calculated the whack. Right. So why.

But it's saying, look, we use their historical volatility but given their current dominant position, maybe we should have assumed there's going to be less volatility. Okay, maybe.

Right. The point is I need to be able to have a, I need to be able to have a conversation with it and understand what it's doing.

It's using.

You know, I don't know how it arrived at the equity beta. Of course, when you do a forward looking dcf, all DCF are forward looking. The beta should reflect future thoughts on volatility, not historical observed beta. So it might need to come down the market risk premium. It seems reasonable the cost. So again it's, it's got a cost of equity here at 15.7%. I, yeah, I don't see anything blatantly wrong as it looks like.

I thought it was interesting the other the number two revenue growth is too conservative. It grew last year at 114% and they're assuming immediately it goes down to 25.

Yep, that's right. So it's.

But also they're assuming for the next six years it's 25, 25% growth rate year over year as that number grows. That's really hard to do for five straight years even for a fast growing public company.

Maybe it is. What did it do for the terminal value? For terminal value it's used a perpetuity calculation here. Maybe it's right. Maybe the stock is so overly priced like a lot of people think it is and it's going to, you know, drop down to a more, more reasonable level. That's a possibility. I guess I'm, you know, continue to believe that, that we're going to have to. Anyone doing evaluation like this is going to have to understand. So I'm impressed. I'm impressed. I love the fact that it's given me a good start on trying to value this company. But, but I really believe that we're going to have to understand it extremely well to know whether it makes sense. I mean is it using, is it using a long term capital structure which is reasonable, which is what you would want to do in a dcf. Is it using. Is the cost of debt reasonable in the wac is the cost of equity reasonable. So we have to be able to go through, ask these questions and challenge it. And that's going to take some time as well. But I certainly like what it is.

It looks like all the formulas are relatively simple. The numbers look right in the sense. I don't see glaring mistakes here.

No.

This is really one where I could take this and go back and forth with the AI and change exactly how I'm comfortable. Like I would be comfortable using this for a basic DCF as a starting place.

This is that little helper on my shoulder that I've wanted. Right. This is the helper that says get something going. And it's actually giving me some good suggestions on the terminal value. Growth rate may be too low. It's telling me that maybe our revenue forecast is too conservative. So. Right. I mean, ultimately, I'm the one that's going to be responsible for this dcf and I'm going to have to go back to my boss, my board, my management team, my client and defend it and explain it. And, and so listen, I like this. Generated this and I like that it is giving us some thoughts on what we might do to. It would be very unusual to present a DCF with a value that's so radically different than the stock price because the stock price is simply the market's reflection of value, which is thousands and thousands of investors all doing their own dcs.

Right.

And so if every investor thought that this was right, then I promise, if every single investor thought that this was right, then of course the stock wouldn't be at $180 because it would trade.

Where the stock would be relatively close to $50.

Exactly. If that's what the world thought. So why don't we move on? This was an interesting test. I'm actually quite pleased with, with the starting point that it's given me. And let's go into this. Let's go take a look then at. Let's go actually see. Let's do this. Let's actually go see how it can build a financial model. Let's go see that one. Let's go see.

Sounds good.

And maybe we'll.

Yeah, I'm really impressed with the dcf. I thought it did a good job.

I thought that as a starting point.

Yeah.

It's not for the faint of heart and it's not for someone who's brand new to valuation. Right. You're going to have to under. And you're going to have to challenge what it did in terms of its choice of every single assumption. But it clearly was able to do that. Now here's the simple reality as well. Given the amount of time I think it's going to take for us to crawl through, I don't actually know how much faster that was than. I mean most of the time in that DCF is going to come from massaging it, tweaking it, playing with the assumptions, checking it. Right. I mean literally all it did is it took some historicals, built a very simple forecast using a revenue growth. So I don't know that that saved me. But if I wanted to get a back of the envelope DCF on 25 companies and I said do the same thing on 25 sheets for 25 companies, it would be a great starting point and give me a chance to play with it. Let's take a look at this.

Go out and look at, you know, the top 50 companies have grown the most over the last month and bring back that DCF to look and see. Hey, do I think there's one over undervalued and get you started? Yeah, kind of like a screener in some ways. But it's built something you can play with.

