Why Spreadsheets Will Stay, but Financial Modeling Workflow and Version Control Must Be Rebuilt with Matt Lee

In this episode of Financial Modeler's Corner, host Paul Barnhurst is joined by Matthew Lee, Co-Founder and CEO of Nummo, a software tool built to help analysts and bankers build financial models more efficiently. Matt explains why he started the company after years in investment banking and as a CFO, and what he sees as the real opportunities and limits of automation in financial modeling. The conversation covers the pressure of dealing with work, common modeling mistakes, and where tools like Nummo can help speed things up without replacing analysts. Matt also shares thoughts on what makes a good model and why hands-on understanding still matters most.

Matt is the Co-Founder and CEO of Nummo, an AI co-pilot for financial modeling that he wished he had during his time in finance. Before Nummo, Matt worked as an investment banker at Lazard, helping to execute sell-side and buy-side M&A deals totaling around $6 billion in enterprise value. He also served as CFO at a digital product studio, overseeing financial planning for emerging tech products launched by Fortune 500 firms and early-stage ventures. With over 40,000 hours in Excel, Matt built Nummo to deliver a banker-grade modeling platform that runs fully on-device, maintains cell-level accuracy, and automates workflows like DCFs, LBOs, comps, and fairness opinions.

Expect to Learn

• Why financial models still need human input

• Where tools like Nummo can improve modeling workflows

• Why version control and hidden hard codes cause major issues

• How to balance automation with accuracy and narrative

• Why it’s critical to never outsource your understanding of a model

Here are a few quotes from the episode:

• We’re about 30% of the way there. AI can help, but it can’t yet build a fully detailed, nuanced model.” - Matthew Lee

• “Even with automation, you still need to check for hidden hard codes and make sure your model tells the right narrative.” - Matthew Lee

• “Formatting, standardization, and QA are where AI is already adding value today.”- Matthew Lee

Follow Matthew:
LinkedIn - https://www.linkedin.com/in/matthew-m-lee/

Company - https://www.nummo.xyz/

Follow Paul:
Website - https://www.thefpandaguy.com  
LinkedIn - https://www.linkedin.com/in/thefpandaguy
YouTube - https://www.youtube.com/@thefpandaguy

Follow Financial Modeler's Corner:  

LinkedIn Page- https://www.linkedin.com/company/financial-modeler-s-corner/?viewAsMember=true  

Newsletter - Subscribe on LinkedIn -https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7079020077076905984

Sign up for the Advanced Financial Modeler Accreditation Today and receive 15% off by using the special show code ‘Podcast’.

Visit https://bit.ly/3AK5dYo and use the code “Podcast” to save 15% when you register. 

In today’s episode:
[02:28] - Matt’s background
[03:54] - Worst modeling mistakes
[06:24] - What Nummo is and how it works
[11:10] - Biggest Challenge in M&A Transactions
[15:34] - Building models that tell a story
[28:23] - How workflows are changing
[31:53] - Favourite Excel Shortcut
[39:44] - Rapid fire: Excel habits and preferences
[45:35] - Final Question & Episode

Full Show Transcript

[00:01:37] Host: Paul Barnhurst: Welcome to Financial Modeler’s Corner. I am your host, Paul Barnhurst, aka The FP&A Guy. And this is a podcast where we talk all about the art and science of financial modeling with distinguished financial modelers from around the globe. The Financial Modelers Corner podcast is brought to you by the Financial Modeling Institute. FMI offers the most respected accreditations in financial modeling. It's used by universities. Many of the big accounting firms are now starting to, uh, have their staff take it. And it's the best accreditation out there. And it's why I completed the Advanced Financial Modeler. Encourage you to do the same. This week, I'm thrilled to welcome Matt Lee to the show, Matt Lee. Matt, welcome to the show.

[00:02:25] Guest: Matt Lee: Thanks for having me, Paul. Excited to be here.

[00:02:27] Host: Paul Barnhurst: Yeah, really excited to have you. So let me give a little bit of background about Matt and then we'll jump into our episode here.Matthew Lee is the co-founder and CEO of Nummo, an AI co-pilot for financial modeling that he wished he had. Before Nummo, Matt was an investment banker at Lazard, where he helped execute sell-side and buy-side M&A totaling roughly $6bn in enterprise value. He also served as the CFO of a digital product development studio, leading the financial planning for emerging tech products spun up by Fortune 500 enterprises and early-stage ventures. Having spent 40,000+ hours in Excel over the last 6 years, he started Nummo to deliver a banker-grade modeling tool that runs fully on-device, preserves cell-level fidelity, and automates heavy workflows like DCFs, LBOs, comps, and fairness opinions. He’s passionate about democratizing structured thinking - bringing the craft of high-stakes modeling to more operators with speed, accuracy, and auditability. So welcome again to the show and love the background.

[00:03:47] Guest: Matt Lee: Awesome. Yeah. No. Ready to dive into anything we want to discuss today? I'm super passionate about financial modeling so ready to go.

[00:03:54] Host: Paul Barnhurst: Perfect. We always like to start with that horror story. Tell me your horror story when it comes to modeling. I'm sure you have a horrible one you've seen or built or worked with.

[00:04:03] Guest: Matt Lee: So yes, as a short answer, I think it kind of relates to where I learned the most. I will say, back when I was a banker, I did touch a lot of fairness opinions, like 100 plus tab models, more than ten scenarios that you can build base upside case, downside case, multiple in between as well across all the variables. You could imagine only 20 days to get it done. So pretty much pulled an all nighter every other day. But I think I learned the most in those high stakes scenarios. And I really think that's where I worked across aerospace and defense, but also TMT and all modeling is different, right? But really being able to be in that demanding setting, creating one of the most detail oriented deliverables, you'll never get that learning experience again.

