Ex-JPM Banker Tests Claude (AI) With Building a DCF in Excel

Hey everyone, it's Ben here and today I'm going to be testing out Claude to see how effective it is with building out a DCF. Now, if you've been following

my channel and videos for the past few years, you probably have seen that I've built a ton of models in the past manually using Excel. And I have played

around a little bit with Clawude Excel.

And I h but I haven't really fully built out a DCF model like the one we're going to build today. And so this is I thought it'd be kind of cool to not just like try building everything on myself and

just showing you guys like step by step how to build it or build or show you what I did because I will be making Excel AI kind of related videos in the future. But I thought this could be like

future. But I thought this could be like a fun little cool memorialish memorabilia almostish type of video where later down in the future I could

look back at this video and see how much Excel or AI has progressed in the future almost like I'm first discovering the iPhone or something like that because I

wanted to kind of give you guys first initial real reactions to how good or bad this technology is at the very beginning.

So, I'm going to basically just kind of show you guys or go through like what I um have planned here, which is first I'm going to give you guys an intro to

Claude in Excel. Build a Google DCF. So,

previously I built out this DCF. This

was the most recent one I built in a few months ago in January 23, 2026.

And then we're going to use the same assumptions and then have Claude build out the DCF. And we're going to compare and see basically what is better or like how I would kind of

use the DCF as an investment banker on the job. Kind of that's what where we'll

the job. Kind of that's what where we'll discuss kind of the pros and cons. But

going to build out the DCF using Claude and then compare with the old DCF, see what values it brings out and all of that and then discuss some pros and cons. Now this is just for like fun

cons. Now this is just for like fun basically not financial advice.

And that's the plan for today. So if you guys have not yet used cloud before or have an idea of how to even like use it at all.

So you first need to just download Oops.

I guess you don't don't need to see that, but whatever. Um, so you you first need to download Claude as an Excel plugin. So if you Google something like

plugin. So if you Google something like Claude Excel for Microsoft or for Apple, then you can download it. And then you just basically

download it. And then you just basically install it and it pops up here just like how you use any other chat. So it looks like chat GPT or like claude chat etc.

And then you just start telling it here like what to do basically. So that's

what we're going to do. And

what I did is I brought in the historical financials and um future estimates that I used for my

previous Google DCF. And then I want to use the same figures to build out the DCF today with Claude. And then we'll

compare the two as I mentioned and see basically what happens. And before I go into actually building out the DCF or asking CL or actually why don't we why

don't we first do this. So I'm going to go in here and with so I think about this as like an investment banker of how I would use this in the future. Let's

say a client or yeah let's say a client gave me some historical financials or estimates and or estimates like management projections of this is what we see where where we see our business

going. It has a lot of key line items as

going. It has a lot of key line items as you see here. You know, capital expenditures, free cash flow. Actually,

we wouldn't use that, but capex, DNA, then we have some historical figures, um, estimates as well for like sales, you know, your income statement, all of that. And then what I would do as a

that. And then what I would do as a banker is basically go into this kind of chat box thing. Um, and then tell it

using these estimates and historical information, build out a DCF for me. So,

that's what we're going to do. And we're

going to see basically how accurate it is. And then what I would do after

is. And then what I would do after everything is built out, of course, is make sure that everything's correct. So,

double-checking everything. That's where

I think like the next generation is going to be very interesting to see where there's going to probably be a divide where a lot of bankers such as uh

basically from like those who started 2023 and be maybe 2024 and before maybe 2025 and earlier they built out all of their

models by themselves in Excel, right?

Checking every building every single cell. But then for those who are joining banking let's say like this year onwards all these new Excel tools have come along and so

increasingly I think people are bankers are going to be less uh they're going to not be as used to probably building models manually and so if they ever do

need to check their models increasingly that skill is probably going to erode a little bit over time but we'll we'll see but that's just my guess. Um, so what

I'm going to first prompt it is over here. I'm going to say, um, build me a

here. I'm going to say, um, build me a D. So I've included in the

D. So I've included in the in I've included historical and

estimated projection or projections for Google Alphalphabet

in the tabs in the document in the Excel right now.

build me a five-year DCF model um using these assumptions or using these figures

and assumptions as needed. Pull in any extra in

as needed. Pull in any extra in information that you need from your own sources

to build out the rest of the DCF.