Very impressive in that regard. For sure. For sure. So let's see if. So this is just the three years of historical finances and I'm not sure why I'm not seeing here. I'm not.

You'll need to expand it out. If you expand out column C for some reason it expanded out D and.

Not C. Well, it's being a little bit odd. Even as I toggle around.

There'S some work that needs to be done when it brings in an Excel file. Yeah, like that's an opportunity.

There's no reason why it shouldn't be showing me the company name, but that's okay. So I'm going to say build a five year forecast model on the model sheet, which is this one for the years 2025, 2029. Build separate schedules for revenue costs, etc, make reasonable assumptions. And this is what I've been asking every other tool to do. Let's see, you know, how it performs. Let's see how the truffle pig tool performs. It's thinking and let's see what it does.

Definitely thinking. Should we pause it and give it a few minutes to work?

Yeah, let's just give another second. You know the tool last week.

Oh, it may ask you a bunch of questions similar to the others too.

Please provide right away.

Use reasonable defaults.

Right. I'm not, I'm seeing not the Same, you know, granular flexibility to slide left and right. I mean, as I move to the right, I'm very quickly flying way off to the side, so makes a little bit more awkward to work with, but that's okay. So is it asking me some questions here?

Yeah, so it's asking you a bunch of questions about revenue, cost and saying, hey, do you want me, do you want to tell me the assumptions or do you want me to use default?

Yeah, it says I need clarity. I need clarity. It says. Right. It wants clarity. So it's got some really impressive. What it's done is says revenue growth for 202529 showed a 13 decline.

But why, how it's saying 2024 showed a decline at 13.4%. Is this an anomaly or trend?

So at 2324 or showed a.

What should I do going forward? It's worded a little weird there.

The forecast for 2024 had a bit of a decline. Right. Is this an anomaly or trend? Should cost remain as a. So I like this asking for clarification, which means I need to know what to ask for. It wants to know. It wants to know what are the annual capex amount, what depreciation method should we use? So it's. It, I like that. It's asking for clarification.

I like working capital. Hey, do you want me to take a simple percentage or should I do.

Historical well, or let me use reasonable defaults. Okay. Please, please. I'm going to say please use a revenue growth rate of 5% per year and make reasonable assumptions for everything else.

Right.

How about that? Let's see what that does. So because I don't want it to project a 13% decline in revenue every year, then we'll quickly get down to zero. But let's see. So it's. I wonder how much it's going to be comfortable trying versus okay, it's so now. Okay, so let's see. Why don't we pause? It's got zero. It's figured out that it needs to do 11 steps. Let's put this on pause and see. We'll wrap up and show people what it comes back with.

All right, it's finished running. It's run for probably seven minutes or so. And we're going to go ahead and take a look live here and see how it performs. So it's asking us if we want to make the changes and in take it away. This is your baby. You're our modeler.

Sure, I'll do that. It is. Yeah, it's. It's so I haven't accepted anything yet. It took about. Yeah about 8ish minutes or so.

7, 8 minutes. Something like that working.

Right. It slowly went through and crossed off its 11 point to do list. It created and then it said it was thinking and then it said done. And what it says is I've successfully built a comprehensive five year model. On the model sheet with seven supporting schedules. It said I did assumptions five year annual growth which is what I asked it costs. It said assumed that cost of goods sold 75% of net revenue. All right. Capex it's made an assumption for Capex depreciation debt. It assumed an annual repayment. So this is pretty good. It's made some good assumptions here. I like this. 30 effective rate dividends 20 payout ratio. This is pretty wild so far what it's doing. I'm not gonna lie. And then net income is going and then it's got some highlights. It's got some investment. Some highlights here. Net income is going to go from 2.2 up 26. Eita is growing. Debt's coming down. Just giving me a nice summary.

It says balance sheet remains balanced throughout. Let's see if we're really good. But that's nice.

Yeah, let's see. The balance sheet remains balanced. So I'm fascinated.

See I don't know if that's a forecast highlight. I would call that standard of your work. Right?

Yeah, I would.

Oh exactly.