[00:04:50] Host: Paul Barnhurst: It sounds like a wonderful learning experience. I learned a lot from those. Did you ever have a model that you inherited or you picked up where you're like, what were they thinking? Whoever built this needs a little bit of help, so to speak.

[00:05:04] Guest: Matt Lee: Yes. And I think, uh, it's where you can find hidden breaks when somebody passes it to you. So, for example, like some figures might be hard coded in random places where you thought there was a formula, someone tried to hide it, and then you obviously kind of need to rebuild it in that regard. But it's definitely happened before. I'm sure you've gotten that too, Paul.

[00:05:23] Host: Paul Barnhurst: Oh, haven't we all got hard codes? And the worst is when you're like, wait, why would there be a hard code here? And you know nothing about it? Why is it there. And yet, like you mentioned, you often have to start over.

[00:05:34] Guest: Matt Lee: Exactly, exactly.

[00:05:36] Host: Paul Barnhurst: Right. So earlier this year, you stepped out on your own and started Nummo  and you've been the CEO there. So first tell our audience a little bit more about what Nummo  is and then what the experience has been like for you.

[00:05:49] Guest: Matt Lee: Absolutely. As you mentioned, Nummo  is really that tool that I wish I had day to day as a banker and as a CFO. Think of when it hits 3 a.m. at night and you get hit up for some please fixes from your managing director from your boss. It's really intended for that purpose. Um, at its core, Nummo  is an AI native app that can guide you through building out financial models more efficiently, more accurately. And we're starting with sell side M&A use cases. So like custom operating models, DCFS, Lbos comps, table pulls as well.

[00:06:24] Host: Paul Barnhurst: And so I'm curious what was the thinking with Nummo  to create your own app versus maybe doing some kind of add-in to Excel or your existing spreadsheet program? What was that? Because I'm sure you had discussions of how to do this as you were going through it, so I'd love to know the thinking.

[00:06:41] Guest: Matt Lee: I think it comes a little bit on matching user behavior, but also usability when you're appealing to financial institutions. And so, for example, when I was at Lazard, I wasn't even able to access things like Google Drive on the company's VPN. So web apps for me were something that we didn't want to venture into. And the same thing with plugins, frankly, like I would use Factset's plugin a lot, but it would slow down my modeling. And I like to think that when it comes into the world of business, the world of finance, the most technical you can get almost relating to an engineer would be working in spreadsheets. And it's such a flow state activity that I didn't want to ever disrupt. Someone who's ready to sit at their screen for ten hours, complete a model like build in every nuance that they wanted to the critical thinking. And I never wanted to disrupt that. Or even worst case scenario, crash their Excel.

[00:07:33] Host: Paul Barnhurst: Wait. Excel crashes?

[00:07:36] Guest: Matt Lee: Honestly, yeah. It's so sad though, because people lose their progress. And I think that that is the worst case. Right. And I never want someone to have to do that or experience that from Numa. No.

[00:07:47] Host: Paul Barnhurst: Okay. Yeah. No, I, you know, totally tongue in cheek. I think anyone who's used Excel for a long time has definitely lost data. How to rebuild something. I know I've had it happen more times than I care to admit at this point. After, you know, 20 plus years of using it.

[00:08:04] Guest: Matt Lee: 100%, 100%.

[00:08:05] Host: Paul Barnhurst: I'm curious, how do you expect the average modeler to use it? So are you expecting them to do some of the building in Excel or walk me through kind of how a modeler would use Nummo  in their day to day work, how you envision it working?

[00:08:20] Guest: Matt Lee: Yeah. So it's really designed to just slot directly into the workflow or the financial modeling process you're already used to. So you can drop in files, say, PDFs of filings, private company reports, research, market reports, even that just get automatically baked into your spreadsheet. And you can even set your own formatting specs. So the models that you're creating will always look the way that you need them to be. Also, by default it should be quote unquote client ready. And ultimately, we want people to have the freedom to either finish things the way they traditionally do with, say, like the alt control, PC shortcuts, manual edits, and also just leverage AI in the most efficient way with like query based changes that impact spreadsheets. Overall, though, I would say the end goal is never to replace the modeler here. It's to make them sharper, faster at their jobs, a little more trusted, so it can turn the parts of modeling that used to drain you into parts that would now elevate you.

[00:09:22] Host: Paul Barnhurst: So is the idea. If I'm starting a model, I go to Nummo , give it all the details, and it gives me something in a spreadsheet to work with, and I can take that back. Or how do I think about that? You know, from just a practical I'm, I'm building my, you know, comparables model or DCF or whatever. Do I go to Nummo  and start there, give it a bunch of information and it outputs something and then I can continue to work on it from there. What do I think about that?

[00:09:48] Guest: Matt Lee: So yes, in simpler terms you can. And that's where you can basically get all of the information that you need. I would think of it as like you're consolidating information into backup tabs that then you might want to restructure if you want to create a more complex model. And if you want to start going into different deliverables, that backup compilation is a place where you can right from there, export to Excel. And everything you see in spreadsheet interface is 1 to 1. Most of the value that you'll get is from being able to start at that maybe 20% place, and then you can kind of take it further with like query based changes, um, pulling in research because we want to be integrating FactSet cap IQ merger market, PitchBook, all types of single sources of truth, if you will. Sure. But then that's where we just want to add value and options. If you want to take it back, go for it. If you want to manually edit and in new mode, go for it too. But yeah, I would say that's kind.

[00:10:46] Host: Paul Barnhurst: If you built a spreadsheet interface in Nummo . So basically you're kind of gathering all the data, whatever the questions, presenting something in a spreadsheet type format, and then they could continue to work on it in Nummo  or they can take it back to Excel.