And then once you do this, if you guys have used it before, like I've tested it around I tested it just a little bit.

It'll start like telling you what it's going to be doing and then it'll start like prompting like asking you like should I do this or not? Like here's my plan, etc., etc. It's kind of cool like

you can basically see what it's thinking and all of that. Um, and you know, this is kind of the cool thing, right? Like

you can kind of like just have it do something and as you're waiting for it, you can do something else. So, for

example, I need to tell you guys about the sponsor for this video, which is Wall Street Prep, right? And as it's just building this model, I could just tell you guys the sponsor and then go

back to it and then the entire DC like a lot of the DCF should be done or it's going to prompt me and ask me a few questions. But we'll go back through it

questions. But we'll go back through it back to it. But, um, so Wall Street Prep and Columbia Business School Executive Education have an 8week online course for AI for business and finance. And so

if you're trying to get smarter on finance and or smarter smarter in how to use AI for finance, this is a 8week program where you will learn a lot of

the fundamentals of how AI is being used in finance. So I want to go through the

in finance. So I want to go through the actual curriculum itself. So foundations

kind of just like an overview of everything. Um then you go over here,

everything. Um then you go over here, predictive analytics and strategic insights, advanced predictive analytics, simulations and optim optimization, generative AI, advanced AI applications

and integration, AI tools for business and finance. And throughout it all, you

and finance. And throughout it all, you also learn um how to code a little bit with Python. I there will be some areas where

Python. I there will be some areas where you'll learn just enough Python, no previous coding or tech experience needed. So pretty cool. you'll learn

needed. So pretty cool. you'll learn

actual like application with theory and then the speakers are Columbia Business School executive education professors and there are a lot of really great guest speakers throughout the entire

program um working at a lot of the top AI firms. So Endex for example is an Excel AI company that only serves institutional investors. Perplexity I'm

institutional investors. Perplexity I'm sure you guys have heard of City Bank, you know Raymond James, all these like financial firms as well. Open AAI. Um so

you guys can feel free to check this out. I have a link to it in my

out. I have a link to it in my description and you can use my code rare liquid to get up to $500 off. Um, so

yeah, if you want like a really professional kind of way to learn about how to use AI in finance, then feel free to check out this program. I think it's

a pretty cool and cool one where there isn't a lot of programs like this where you have a top tier business school and u Wall Street Prep obviously is well known for financial modeling. So check

that out in the description if you'd like. Going back to the model that's

like. Going back to the model that's being built. Let's kind of go through a

being built. Let's kind of go through a little bit of the uh actual thing that's being built here. And I'm going to I realize that the bar is not being fully

shown. So let me

shown. So let me adjust my settings a little bit so you guys can see everything.

Not sure if I can really like zoom into this part on the right hand side.

Hopefully, it's not too hard to read, but right now it's saying here just to kind of go over it says, "Let me go to uh get the current stock price, like gathering some information,

um getting the balance sheet figures, have enough data." So, it's telling me what it's pulling from my existing sheets, all this stuff here, and then what it's pulling from it's the SEC

filings, and then the plan that it has for the DCF. So, build a new DCF sheet.

Create a key assumptions section. Whack

inputs. Section two, whack calculation.