Now the. It's put in all of these pieces here and it's put a bunch of schedules at the bottom. I am going to say yes, let's accept. So I'm going to accept and I guess that means it's just here now. Now again. Yep.

Yeah. Yeah you can just There now.

Bit awkward. Tiny bit awkward that I can't easily see all my columns formatting.

You have to expand them. That's annoying that it doesn't.

Yeah.

And you didn't get a label.

The labels don't go into the blank cell. So I have. I would have to really make one column wider to see another column here. What I could do. We talked about what I could do is I could download this into an Excel file.

Let's go ahead and see what it looks like. I could leave it here and it gives us a download separate. So you can just open the download.

Sure. So let's. And you can choose. And this is probably why we talked about. I think we talked about this. Right. Why they've built it as a separate tool is it allows you to. Allows you to Download it into either Google Sheets or Excel, which is nice. Let's export it. Now, this might require me to change my screen share because I was. Because, yeah, it is. I'm going to have to change my screen share quickly.

We're going to pause here for a minute as he gets up, brings up the Excel file. We'll be back with you in one second. All right, we're back. We've opened up the work it's done in Excel. Let's go ahead and take a look and see how it did.

So just downloaded the straight into Excel. It's linking to schedules, which is nice. It's obviously built a schedule down below. We'll go take a look at it. This is the first time seeing of this here and you know, obviously the formatting is not ideal. I would love it, but that's not a big deal at all.

And we probably could have asked it to format and got a little better with some problems. Absolutely.

It's got. Let me just, you know, deal with some little consistency here, of course, decimals, so that I can at least read the numbers and notice how much faster.

You'Re working now that you're in the tool you're used to.

Yeah, that's funny. No surprise. Now, the net income, interestingly, is not coming from where I would expect it to. And this is what we found. Some of the other tools I've struggled with is understanding some of the account when you ask it to build free form. Sometimes they struggle with the accounting ideas and the accounting system. So net income, most people now it's the right number. I'm not sure where it's grabbing it.

From, but linked to the schedule is my guess. Or where. No, it wouldn't be the schedule because net income to comment, doesn't come from a schedule, per se.

Net income should just be right off of the income statement.

Exactly. Yeah. I was just gonna say.

I'm like, so let's see what it did on that front. But I do like the fact that everything is just straight links so far. So we're gonna see what's happening.

Why did it go. One second. Why did operating cash flow go so negative? Can you scroll back up? Not sure.

We're gonna have to come back to. Well, because we can come back, but I just noticed a massive.

It's a working. Okay.

A monstrous. A monstrous investment in working capital every.

Year from zero to that. That. That is definitely.

That's not likely. That's not likely that it's going to be a. That it's going to have A enormous change in working capital. Let's see, though, what is going on?

The balance sheet doesn't balance. Despite it saying it balances throughout. It's out by about 50 million almost.

To working capital does. Obviously, the historical balance sheet balances here. Right, we can see that. Let's carry it across. Right, drum roll, please. And of course it doesn't. So it's off. Even though you're. As you just.

If you go up, it's almost exactly off by. I know it's not just that, but it's almost exactly off by the working capital.

Well, that could be a big part of it. And again, the point is, I'm sure.

There'S more than that, but it's just interesting that that's the rough number.

And. Yeah, in every episode we've been saying. So now the key is to understand what's going on, what's it doing. It's built some schedules and.

Wow.

This is not. This would not be recommendation for best practice on a. On any formula to have an enormous change. And I have to check and see if this working calculation is actually working correctly. But what it did is. It is. Right, so it's. Okay, so it's forecasting the receivables based on days. It's forecasting. So I can tell you just instinctively by looking at this, right, if I look here, this is for 2025. I can tell you that if receivables is going down, that should be a slight inflow of cash. Right. Inventory is about flat, payables are around flat and the prepaid expenses are approximately flat. Everything is about flat. So how is it possible that if everything is flat, the change is so negative? That's an error. It makes no sense. And so there's a huge, huge error, obviously, in what it's doing when it arrives. The change in working capital. And is this number actually.

So. Yeah, I mean, it has some.