[00:10:59] Guest: Matt Lee: Correct.

[00:11:00] Host: Paul Barnhurst: Got it. No, that's helpful. So I'm going to step back. I have other questions I want to ask about that. But we'll jump into your early career a little bit before we get back to Nummo . So you started your career as an analyst. As you know, your bio mentioned Lazard, working in M&A. I'm curious for you what was the most challenging part of this kind of M&A transaction? What was the hardest?

[00:11:21] Guest: Matt Lee: I would say there's a lot of things that you have to do, and I'll be a little more concrete there. And I really just think it's a matter of managing your time, because when you're facilitating M&A, yes, you are kind of navigating that buyer-seller relationship depending on what type of deal you're on. So there might be multiple financial sponsors that you're interacting with, multiple strategic acquirers that you're also interacting with. And I'm not really talking about trackers and, you know, the contact logs for those who are in banking, that's not really what it is. It's more just like having to navigate while also having to create a lot of deliverables. And so that would be PowerPoints, Excel models. And overall, that's kind of where there's only so much you can do in a given day. Yes. Like you'll likely be working 100 plus or 80 to 100 plus hour workweeks. But you need to be able to navigate your time, which is always a challenge.

[00:12:14] Host: Paul Barnhurst: Yeah, no, I can imagine. So it sounds like the hardest part was just managing the time for the sheer volume of workload.

[00:12:21] Guest: Matt Lee: Correct, yeah.

[00:12:24] Host: Paul Barnhurst: What was your favorite part of working on the M&A deals? What did you like about it?

[00:12:30] Guest: Matt Lee: Little on topic here. I would say the financial models. And it's just because you can get into a flow state. And I'll say this personally, nothing makes me happier than like a 100 plus tab well structured, detail oriented, well sourced, cited model or like Excel output. Um, and it does get a little challenging, right? Like so for some of the very complex deal deliverables like fairness opinions, it does take time and you might lose sleep. For me, nothing's more rewarding than getting to that VF. The final version of your model being ready to send out, to distribute to clients, to send out to potential buyers, or to even show the rest of your deal team. Nothing makes me happier than that. So I would say modeling.

[00:13:12] Host: Paul Barnhurst: I like it. I was going to jokingly say, is there anything? Is there really anything such as a final? Right? Have you ever seen the joke you know, in Budget and Forecast? It's final. Final really final? Final version seven. You know what I'm talking about. It feels like you think you're done. And it's almost always one more thing. And I imagine you dealt with a fair amount of that in M&A, just like I did in Fpna, I would guess.

[00:13:36] Guest: Matt Lee: Yes. Updated VF v distribution I totally, totally agree. That's so funny. Yeah.

[00:13:45] Host: Paul Barnhurst: Yeah. I think universally anyone who's built models has dealt with it. Okay, how should we be naming this? And the naming convention is not being clear. And you're like okay what's really the final version. Because this says final version three dated this day. And then there's another one that says final version two that's dated a week after this one. So which one is it?

[00:14:03] Guest: Matt Lee: Yeah. That's why unless you are in the seat only you will know. Right. Which one is it?

[00:14:10] Host: Paul Barnhurst: Yeah. And. Well, and that's why naming conventions really are critical, even though we sometimes forget that as we're quickly trying to do stuff. At least I've, I've done it a few times. And so you said you love the flow state and building a model. Is there a certain type of model you like? You like building? Is it, you know, DCF more complex, the deal the better or what? Did you know what is within modeling? You mentioned, you know, big huge files with 100 tabs. So are there certain parts like is there a statement you like best, or is it just really kind of at the end being able to show, hey, here's the valuation.

[00:14:45] Guest: Matt Lee: I think a little bit of probably DCF where you can show the valuation, but the part that I like the best is when you can put your own spin on it. And that's where it comes into the operating model, the operating models, where for, you know, aerospace and defense models, it looks completely different than like a TMT model. Same thing across any sector, any business type. And that is where as a modeler, you can really showcase how good you are or like how much you've thought about the best way to structure and showcase, you know, a company's operations. And I think that is where you might get into monthly models, you might get into annual looking models. Structure is always going to be different. And you can get as detailed as you would like to be. But the more detailed you are, the better, the more complex the better. You can kind of showcase all of the considerations that went into making it. So those are my favorites.

[00:15:34] Host: Paul Barnhurst: Yeah. As you mentioned, the more detailed you get, the more complex. And there's that balance of how complex you want to go. Is it really adding value? But I like something you said there of just the joy of being able to think through it and helping present the operation. So it's almost like telling the story of, hey, this business is valued at, you know, $1 billion, but here's why and here's why. We think, you know, if you guys merge, it can be valued at 1.5 or whatever. It might be kind of really being able to lay out the story and share it in such a way that people understand it sounds like you really got a lot of joy out of that, if I'm hearing you right?

[00:16:10] Guest: Matt Lee: I do, I always think of financial models as like, what is a financial model? It's just a table. It's tabular data, right? But what it ends up showcasing is an overall narrative of how you got to a figure through, like a mix of an art and a math. And I think that's kind of why I think in tables and I love tables so much. And going back to what you said in my intro, it is like a form of structured thinking and being able to paint that picture that is auditable, that is transparent, and that is very clear and detail oriented. I think there's nothing better than that in the realm of finance.

[00:16:46] Host: Paul Barnhurst: Yeah. And so, you know, leads me to a couple questions. The reality is a lot of models that are built are not very good. You know, many models are bespoke. It's a very time consuming activity. Why do you think that still is. Why do you think it's still one of those things that most models have errors? Many are bespoke, take a lot of time and frankly, a lot of them are hard to follow. Not always, but a lot of them out there. So just kind of share some thoughts on that. I'd love to kind of get your thoughts.