Third is unlever free cash flow projections terminal value. It's going

to use both perpetrator growth method and the exit multiple method and enterprise value to equity value bridge sensitivity table. So I'm going to first

sensitivity table. So I'm going to first tell it yes and then kind of discuss with you guys a bit like what I like about this which is first of all you can see that it's fairly

specified for a financial model right like if I were to think about when before I started working in finance if I needed to learn how DCF was built and it gave me these instructions at the

beginning like I can actually learn how to build a DCF just from from reading this and kind of going through it versus like in the past you actually have to learn by learning the concepts for

through like the 400 questions or something like that and then building things out. Um and then it let's see

things out. Um and then it let's see next here it has like a plan. I'm just

going to allow it. Um so yeah, it's pretty much now just telling us what it's doing and then we're going to now just have to wait and see it build the model. But it kind of it's kind of cool.

model. But it kind of it's kind of cool.

It's like magic. I'm not going to have to build anything myself and it's just going to start building out as you guys will see.

Um, let me just also as we're waiting for it to build out, see what questions you guys have.

So, let's see. Thoughts on data accuracy

let's see. Thoughts on data accuracy when leveraging cloud AI and Excel? So,

that's what we'll kind of take a look at together after it's built out the model. Will

this be available to watch after the live? Yes. Have I tried switching to

live? Yes. Have I tried switching to Sonnet? Is there a big difference with

Sonnet? Is there a big difference with Opus 4.6?

I have not tried that. No. Um, should I drop out because it doesn't look like they're going to hire me to do any of this if Claude can do it. So, I actually just spoke with like an Evercore banker and I spoke with several other bankers

over the past few weeks and asking them like, you know, is headcount going down?

How is AI affecting your current job?

What's the outlook like? And there will always be a need for analysts and associates like junior bankers in general. You're never going to have a

general. You're never going to have a time where VPs or EDS and MDs are going to want to even prompt Excel to build a DCF and whatnot. And even if they are

able to do that, they're probably going to, you know, think about it. They're

they're senior bankers, their role is to maintain relationships with clients and run deal processes. And so that's what their focus is. It's not really

building out models themselves, even if it's super easy to prompt. And as you guys might see, like there's a lot of kind of issues that will probably still

pop up as we go through this model like formatting and eventually putting this into like a PowerPoint page which will be automated as well. So can I see a future where there will be less bank

like junior bankers and a reduction in headcount? I do think so. That's

headcount? I do think so. That's

definitely definitely possible. But I

think just as how with sales and trading for example, there was a lot of automation and there was a reduction in headcount, there was still there still are people actually working in sales and

trading because you still need people to do things, right? So um that's kind of like my view overall. Of course, we'll see what happens.

But I think also the important thing is if you really upskill yourself while you're, you know, before going into banking, learning how to use all these tools, you'll actually be able to be

ahead of a lot of other people, right? I

think there's like a very very very small people of pe small number of people actually using Excel AI tools right now.

Um, okay. So that was let's see next one. This is crazy. I'm starting my

one. This is crazy. I'm starting my analyst job at a BB and this is scary.

Uh, one other thing that I think is worth mentioning actually is that my MD would always talk about how when they when bankers used to get financial

figures, they would have to go to the public library and print out all of the financial statements and all of that, bring it back to the office and build out models by hand. That was, you know,

of course, like 50 years ago or something like that. But now you're able to download all the financial figures from databases like Bloomberg, Cap IQ,

Faxet, and you're able to instantly get all those numbers. But even with that automation that literally made your life a thousand times easy, like faster with

building models with the with Excel and with all those databases I mentioned, you still have bankers working like 80, 100 hours a week. And so I think even

even even though you're going to be a lot more efficient with all of this stuff, I do think that it's just going to result in bankers having to do a lot more work,

junior bankers, I mean, and just a lot more volume. Like even for my business

more volume. Like even for my business right now, even though we have AI and so many things are automating a lot of things, like I'm I'm still working a lot actually. And then uh a lot of my

actually. And then uh a lot of my teammates still we're kind of just all expected to do a lot more versus um just kind of like being able to automate everything and not have to

worry.

So yeah, we'll see. Overall, this looks like you know pretty good for a first just prompt. If you guys are just

just prompt. If you guys are just joining, you know, I prompted this with just like two sentences. I asked it.