I like that. It built it and it linked it. This is better than some other tools are done. Right. It built a working capital schedule and it linked it in here. This is the correct flow. It's the right linkage, but that we'll notice in 24.

It also. It tried to show it's 55 million, but the real thing is zero. So it's formula and each. Why is the formula, if you go to G, G177 is zero, which is what it shows up above. Oh, it just linked that one. And then it tried to do its own formula and messed up.

Yeah, there's no previous period here, but let's take a look at this one here. So we can all agree, right, the change in working capital. If I just make these numbers again, this is the, this is where we're seeing, you know, a lot of challenge in these tools. They don't seem to really understand the purpose of these ideas in these concepts. So what I would say is now first of all, this working capital schedule is not optimal here. But what I would do is I would start by adding up the. Let me just add a couple rows here. Let's see what the actual working capital is. We do have our assets, right? We have our total assets, receivables, inventory payables, our assets, prepaid expense assets and other current assets. There are payables, prepaids payables, accounts payable.

Should be a liability, right?

Where, of course, yeah, that I, okay.

I was like, they're going, we said asset. I'm like, you must be right.

And I'm like, wait, yeah, no, no, I don't like it. They had that order. So where are the assets? The assets are my receivables, my inventories and my prepaids, of course. Right.

And other current ass and then other liabilities here.

So I don't love that. It's, that's. It threw me off because it doesn't have them in the right order. And then of course we're going to subtract out the payables, the accounts payable and the other current liability. So this is my networking capital. Would you agree with that?

I agree.

That's my net working capital and is just the assets minus the liabilities. And then the change. Right. Should of course be the last year minus the current year. Right. It should be in this case. So this is the change in working capital that should be flowing into the cash flow statements. Not. Well, I don't even know why this, this is the, this is the net. This is. I'm going to call this the networking capital. This is my networking capital here. This is correct. Assets minus liabilities. The change should be the difference year over year, last year, minus this year. Because if, if the working capital is going down, cash is coming up. So cash is coming up here. So, gosh, we. It appears, well, it appears it's got the sign wrong. It took the negative of this minus that which ended up. Look what this little piece of calculation does, and this is just a little bit concerning by taking. I'm just going to put these two in brackets. So this is what it did here is it took the difference in the receivables Right, Agree with that. And if I do an F9, just that in itself because it did A minus. Minus. A minus. Right, A minus. A minus. It flipped them, used its mathematical operations incorrectly. It's ended up with negative 55. So that's a big error.

So that.

Let's do this. Let's link the cash flow statement to row 180 here and see what that does. So that. But again, you really have to understand modeling and how these calculations should work. So really that should have been. I'm going to copy that over. So now we've linked it correctly to the change in working capital. Let's see now that should be fixed. Let's see now if that did anything to get. Look at that. That was the only thing. Paul, the balance sheet balances now. Everything else is. That was the only error here. Otherwise I got to pay.

Found that itself. We could go back in real quick and ask it to see if we could find why it didn't balance in the AI we wanted. But.

Oh, sure, I have to go back to the AI. Otherwise I have to tell you, I'm. It's doing a really fine job here. Right. Look at. Let's look at the depreciation PPE schedule. Beginning balance plus capex minus depreciation. Excellent job on that. The cost schedule, the revenue schedule. I'm not going to crawl through them all because we're already probably, you know, at time here. But other than that one error in the working capital schedule, it feels like it's done some really nice work. And other than the formatting and the presentation of it. Let's just see here. Beginning retained earnings. It took. It added in net.

What did it. Did it do a revolver?

Why is it calculating net income as. Oh, did net income as J153 minus J155? It is. Oh, interestingly. Okay, that's interesting. That's interesting. It's calculating net income on the equity schedule, even though net income already exists on the income statement above. But it appears to have gotten the.

Right answer and that's why it was linking wrong above. Remember net income? You're like, why is it linked down to the bottom?

Yeah, it's coming down to here. So it calculated net income. So listen, overall I think that this is done a really nice job. I think with a whole bunch of additional prompts we could get it, we could clean it up. We could pull out the assumptions, we could refine the assumptions, we could add a summary page. We could probably add scenarios. It's not taken us zero time because We've had to find it. We had to troubleshoot an error which we had to understand. We had to. And again, would everybody would any quickly realize that this was the obvious answer problem?