[00:17:15] Guest: Matt Lee: Great question. So kind of live through this directly. But I think the truth is no two financial models are ever the same. And that's kind of why even if Paul, you know this too, like even if the output looks almost identical, each of the models carries its own nuance like sectors, mentions revenue cost drivers that are particular to that business model, maybe deals with specific assumptions structure that just reflects the right way. Somebody who's completely different than you thought about the problem there. And I really do think that that is the main basis for why, you know, modeling is so manual because you're kind of approaching a new analysis every single time everyone is thinking about a business differently. Um, and plus, like, it's really not just despite you being able to use all of the PC shortcuts and knowing all about finance, building stress testing, and doing your own QA work of your own model. It always ends up being different and currently stands pretty manual. Um, plus, I do think their models do break a lot, right? And that's kind of where it's an element of trust. Like people don't believe in. Automated systems can spit out a perfect model for you. Figures might not line up, formatting might be off, and it might miss a couple of pieces of nuance to where you might have been better served by, you know, starting it from scratch yourself instead of getting like a one shot solution. So kind of going into pneumo, that's kind of where we see a lot of opportunity. Instead of looking at it as a barrier, like we're not trying to replace modeling, we would rather create a copilot that preserves those aspects and then leverages AI to the extent possible for what it's good at.

[00:18:58] Host: Paul Barnhurst: Yeah. And so speaking of that, what are the areas you think AI let's talk about today. I mean, we know it's quickly changing. It's always getting better. But let's focus on where we're at today. Where are the areas that you think AI can help us the most with model building?

[00:19:15] Guest: Matt Lee: So I've been thinking about this every single day, as you probably know.

[00:19:18] Host: Paul Barnhurst: I would hope so, since you're running a company that's all you do.

[00:19:22] Guest: Matt Lee: If not, we'd be screwed. But, um, I think concretely, formatting is very obvious, one that it can help with when it comes to standardizing outputs, cleaning up models, making them presentation ready. Next would be converting natural language processing into spreadsheet or cell edits. I think that's like a very clean one that it can work with. And then on the pressure cooking, sensitizing the QA side of things to identify errors like narrative issues in that picture you're trying to paint. Um, and overall, that's where I think AI can be the best at helping be the reviewer, translator, and accelerator. But I'm sure one of your next questions is probably like where it falls short. And I would say it does fall short on being a calculator for now.

[00:20:11] Host: Paul Barnhurst: Yeah, the math side. So I'm going to drill in a little further on, you know, kind of where you talked about what it's good at. So I didn't hear anywhere of basically taking the prompts and building the entire model for us. Do you feel like you're at the point? We're at the point where any complex models take away some very basics. We've seen that, but anything of level of complexity that you would use in M&A or Fona, you know, in the corporate world, an experienced modeler is AI at the point where it could really build any of those through prompts and get us, you know, kind of get get a presentable model. Do you think we're there yet?

[00:20:48] Host: Paul Barnhurst: Bad financial models can lead to bad decisions or worse. If you want to learn to build better models, impress your boss and your clients, get the Advanced Financial Modeler accreditation. Podcast listeners save 15% on AFM registration. Just use the code Podcast at http://www.fminstitute.com/podcast.

[00:21:11] Guest: Matt Lee: I don't think we're fully there. The reason why is because it might give you a couple tabs, right? Of course, it could be the basis of what you need, but it won't give you that full, detailed, 100% operating model. Neurons. Um. Mhm.

[00:21:28] Host: Paul Barnhurst: Yeah. So how close do you think we can get right now? I mean do you think we're at 5060. Just kind of high level. We're not going to hold you to the exact number. But just like your thoughts.

[00:21:37] Guest: Matt Lee: I would say we're probably around like the 30% mark right now.

[00:21:42] Host: Paul Barnhurst: And I think we have a long way to go. So if I. So when you say 30% let's take an example. Let's say I build a complex model. It takes 100 hours to build. Mhm. If I use Nummo, how much time is it? 30 hours I'm saving. What do you think? Is it 50 if I'm doing a complex M&A transaction? This is going to be something with, you know, 5000 tabs whatever or a lot of scenarios, schedules, you know, relatively large transactions. How much time do you think one would save?

[00:22:12] Guest: Matt Lee: I think at the current moment you could probably save like around half. I wouldn't say it could be done in minutes for you. But the reason why is because with Nummo  and I like to look at it from what is the modeling process? Right. Like you're taking all the resources that you need. Then you'll put it into backup tabs. Kind of the use case I was mentioning before. Then you'll start structuring. We can help you with that structuring. We can help you with that extraction. But we want you to be in the driver's seat. Analyst, modeler in the loop, if you will, to be able to dictate okay. At all points. Right. I want to project in a certain way. I want to run these calculations. I want to split up the drivers. Yes, we can guide you, but I want to split up the drivers, make these particular assumptions and it will still take time, right? Don't get me wrong. And then we can help on identifying any errors along the way so we can cut savings there. But you'll still need to be in flow state to get the model to where you want it to be and where it should be.

[00:23:10] Host: Paul Barnhurst: So why do you think? And I just want your opinion. A lot of these tools are hey, we can build a whole model for you. I've seen things as far as, hey, analysts cost this much to build your DCF model. We can do it for $1.20. You know these AI tools. Why do you feel like we're hearing so much hype? Because what I'm hearing from you is the hype we're seeing in a lot of places in reality are not aligned. So I'm just curious, your thoughts, is it because, hey, if it's basic, yes, AI can do it, but that doesn't mean it's at production level ready? Or are people just, you know, is it just everybody trying to market and oversell what they can do or what do you think's going on. Because I see a lot of stuff out there that feels like it's promising a lot more than reality.