I've included historical projector uh figures and projections for Google/Aphab in the tabs in Excel right now. Build me

a 5-year DCF model using these figures and assumptions as needed. Pull in any extra information that you need from your own sources to build out the rest of the DCF.

And then we kind of went through a little bit about what exactly its plan was and then now it's kind of just building everything out. Says it has the layout. So, it's kind of cool. It tells

layout. So, it's kind of cool. It tells

you exactly what it's doing. um showing

you DCF projection table. Yes, it's

going like basically step by step and telling you what's being built, perpetrator growth method give 175 at base. So it's telling me a little bit

base. So it's telling me a little bit about valuation.

Okay. So yeah, it gives you like a summary of everything. Um I guess for our purposes though, it'll just be good to basically double check everything now.

All right. So, let's

over here. I'm going to move this DCF over here. And I discussed at the

over here. And I discussed at the beginning, I gave you guys a brief intro to Claude in Excel. We built out the Google DCF with Claude. And now I'm going to kind of go through it and make sure that all the assumptions like make

sense and whatnot. And then I want to compare it with my old DCF that I built uh about about two months ago. And this

we're using the same historicals and financials. So, let's it'll be

financials. So, let's it'll be interesting to see what AI or what Claude said. Um, so let's see. Risk-free

Claude said. Um, so let's see. Risk-free

rate, it's using a 10-year Treasury yield, 4.3% as of March 2026. So, let's go to 10-year Treasury yield.

NBC today is 4.283. So, that looks accurate.

Equity risk premium, Damodorian implied equity risk premium 4.6%.

So, Damma Dorne, he's like this um NYU professor who just publishes this all the time out of honestly like the goodness of his heart.

Uh let's see where this is. So,

implied equity risk premium 4.38% 4.6%. So, this looks

4.6%. So, this looks a little bit off.

So, I'm going to I'm going to adjust this. Well, I'm just going to have a

this. Well, I'm just going to have a note here. This should be 4.38%. So,

note here. This should be 4.38%. So,

yeah, seems just a little bit off, but not not crazy. Levered beta.

Let's go to Alphabet because we're looking at Alphabet right now. Oops.

This number is probably going to be different just because I have access to faxet and then claude

is probably using like Google or Yahoo for beta.

Where is beta?

Um beta beta Okay, right here, beta 1.05. So that's

pretty close. Oh, that's actually exactly right or exactly the same. So at

cost of equity, it gives you the formula here, which formula looks correct.

Pre-tax cost of debt. This is alphabet weighted average coupon on outstanding notes of 3.5%.

Let's see what we used last time. I

actually did like a quick and dirty method, so it's a little different. I

use 4.07%.

This 3.5% is probably pretty accurate.

Average coupon on outstanding notes, 3.5%. Okay. Marginal tax rate of 17%.

3.5%. Okay. Marginal tax rate of 17%.

It used its effective tax rate from 2024 rounded to 17%.

We used a tax rate of 16.89. So that's

pretty pretty close. Posted that looks good.

debt over total capital. I'm not sure why. Yeah, this is like a hard code

why. Yeah, this is like a hard code which I wouldn't do personally. So, let

me just um let me add like basically a column here

where I would say like what what I would change or let's call it like errors.

Maybe I need to zoom in a little bit.

So modeling always it's not exactly uh sometimes you know it's an art not just a skill. So

a skill. So not necessarily purely an error but some well some things are errors like this I think is an error

but okay so this 2% I would not hardcode.

So what I'm going to do maybe I'll do like kind of want to explain what it is so that it makes sense. So 4.38% is the

current equity risk premium and then that this is hardcoded which is a big mistake and an error in a

model.

Okay, I wouldn't say big mistake, but if you hard code a lot, then it's not good.

So, what we could do in instead what you want to do is kind of find the amount of total debt and then find your debt over total capital and your equity over total

capital versus like using this as like a 2%. So what I would do is

2%. So what I would do is what I would do is find the total debt and equity value and calculate

the total that 2% um equity figures. But we can actually have claude

figures. But we can actually have claude adjust this. So maybe we'll like check

adjust this. So maybe we'll like check uh have it adjust all the errors in the future or as a next step um for whack.