I'm not sure.

I mean to me I was, you know, I was able to hone in on that quickly because it seemed like a huge. Because the number was so huge. But you.

You have enough experience, you could quickly look at it. Would everybody have that?

Overall though, I think it's done a very, very impressive job and I'd be thrilled to use it as a starting point to. To get a model built. So I think you can stop kind.

Of my take what I think what I find really interesting. And this gets back to if you remember, they said, hey, we may not demo as well. You know, demoing, showing the Excel Esports, that's kind of sexy. We can solve that. But that's not real work.

Yeah.

What we've done here is a little bit more of the real work and it has performed better. I think what they said is true.

Yeah, no, I think I agree with you. 100. This is, you know, outperformed probably virtually every other tool. You know, I'm noticing here things like it's showing negative cash, which nobody would ever want to show on a balance sheet. So we'd have to understand why was the revolver. So it obviously it didn't know to build a revolver on the debt schedule. So again, these are. These are modeling concepts and accounting concepts. Probably you need to re upload this, go back to the prompt and say, hey, we're showing native cash. So can you build in a revolver, give it some instructions so that the revolver starts to kick in and as a result we have, you know, zero cash. But its ability to build schedules and mechanically link them all together was great. The only blatant error that we found was the way it treated working capital. Overall though, I give this a very, very high grade and it's done a really, really nice job. So I agree with you.

You know, it linked a few things like why did netting come down at the bottom? So there's the revolver. There's some quirks in here. There's some things you need to work out. But on the whole, if you know what you're doing, this is a tool. I could feel comfortable having to do some stuff and then me massage it. It could save me some real time. So I.

This would empower. This would empower a good analyst. Could get better. But you have to know how to challenge it, push it, find the issues right. And like we keep saying, you know, every episode. So obviously your boss would obviously eventually want to see that net income is coming right off of the bottom of the income statement. That's more standard convention. But again, if you understand that, then you can do that and clean it up once you get it to work. Same thing like I did here with the working capital. But yeah, I'll turn it back to you. Then I'll stop my screen share because.

I think that yeah, we're, we're kind of to wrap up point. So why don't we stop. We'll stop your screen share and just final thoughts before we go. I mean, I very impressed with Truffle Pig on the whole. I was a little nervous at first. Like I said, it struggled on the esport cases, but on the work tasks I give it pretty close to an A. Not quite there. But on the whole it might have been the best one we've seen. It's right up there for sure.

100%. Definitely up there. Didn't again, maybe it's just getting used to it. They don't love working in a separate platform because that was. But I guess you have to get used to working in their own app and then their own spreadsheet and then downloading to Excel. I guess if it's very iterative, if you're going to do that, you're going to work on something, download to Excel and probably make some changes, tweaks, formatting, improvements, re uploaded if you want to make more changes. So I don't know if that would be a bit of a pain to be downloading and re uploading.

There is a little bit of pain there. So you got a little bit of pain with just learning a new spreadsheet. If you're going to go back and forth, I love the fact that you can link out to all the websites and it shows them all. That was really nice. I love the way it showed changes that are very easy to see. So impressed on the whole, I mean, something in beta it's going to get better. If you're wanting a third party spreadsheet AI tool, definitely one I'd consider. No question.

100%. Yeah. No, I think it's a great job and I give kudos to them for developing this tool. It's certainly the best one that's been able to integrate the financial statements and actually do it correct from sort of an accounting standpoint.

Yeah. So great job, Truffle Pig. We'll be back again testing another tool next week. We're not done. We have several more to go and a few other exciting things we're thinking about, so please stay tuned with us. We'd love to hear from you. If you've watched an episode and you enjoyed it, reach out to us on LinkedIn or send us an email. We always love to hear from an audience. And thanks for joining us. So, thank you and it was a pleasure. Giles will be back with us next week. Or in. Will involve his dog even more if Giles doesn't.

Yeah, yeah. Benny might be back if Giles wants it. Needs to keep skipping off here, but we'll see. Anyway, thanks, everyone.

All right. Thank you. The mouthwash. We are the My Squad.