[00:23:55] Guest: Matt Lee: Yeah. So the one thing I will say first is that I think anybody who's working in the automated financial modeling space is impressive, I agree. Awesome. Yeah. And I do think that, yeah.

[00:24:08] Host: Paul Barnhurst: I don't want you to badmouth anyone. I'm not saying anything. Anyone's doing bad things. I think a lot of great work and they have to have that vision to get there. And it's amazing what many of them are accomplishing. But it feels like AI in general. And I think, you know, the modeling example is no exception. There's a lot of hype and there's reality and they're not always aligned, I guess is what I. The bottom line.

[00:24:31] Guest: Matt Lee: Yes. And that I also agree with I think like in practice if we just take investment banking as an example. You never really want to strip away that learning experience. Also like there's a ramification that happens with other people who are not touching modeling within even one deal team, right? So like if I tried to replace an analyst, the onus of who owns that work stream falls upon seniors. Senior level deal team members who don't really want to be responsible for it today, but two, they've already kind of graduated out of it. And they have other things to focus on, like winning new business. Right. Like navigating high, very high touch point relationships with, you know, everyone in a potential universe of who would be touching a new company. And that's just one aspect. And the second is also just like there's a humanistic touch that I think is super important here, where I would have to go back into a model, no matter what, to be able to make sure it's showcasing the narrative that I wanted to show, and also to make sure there are no hidden hard codes. Right. As we kind of touched on before earlier, and I just don't think we're currently ready for me to just be in a chat bot, for example, interface and say, hey, create me this 100 plus tab DCF, because I can almost guarantee right now there's going to be errors, and it would take a lot longer for me to navigate it than if I had been involved at every step of the way to get to the 100 tabs. And maybe I started with 20, for example.

[00:26:00] Host: Paul Barnhurst: Yeah. And so what I'm hearing, and you've said this earlier, but really a key tenant, key belief of yours is this is human. In the loop. Ai can assist with different areas. And that number over time will go up. But there's always going to be the human, if nothing else. You have to review it and make sure you're comfortable with it and make sure it's telling the story you want so that you can lay out the story. Even if it could do everything else via chatbot, you still are going to have that human involvement. And not saying it can do everything. We know it can't at this point, but even if it could, your view is there's still very much a human in the loop element to get the best out of this.

[00:26:42] Guest: Matt Lee: Yes, at least for the foreseeable future right now.

[00:26:46] Host: Paul Barnhurst: So yeah, ten years from now, all bets are off, I get it. Or maybe five years at the rate we're going or whatever. But where we're at now, and at least in the short term, it's very much, hey, let it help assist you, make you better, save you time versus, hey, I can kick back and let it build my model by just a few prompts or whatever.

[00:27:08] Guest: Matt Lee: Exactly. Exactly the point there, Paul.

[00:27:11] Host: Paul Barnhurst: That's one of the reasons I wanted to have you on is I feel like we have a similar philosophy. I know I've talked about it. You're a co-founder and many others in modeling that I think this is the philosophy of, hey, there's areas where it can be helpful, but I think everybody's struggling to understand where that is. What are those areas? Because and not just modeling but excel in general because things are changing so quickly. You know, by the time this comes out, it'll have been out a while. But like the copilot function, you probably saw that on LinkedIn. You know, Microsoft just released that. And that's been an uproar because some people are like, well, that's probabilistic and you can't have that with all these scenarios. So we shouldn't have that function. And others like this are a game changer. It changes everything. And I'm like, okay, the answer is somewhere in the middle.

[00:27:56] Host: Paul Barnhurst: And let's not poo poo it because we're not thinking like AI thinks because, hey, everything I've done is finance. I want every penny to tie out. Well okay. For those use cases don't use it. But that doesn't mean there aren't these 500 over here where it could be really helpful. And then I've heard other people saying, I don't need to learn Excel anymore. Let's not oversell it either.

[00:28:19] Guest: Matt Lee: Yeah.

[00:28:21] Host: Paul Barnhurst: And so I appreciate you sharing that. So in your mind, what's how far we are from it and what's the workflow of the future for modelers. Like how far do you think we're away from being where most modelers are using AI tools? And if you had to guess, what do you think that that looks like? I mean, I know too far in the future all bets are off, but in the near term, maybe give us those two.

[00:28:46] Guest: Matt Lee: So I'll start with the future of what I think it looks like. I think the future of AI powered financial modeling is never going to abandon a spreadsheet environment. It'll just be a different experience. And that's kind of like as you've seen with Nummo , right? That's the whole point, right? Like we don't, we want someone to be able to ask AI to assist them on one cell range of cells, cells on multiple tabs all at once, and that's kind of the only way to really get that level of help. Um, in terms of when, um, or like, where are we exactly in that regard? I think, like a lot of people, they view that future similarly right now, um, in terms of where it's going to fully change, I think it's more imminent than people believe of where they're. I'm not saying Excel is going to die, right. Because copilot does, to your point, help some people, but it doesn't help a lot of people who are sitting in like a high finance, deep finance seat right now, um, and being able to not replace like an XLS format, but being able to say give a lot more enhanced collaboration with like version control, being able to give a lot better or more nuanced understanding of financial technicals with the help of AI. And then overall, just getting someone to the point A to point B of where they need to be a little more seamlessly, there is an opportunity to kind of step in and overhaul that workflow in that regard.

[00:30:15] Host: Paul Barnhurst: I appreciate that. And kind of along those same lines, what should modelers be doing today to prepare themselves? Because obviously the workflow of the future is going to look different than the workflow of the past. So what would you recommend to those people who are modeling today and trying to prepare themselves? Because I think everybody's trying to figure out, what does it mean? How do I make sure I'm not left behind?