Let's see if this makes sense. your

percentage of uh B15, your percentage of equity times your cost of equity plus your B14, which is your percentage of debt

times your cost of debt. So that looks good. Terminal value assumptions says 3%

good. Terminal value assumptions says 3% 3% is always kind of high. I would say 2.5% is more conservative slash

accurate.

Keep it the multiple. This is going to be hard for us to Well, actually, let's see. Did I do comps last time? I didn't.

see. Did I do comps last time? I didn't.

Okay, so this one we we'd have to actually do the comps ourselves to see if it makes sense. But 18 times might be

okay. And now net debt. These are just

okay. And now net debt. These are just Okay. Gave us figures.

Okay. Gave us figures.

All right. What I don't like about this is that it gave us figures from 2024.

Now, if we were to go now, I'm sure like the good thing is that it pulled from an accurate figure, but you want to look at your latest

look at latest 10Q or 10K, but plot fold from uh 10K 2024. it probably is like

prioritizing 10Ks because it's like the longer full annual report and the 2025 10K hasn't come out yet. Or if it has, that's probably like a much bigger mistake. So, let's check that.

mistake. So, let's check that.

Oh, okay. That's even crazier. So, the

2025 figures have come out already, but for some reason, Claude pulled from the 10K 2024.

So yeah, that's kind of odd honestly. And

then the diluted shares outstanding again is pulling from 2024. Not ideal.

So maybe maybe we can ask it to pull from the 2025 10K. We can try doing that later.

And then let's go into these figures here. So let's see first of all if

here. So let's see first of all if they're pulling from the right information.

2024 looks good. 2025 estimates. So even

though right now it's 2026 and 2025 figures have come out, we are assuming that we're modeling from this DCF that I

built previously on January 23rd, 2026.

And on that date, you can see here that the 2025 figures came out on February 5th. So that's why

2025 right now is an estimate.

So, these look like they're correct, but what it's probably doing is just straight pulling from these figures, which is what I told it to do. So,

that's okay. Revenue growth.

Yeah, this is But this is weird because if you've been following my models at all previously in the past, what I don't do is use these figures here because they're not reliable.

Um, 2030. Yeah, but it's okay. This is what

2030. Yeah, but it's okay. This is what I told it to do. So, it's doing everything correctly. EBIT. Let's see if

everything correctly. EBIT. Let's see if this is correct. And this is exactly what you would do on the job as well, right? You build out with Excel and then

right? You build out with Excel and then you basically tell it what it's you see if anything's wrong, but all of this looks correct. Taxes on EBIT

looks correct. Taxes on EBIT multiplying by the marginal tax rate.

I don't know why it's m marginal tax rate not effective, but notepad is ebit times one minus tax rate, which is fine.

DNA is pulling from the estimates that I gave it. Capex

gave it. Capex is pulling from this the estimates I gave it as well. That's good. And then

change in networking capital. It's just

using this two point two 2%.

Assumption is unclear where it pulled from this like 2%. We can compare with what I used, which may be similar, but for me, I used around like a what did I

use? I I used an average. So, it's not

use? I I used an average. So, it's not like super far off or different, but still.

Um, and for capex, it's kind of just using the, let's see, it's just using the estimates I gave,

but it's not actually color coding correctly because green means that you pulled from another sheet,

but it's not green here. Um, and I know I'm being nitpicky with that, but I'm just trying to figure out basically like everything that I would have to change if I was actually on the job. So, now we

have unovered free cash flow, which looks correct here.

Um, we have your notepad plus your DNA minus your capex minus. Oh, wait. Oh,

wait. Oh, yeah. It's negatives already, so it's just adding them, which is fine.

Unlever cash flow growth. All right.