[00:30:40] Guest: Matt Lee: Yeah. So great question. I think there are a couple things that you can do. One is like the more reps you get, the better, right? And so like a lot of these tools you should try what works best for you right. And then the more familiar you are with them I think everyone even said that like ChatGPT the more you use it, the better you're going to be. And I think what's particularly important here as to why, is to build a strong foundation, like a strong foundation of Technicals recognizing what's changing in that future of, you know, financial modeling or the workflow is really about how certain repetitive and manual tasks won't be necessary anymore. And that doesn't make modelers less valuable at all. It just means that the clarity of thought, the ability to ask the right questions or the right queries, the discipline of knowing what clean assumptions look like or accurate assumptions look like are going to be even more important than ever. And so if you're able to kind of focus on those fundamentals, that's where you'll really be able to thrive in a workflow where AI takes care of the repetitive tasks that don't need to exist.

[00:31:47] Host: Paul Barnhurst: I think the big thing is to really focus on the fundamentals.

[00:31:50] Guest: Matt Lee: Yeah, I would say so. I would say.

[00:31:52] Host: Paul Barnhurst: All right, so now I want to move into a more standard section where we ask, uh, guests the same basic questions here. So the first one is my favorite Excel shortcut.

[00:32:05] Guest: Matt Lee: Hands down alt W w.n, where you can open up a new window of that same file. Concretely, you can work in two tabs at once in the same file.

[00:32:16] Host: Paul Barnhurst: Or a big. Yeah, I think you're the first one who's given that answer. I know exactly what you're talking about. The new window. I was using it the other day, but fun, I like that. So you're obviously often wanted to work in two, two different screens. Same tab.

[00:32:30] Guest: Matt Lee: Correct. And I think it's like I would always tell that to my friends too, because they would say, okay, like you must know all the shortcuts. What's your favorite? I tell it to other bankers, tell it to other consultants. They'll get kudos from their bosses, which is nice. Um, but I think that principle in particular is where there's so many things going on in a spreadsheet. Right? And like being able to have two screens where even with pneumo, you're actually able to have your exported Excel file open, pneumo open. You can run a query in pneumo and we can dynamically update in that Excel file, um, 1 to 1 update on disk. And that kind of philosophy, I think is super important where there are moments, even in your own head, even just you in the driver's seat where you want to do two things at once, and it's important to be able to do so.

[00:33:17] Host: Paul Barnhurst:  While my background is in fact, I am also passionate about financial modeling. Like many financial modelers, I was self-taught. Then I discovered the Financial Modeling Institute, the organization that offers the Advanced Financial Modeler program. I am a proud holder of the AFM. Preparing for the AFM exam made me a better modeler. If you want to improve your modeling skills, I recommend the AFM program podcast listeners save 15% on the AFM program. Just use Code Podcast.

Got it. What is the number one lesson you've learned during your career so far that's helped you the most?

[00:34:06] Guest: Matt Lee: That's a tough one, Paul. Um. Let's see.

[00:34:09] Host: Paul Barnhurst: That's the goal. I gotta make you think, right? I can't make this all easy. We had a fun conversation, but.

[00:34:14] Guest: Matt Lee: True. True. I would think it's to never outsource your understanding. And so I'll give a concrete example where? Back when I was an analyst and this was pretty early on. Like we caught a small error with a discount rate in our DCF. It was only, what, 30 basis points or something. But being able to identify that and you know, overall the model would still run right. But being able to understand that, okay, the decision of what we're trying to achieve, what our end up purchase price that we want to go with is going to be could ultimately be wrong based off of like that wrong, you know, 30 basis points or the the mistake 30 basis points. And so that's kind of where ever since that moment, I almost changed the way I work, where I might be a little OCD or a perfectionist in that regard, but I trace numbers exactly to the source. I want assumptions to be labeled very cleanly. I want to always gut check and ask myself, okay, how can we know very quickly if something is wrong? Um, almost like within a few hours or 24 hours, just because those types of habits would never slow me down or slow others down, it would just create an in flow state, that level of trust and velocity that you're going to need. And just tying it to pneuma really quickly, like we need to do the same thing here. That mindset needs to bake in. And you know, always having that obsessive attention to detail, which has been kind of my mode or my biggest strong suit as a top bucket analyst, it's what you need to encode in a tool like pneuma itself.

[00:35:51] Host: Paul Barnhurst: I really like the line you said never outsources your understanding. Good. Really good. All right. So this is kind of a fun question. What's the most unique or most fun thing you've built in a spreadsheet for your personal life model? Or could it be anything?

[00:36:10] Guest: Matt Lee: My co-founder always teases me for this. I'll be honest. It's, uh. So I'm a huge Pokemon card fan. I'm a huge Pokemon avid Pokemon card collector, um, TCG game. I built my own personal financial model for my own collection. Also to forecast in the most informed way that I could possibly expect. Like what future value of cards would be? Um, and it's cool because when you're really passionate about something or something you just really love, right? I like what I like. If people bash me, that's okay. Um, you can almost think so critically about where there are macro and micro drivers happening altogether simultaneously. And so like macro drivers Pokemon releases new sets every single multiple times a year, every single year. Yea, some of them have limited distribution, some of them have massive distribution. Demand is higher based off the popularity of certain cards and styles, and they release in different languages. So when it comes to the Japanese economy, also America's economy and all around the world, those are the macro drivers on the micro side. I'm going to seem a little geeky here, but you can get super nerdy with like, what type of Pokemon the cards are about or are included within a set, how old they are, the type of card, whether it's a full art card or it's not a full art card, whether it's, you know, a limited release, whether it's it's like some Pokemon are worth more than others, like the popular ones.

[00:37:37] Guest: Matt Lee: Charizard. Pikachu. Gengar. Right. Those are always going to be worth a lot. Some particular cards where it's like it matches a historic artwork piece of, you know, a Japanese style, like a certain type of artwork, like a Van Gogh Pikachu, if you will. They end up being worth a lot, and there's so much you can do with and it's surprising you can. There's so much you can do with just a hobby or like a, you know, personal passion of yours to turn it into a very well functioning, detailed financial model. And I would say that is like the most fun thing I've ever done with applying finance to a passion.

[00:38:10] Host: Paul Barnhurst: Got it. I am curious, what's not necessarily in your collection, but what's the most expensive or highest value Pokemon card out there? What's it kind of worth? Do you have an idea?

[00:38:20] Guest: Matt Lee: There's always the first edition PSA ten graded like Charizard. So the first edition. First edition base set. Charizard I believe right now it's around $250,000, but there are a few limited copies of what's called a Pikachu Illustrator card that I know have sold. Don't quote me on this, but I know it was at around a million figure. It might be more depending on the grade. Um, and so it's extremely expensive. And Pokemon's actually like a very hot market right now. Um, I would recommend you look it up after not saying you should collect, but recommend you look it up after too.

[00:38:54] Host: Paul Barnhurst: Yeah, my daughter's a huge Pokemon Pokemon fan. She has several books and different things and yeah, no, I knew they had to be high. I'm not surprised by any of those numbers. I saw one the other day. The highest sports card. It was, uh, one that had both Kobe Bryant and Michael Jordan on it, and they'd each signed it. Yeah, it just sold for, I want to say 12.9 million, I believe.

[00:39:18] Guest: Matt Lee: I think I saw the headline, Kevin O'Leary bought it. Right.

[00:39:21] Host: Paul Barnhurst: Or maybe maybe it was Kevin O'Leary. Wouldn't surprise me. I didn't. And this one I didn't see who bought it, but I saw they showed a picture of the card and it was 12 something. I think it was 12, nine that it sold for. And I'm just like.

[00:39:33] Guest: Matt Lee: Yeah, crazy, right? It is crazy. Well, Paul, I didn't know your daughter was a big Pokemon fan. I'll send you some Pokemon packs so she can rip them, huh?

[00:39:40] Host: Paul Barnhurst: Yeah, sure she would. Take that. I love it. All right, so now we're going to move into rapid fire. Um, and so how this works is you can't give me the answer. It depends. Okay. You have to pick a side. And then at the end you can elaborate on 1 or 2 of these, because I recognize all of these have nuance. But if you said it depends, all of them, we'd just be listening to a consultant for an hour. And what fun would that be?

[00:40:03] Guest: Matt Lee: No fun.

[00:40:03] Host: Paul Barnhurst: So we got to pick a side. You can plead the fifth one. If you don't want to answer, I'll give you one mulligan, as I like to call it. Okay. And we're just looking for quick answers. We'll go through all of them, and then we can elaborate at the end. Are you ready?

[00:40:16] Guest: Matt Lee: Yeah. Hit me.

[00:40:17] Host: Paul Barnhurst: All right. Circular references. Yes or no? No VBA. Yes or no?

[00:40:23] Guest: Matt Lee: No.

[00:40:24] Host: Paul Barnhurst: Horizontal or vertical. Do you prefer all on one sheet or multiple tabs?

[00:40:30] Guest: Matt Lee: Horizontal.

[00:40:31] Host: Paul Barnhurst: All right. Excel dynamic arrays in models. Yes or no?

[00:40:35] Guest: Matt Lee: No.

[00:40:37] Host: Paul Barnhurst: All right then I don't need to ask you about fully dynamic, um, external workbook links. Yes or no?

[00:40:43] Guest: Matt Lee: No.

[00:40:44] Host: Paul Barnhurst: Named ranges. Yes or no.

[00:40:47] Guest: Matt Lee: Yes.

[00:40:49] Host: Paul Barnhurst: Okay. When you were building models or when you build models, do you follow formal standards like one of the boards out there, like fast or Smart or any of those?

[00:40:57] Guest: Matt Lee: Better rule of thumb to do that? Yes.

[00:40:59] Host: Paul Barnhurst: Okay. Should financial modelers learn Python in Excel?

[00:41:03] Guest: Matt Lee: No. No need.

[00:41:05] Host: Paul Barnhurst: What about Power Query?

[00:41:07] Guest: Matt Lee: Also no power BI. Also no need.

[00:41:12] Host: Paul Barnhurst: All right. Will will Microsoft Excel ever die?

[00:41:17] Guest: Matt Lee: I need to elaborate on this one. I don't think it'll die, but I think it could be replaced. And I'll elaborate later.

[00:41:22] Host: Paul Barnhurst: All right. We'll come back to that one. Okay. And I think we have the answer on this one, but we've talked quite about it the whole episode. Will AI build the models for us in the future?

[00:41:32] Guest: Matt Lee: Potentially. Depends how far out. But yeah potentially.

[00:41:36] Host: Paul Barnhurst: Yeah. So it could one day. All right. Do you believe financial models are the number one corporate decision making tool?

[00:41:44] Guest: Matt Lee: Yes.

[00:41:45] Host: Paul Barnhurst: Okay. And what is your lookup function of choice? What's your favorite lookup function?

[00:41:51] Guest: Matt Lee: I'm an old guy. Index. Match index. Match. Match 100%.

[00:41:54] Host: Paul Barnhurst: All right. Not. Not surprised. How about an index match?

[00:41:58] Guest: Matt Lee: Yeah, I just mainly use indexes. Index.