And then DCF valuation.

uh looks like what 0.5 1.5 so it's kind of just like brute forth brute force using discount period or the midyear convention but what I would like to see

is it incorporate like we did in our model midyear convention of seeing how much is left from the current date to the end of the year but

I guess I didn't give it the date so I can't fault it too much for that um but then it calculates the discount factor and and is figuring out

present value of free cash flow is okay.

Sums the free cash flow and then terminal year unlevered free cash flow.

This is correct using uh okay let's see if this formula is correct which should be terminal value last year free cash flow times 1 plus B20 which should be your terminal growth

rate divided by your whack minus your terminal growth rate which is good present value of terminal value you divide by 1 plus your terminal one plus

your whack to the power of your last projected period which is good exit multiple it's getting your EBIT from here. See if that matches with your

from here. See if that matches with your 391.

You have your DNA and your EBIT here 352. So that's

something that's definitely I would say is a big error as well. You have to be using the same EBIT across your DCF and

your um exit multiple.

Ebbit thought is inconsistent with ECF.

Uh so now you multiply this to your uh exit multiple which is the 17 or 18 that we had up here

and then you discount it. So very very different figures enterprise value using your terminal uh yeah perpetrator growth

method and exit exit multiple method just add them together then you have your net debt to your equity value diluted shares outstanding and then implied share price. So I'm

going to just add a little nit that color coding needs to be purple. This is

I didn't tell it to do this so it's not entirely incorrect there. Um, this

formatting is a bit weird. This left

align formatting change to right align.

And then we have the implied upside and down or and or downside.

So that's pretty much what it came out to. So implied share price 175 to 400

to. So implied share price 175 to 400 dep depending on the method. What we had previously was implied share price of 202

using the perpetrity growth method. So

it looks like this DCF is if we compare just purely with the perpetr method which is what we used. a little bit more conservative and

used. a little bit more conservative and having built the model I know it's most likely because of the capex spend which it just pulled from our estimates

purely versus um what we did was we actually grew it down over time because right now Google is spending way like a ton on

capex right now because of AI but then we assume that it would actually go down over time as a percentage of sales versus like if you look here 149 on year

2030 we have 128 right and this is like $128 billion and so like if you this all adds up a lot over time basically

so maybe one thing I can do is I mean we can kind of see how smart it is maybe um I'm going to be a little bit more

specific make it green so this assumption is clear go where 2% was or let's see where what this comment says

okay estimated networking change of capital is 2% of revenue consistent with alphabets as light model okay so I guess it does give it some

backup but we need kind of like backup information need backup figures what we can do is the

if it can take everything from column column K and then make the adjustments needed. Uh by the way sensitivity

needed. Uh by the way sensitivity analysis this is like using like for formulas to create it. So this is weird like I would actually use the data table

functionality within Excel.

um data table uses formulas instead of table

if what if analysis in built in Excel.

So, let's see if I'm going to make this one a copy and just call it um control like like a control group

basically and then we'll make adjustments to this. So, I'm going to prompt it now and say, "I've added errors I found

in column K.

Make revisions to the DCF based on the errors I found. If you have any questions

or need me to clarify how to fix an error, ask me. That's like a tip that I've come across when prompting

flawed AI ex um chat, etc. is to ask it if it has any questions if if it has any questions to ask you.

So, let's see what happens here. And as it as I do that, I will um answer any other questions you guys have. Um so, what percentage of your first year junior

analyst class at JPM do you think came from non-targets?

Uh I don't know where everyone went to school. Like my group was really small

school. Like my group was really small because I was in the SF group. So, one

was from UCLA, I was from Berkeley, one was from Morton Wharton.

Um, and one was from a non-target actually. I do know that.

actually. I do know that.

But we had only when I got there like I was one of two analysts starting for the SF Healthcare group.

Uh, all right. Then

there don't seem seem to be too many other questions actually. So now let's Okay. Okay. So on the bottom right here

Okay. Okay. So on the bottom right here we it says okay IDA is inconsistent with DCF. The terminal value section

DCF. The terminal value section currently uses the 30 IDA from income statements estimates while the perpetrator growth method uses a terminal value. Uh how would you like me

terminal value. Uh how would you like me to fix this? 853. What did we put here?