[00:42:01] Host: Paul Barnhurst: I've started adopting X match. It's been really hard for me when I do index matches. But yeah, because yeah, you get in a routine. All right. You wanted to elaborate? I know you wanted to elaborate on Will Excel ever die. So let's start there.

[00:42:13] Guest: Matt Lee: Yeah. So I think about this similar to, you know, like Figma and Adobe or when it comes to design, right. Like PDFs, PNGs are still the output of Figma. But from a collaboration standpoint, they overhauled the process of design. And I think the same case goes for something like working spreadsheets, where if there are very painful moments or there's a lot more you can revamp in that regard, maybe everything will still export to XLS. But, you know, even with what we're building here at Nummo , there is an opportunity for a new entrant to redefine that process right when it comes to version control. A huge issue touches anybody, not just in the banker seat, private equity seat or the donor seat touches anybody who's working in spreadsheets. Version controls a problem even with LLM interactions giving a little more nuance, providing self-learning guidance. I don't believe Excel does the best at that. I love Excel, but so many people are afraid to touch it, right? And I think that's kind of where it does fall short in those regards. And I think something like pneumo could ultimately replace Excel as that core workflow environment for spreadsheets.

[00:43:24] Host: Paul Barnhurst: Okay. And then the next question I want to ask you is where. So we finished through all those and actually let me step back. I'm going to cut that. You know as we've gone through that list. Any others you want to elaborate from that rapid fire list? Is there anything else you wanted to share any more context on?

[00:43:40] Guest: Matt Lee: I would say just on the horizontal or vertical model. I think it needs to be a mix of both, where some sheets like you do need to scroll down, create very detailed revenue build ups on like individual revenue segments across tabs. But ultimately it's probably better to minimize scrolling. So maybe like when needed, build up in a vertical fashion, but try to, you know, have multiple tabs to the extent possible. Hundred plus tabs is okay. Sometimes preferred, if you will.

[00:44:08] Host: Paul Barnhurst: Interesting. Yeah, we get some people. I had one go uh, be vertical as much as you can. The rows are free. Really. Tabs are free too. So we definitely I'd say most people go more vertical, but it depends on the type of model you're doing. My view is if I'm building a basic three statement, I'm going to do it almost all vertical. If I'm building a Fiona model, I'm going to tend to be more horizontal because I usually have a bunch of cost centers. There's 30 different cost centers, and I'm not putting that all on one sheet. It's much easier to have it be the exact same on every single sheet and just update the numbers. And so I think it depends on the type of model you're building preference. But I'd say most people have on the show probably lean a little bit more toward vertical than horizontal, but it's definitely one of those that there's a lot of opinions. It's not like circular references is 90% or no. There's a few that say yes. You know, that's probably the most clear one of any on this list where there's don't do it or an external workbook links is probably next because everybody's been burned by them.

[00:45:09] Guest: Matt Lee: Yeah, well, that's external work. Links like that should be no right. You have to alt, ESV, alt, ESF instead of, you know yeah I yeah, yeah.

[00:45:18] Host: Paul Barnhurst: It's one that's almost overwhelmingly no. And it's not only no I get I get hell no. You know every so often like just don't do it.

[00:45:27] Guest: Matt Lee: I try not to curse on here. So yes, I, I share the same philosophy.

[00:45:31] Host: Paul Barnhurst: Yeah I try not to. But if that's the worst, I said we'll live. All right. So tell our audience where you're at with pneumo if people want to learn more about it, you know when it will be available. So just take a minute and talk about that. If people want to learn more about the kind of Nummo  and where you're at with the tool.

[00:45:50] Guest: Matt Lee: Yeah. So where we're at is we are going to be releasing it to like a closed list. Think of those who are either in banking, some people in PE, but that kind of banking is mainly our core audience right now. And we do want to end up releasing to anybody touching spreadsheets within the realm of finance. Think of everyone in financial services that is the vision but starts with them. We're going to be releasing it pretty soon in terms of functionality available like our MD model reviewers available which pressure cooks models also like being able to have, like all of that private file extraction that I mentioned, kind of getting you to the starting point to then have backup tabs ready to then create further tabs, whatever deliverable you're trying to create. Chat is working well leveraging queries, manual edits, shortcuts. Think of everything like that in terms of the product that should be coming out. Don't quote me on this, but like within a month or two to those individuals. And so that's really nice where we can start getting user feedback. Have a couple like boutique banks or middle market banks like going to be piloting this. And then from there we're hitting the ground running. And ideally we can get it to everyone's computer, including yours. Paul.

[00:47:03] Host: Paul Barnhurst: All right. Well I'll wait for it to come on my computer. I have it on my list to be able to check out. Thank you so much for joining me, Matt. It's a real pleasure. We'll put your website, your LinkedIn profile in the show notes so people can find you. And, uh, excited to see how, you know, the AI workflow of the future is becoming a reality and what you and all the other founders are doing, even though sometimes I feel like the hype is a little much from some, I'm grateful they're all in here because at the end, innovation and competition helps everybody raise the level that they're, uh, they're playing at and that helps us the end modeler. So kudos for you for jumping in and doing this and we wish you nothing but success Matt.

[00:47:50] Guest: Matt Lee: Thank you Paul. And again appreciate the time. Happy to be on. And then I will send you those Pokemon cards and also Nummo  I'll ship it to you.

[00:47:58] Host: Paul Barnhurst: All right. Perfect. Well thanks again Matt.

[00:48:01] Guest: Matt Lee: Thank you Paul. Thank you.

[00:48:03] Host: Paul Barnhurst: Financial Modeler’s corner was brought to you by the Financial Modeling Institute. This year. I completed the Advanced Financial Modeler certification and it made me a better financial modeler. What are you waiting for? Visit FMI at. Backslash and use Code podcast to save 15% when you enroll in one of the accreditations today.