EBIT dot is inconsistent with DCF. So

okay cool. So this is what I wanted to do, right? I wanted to use the EBIT dot

do, right? I wanted to use the EBIT dot plus DNA from the DCF so that everything ties. That's good. So K70

ties. That's good. So K70

says these data tables.

So what if analysis requires specific cell references and manual setup that I can't fully automate?

So, I'm just going to leave it as formula based and then it should kind of build out everything.

So, let's see what else we're going to do. So, I've kind of already done two

do. So, I've kind of already done two and three a little bit and then we can discuss some pros and cons and then we're it's kind of we're kind of done with it, right? And I think that's it's definitely like if you have been

following my models, you know, like I build out a lot of the DCFS completely from scratch just to show you guys how to build it. And it takes usually like two hours to be able to like build it

all. But with this, like I could build

all. But with this, like I could build out the model and I would have to actually make a lot of adjustments to the formatting on all that stuff on my own to make sure that it it's aligned

with how I normally like things. But

actually, if I I'm sure over time, you can actually have it for probably now already have it just format in the way that you like things being formatted.

But I'll do that kind of in the future.

Um, I just wanted to this to be like my first test with using Claude for building out DCF. And in the future, I'll build out more videos about how to

or kind of like a tutorial for how to use like a lot of these tools because I'm looking a lot into them these days.

Um, and also going to like make videos about like testing different Excel AI tools and like kind of grading them and seeing like which one creates the best VCFs or financial models by all kind of

using the same prompts.

So yeah, stay tuned for those types of videos in the future.

Um, if you guys also I'm always curious to know like if you guys have any AI tools you guys are using right now that you really really like. Let me know in the chat because I other ones I've heard

I feel now I heard nowadays everyone is kind of using voice to text. Um, so you like press like a command shift or something on your

on your keyboard and then you just start talking to your computer and then you can basically, you know, you talk a lot faster than you type. So it just saves you a lot of

type. So it just saves you a lot of time.

Um, okay. So I see someone said, I think you're also using it rather inefficiently. I think there are more

inefficiently. I think there are more cohesive ways to go about creating Excel models code.

Maybe I mean this is just like a test but I'm not too I'm not proficient with cloud code. Um

cloud code. Um I don't know this was pretty fast so I don't know like if you have to like build out something totally custom then I don't know if it's like really the

best. Um,

best. Um, but maybe I just, you know, I don't know. But I don't know if it'll be

know. But I don't know if it'll be interesting to see maybe if bankers in the future are going to need to learn how to use cloud code in order to build Excel models. That'll be interesting.

Excel models. That'll be interesting.

But I think also like there are probably going to be better and better tools in the future. Um,

the future. Um, and that'll help make things a lot more efficient as well.

Uh, I see someone says using Whisper.

I've definitely heard of Whisper. Any

advice on how to prep for hitting the desk? Want to make sure you can hit the

desk? Want to make sure you can hit the ground running? I would honestly play

ground running? I would honestly play around a lot with the Excel AI capabilities and have it build models for you. Just double check the models

for you. Just double check the models and make sure you know how to build everything from scratch as well. I think

that's super super important.

All right, but let's see. It seems like Oh, maybe I built it out here.

4.38%. Okay, cool. So, it made this change or this hardcoded figure and now see previously it was hardcoded

at 2%. And then now I I told it

at 2%. And then now I I told it basically like I want the actual figures. I think it's still

figures. I think it's still pulling. Oh, as of March 2026. Oh, so it

pulling. Oh, as of March 2026. Oh, so it even actually started pulling from more recent figures. Although the 10 Q3

recent figures. Although the 10 Q3 I don't know why I pulled from the 10 Q instead of the 10K 2025.

Um but now at least it fixed this first thing which is like uh showing me how exactly you're calculating your debt over equity or debt over total capital.

And that's closer to what we had previously when we built the built the model 0.54%.

Um, okay. But versus like in the past here it has 2%, right? And every little like this might not seem like a big difference, but when you're building when you're doing this for an actual client, you need every single number to

be perfect. The 2.5% is more

be perfect. The 2.5% is more conservative accurate. So, it fixed that

conservative accurate. So, it fixed that as well. And I told it to look at the

as well. And I told it to look at the latest 10Q or 10K. And it's pulling pulling from Q3 2025. Obviously, as I mentioned, there's the 2025 10K. So I

would actually have to go in and fix these numbers.

So it didn't it still didn't really fix that. Maybe I should have been more

that. Maybe I should have been more specific. Tell it to use 2025

specific. Tell it to use 2025 told it to make this color coding green which it fixed. So the assumption is unclear. I need backup figures. And so

unclear. I need backup figures. And so

what it did here, what did it do? So, we got the previous year sales minus your current sales and then it just used the 2% as an

assumption and it's basically just saying okay it's using historicals and it used an average so I guess that's fine. I

thought was inconsistent with DCF but now it's consistent.

All encoding needs to be purple since it's from the same sheet and then formatting change to right align which it did. So I think that's actually

it did. So I think that's actually pretty it was pretty good at find at correcting most errors but then as you can see there was like a few things in terms of

like data sources that was pretty inaccurate and not great.

So let me now go into kind of like the pros and cons that I see. Um because I started off by telling you guys a bit about if I were to use this in banking

you know like how exactly would I use it. I think first of all there are other

it. I think first of all there are other companies that I've come across and spoken with myself where they actually have a template that you give the

company a lot of your templates and then the program or the model or um whatever tool you decide to use will actually build out the model in your format

versus this is kind of just like a stock format that it used, right? And then a lot of the I would say like the pro is obviously that it built this out like super super

fast. Now if I were to try to make it a

fast. Now if I were to try to make it a lot more complicated by having like 50 plus assumptions or um making it kind of

more similar to the standard DCFS that I have here where I have like different cases and stuff like that. I'm sure it could do that for me as well. Um, but I guess like you just have to be a bit

more specific with all that stuff.

Cons is that it seems like the data source is a little bit unreliable. We

could try maybe I can try asking like pull from the latest 10K and update all relevant figures.

Maybe that'll fix it. But I would imagine that I would want my whatever I'm using to automatically pull from the latest. It's kind of it's kind of dumb

latest. It's kind of it's kind of dumb honestly that it would pull like a 2024 financial metric or Yeah. Yeah. 2024

financial metric when it's already 2026.

So the fact that I had to prompt that was kind of a waste of time. But

um yeah, and then a lot of like little things like it it mess it's a little bit more inconsistent than I thought it would or could be. Like you just expect

like if it can do a lot of these things correct like harder things correctly, why can't it, for example, have like color coding correct throughout the entire model?

Um, the way it built out some of the figures with like the assumptions and whatnot, I feel like previously it was just like multiplying by like a random 2%, but you

would never want even if that's the way you want to do it, you want your 2% assumption to be here, not like embedded inside the actual figure here, like it says times 2%.

So, um, that's kind of like my initial take. you know, later in the future,

take. you know, later in the future, I'll definitely plan a much more kind of thorough like building out maybe some of the parts DCFS or different types of financial models, but I wanted to just

this to be like a very quick uh intro to like building out a DCF with Claude. Um,

it seems like I also used up all of my tokens already when and I barely prompted it. So, I

guess that's like another huge limitation. you work at like an actual

limitation. you work at like an actual company, you probably can, you know, have like a limitless amount of capacity for for using this stuff. But anyway, um

let me know if you guys have any other requests of what kind of like AI XL or AI uh finance type videos you guys want me to make in the future down in the

comments below and I'll definitely take a look at all of them. I really read read like almost every comment when a video is published within the first few days. So, um, yeah, feel free to do that

days. So, um, yeah, feel free to do that and then hope to catch you all in the next video and stream.