Financial Modeler's Corner
00:29:04 · Paul Barnhurst
[00:29:04] S0: I haven't given that a ton of thought. I think the thought leadership will continue. What I fear is it's gonna get watered down over time. So a good example is Carl Seidman recently shared... He he had shared a financial model, and then he was attending some... Something Someone was sharing. The financial model they were sharing looked almost exactly like model he had built. There's, like, 90% the same. And what he realized, he had recently done a training where he built a lot of it with, I think, Claude or, you know, one of the AI tools. And we're furious. Everything's gonna look the same. We as humans, one of the things that makes us great is the diversity via thought, the diversity that's out there. I would say even in the modeling field, even though, yes, there are standards, we all have a signature. As, Jeff Robinson said, he's done... Trained thousands of people, right, there in The UK, he made a comment once. He goes, I could usually tell who trained them when I'd see a model because... Or, you know, what style they had been trained in because everybody has a unique signature. Do we really want that to go away? That's what I worry about, whether it's thought leadership, whether it's modeling, whatever it might be. AI doesn't have the creativity that a human has. And so there's just something to be said value. Right? We've all seen it. Where you're reading a post, and the information may be valuable, but you could tell it's 100% written by AI, and it just turns you off because you're like, I want some original thought. I wanna connect with the person here. That's what that's what I fear more than anything even though I think we'll continue to have content develop, we'll continue to have thought leaders regardless of how good AI is. Is the next generation gonna lose a lot of that creativity and part of what just makes us who we are?
00:30:50 · Giles Male
[00:30:50] S1: Yeah. I agree. And, actually, the creativity words are really important one. We... We've mentioned that in some of our training recently as well. Even when you think of the use of AI, ideally, you need to be quite creative in your thought processes and interactions with AI to try and get the most value out of it. So that's something that I try and keep in my mind, like not to just fall into a lazy mindset with modeling and AI of just that, okay, we'll build me a free statement model. Actually, you've got to engage with these tools quite creatively to get the best outputs. I don't know if if that's something that resonates with you.
00:31:27 · Paul Barnhurst
[00:31:27] S0: No. I do agree. I mean, I... You have to spend time with them. There's an art and a science. And for certain parts of the model, you want it to be science. Right? A debt revolver. There are other areas within a... The balance sheet balancing. That's not an art. There is science to that that you need to follow or it won't balance. But how you show the summary, how you do other things, there is that ability to bring art into it. You know, exactly how you do the scenarios, yes, there's some structure to that, but there's also some freedom. What assumptions you use? How many? These are all things that... Whether you wanna call it creativity, knowledge, they're human judgment. So even if you're... You wanna say, oh, I don't like to be creative. Fine. Take out creativity. You still have to involve human judgment in the process, and that leads to a different outcome than turning over your judgment to AI, which is happening. Now is AI building models for us? The answer is unequivocally yes. Is it building good models in the hands of a right person again? The answer is unequivocally yes at this point. Can it do a lot of the work and save us hours? Right? Those are all yeses. Those are no longer a debate like when we started the Mod Squad a year and a half ago and we were wondering where it would be. Those have been settled. The question is how you're gonna adopt it, when you come on board, and how you responsibly use it at this point to meet.
00:32:56 · Giles Male, Paul Barnhurst
[00:32:56] S1: And you mentioned The Mod Squad. I mean, god, if you and I reflect on the year that we've had... What was it? Pretty much a year ago, I think, we started The Mod [00:33:04] S0: I think the first one was September, late August or early September of last year. We started talking about it, I think, the end of July. It took us about a month to find our logo and our name and all that. [00:33:14] S1: Yeah, which we used AI for, I think, in part, at least, didn't we, to try and brainstorm the name? But you think when we stand, for anyone that doesn't know, listening or watching, this is Paul, myself, and Ian Schnoor, who's the founder of the FMI. And a year ago, we were just testing third party AI tools in Excel because we didn't have Claude and Copilot in Excel. There weren't agents. The only link to, like, in app experiences with with these third party tools. [00:33:44] S0: Right? Copilot came out about a week after we started. We did our first episode. I think our second was testing Copilot online. Online. I remember. The agent. Not not Copilot in Excel, but the actual agent that would build for you, just so everybody understands. Copilot had been out for a while, but it could only give you... Answer your questions. It didn't have an agent that actually could build for you. So the first one was online. And do remember how awful it was the first time we tested that thing?
00:33:44 · Paul Barnhurst, Giles Male
[00:33:44] S0: Right? Copilot came out about a week after we started. We did our first episode. I think our second was testing Copilot online. Online. I remember. The agent. Not not Copilot in Excel, but the actual agent that would build for you, just so everybody understands. Copilot had been out for a while, but it could only give you... Answer your questions. It didn't have an agent that actually could build for you. So the first one was online. And do remember how awful it was the first time we tested that thing? [00:34:09] S1: Yeah. But I remember there's been a few points. There was definitely the moment when we tested the in Excel online agent where we were like, oh, hang on. That's better. And then I think the beginning of this year, I can't remember if it was just certain model releases It was [00:34:23] S0: open kind of four, five, and five late last year. We tested it at the beginning, I still remember Ian's comment, this changes everything. He was like Yes. Kind of his jaw almost dropped like, wow. This is actually good. Not, I could see where this could help, but man, you gotta be really paying attention. It was, yeah, this can really help. You still need to be paying attention, but it's not this can kinda help. [00:34:48] S1: Yeah. Hell of a journey. Alright. I'm I'm gonna I'm gonna move some questions you will be very familiar with. Not quite quick fire round, but we're definitely going to hit you with those. The first one, what's your worst modelling experience, whether it's yours or somebody else's?
00:35:06 · Paul Barnhurst
[00:35:06] S0: So I would say probably this is the greatest learning, but the worst one. I inherited a business that was the most complex business we had. So I'm gonna tell a story because I think everybody will appreciate this. I'm working in FP and A. We get a new CFO. He basically tells all of us, all your models suck. Everybody's coming to headquarters next week, and we're all building new models. That's litter... And if you know him, that's basically the way he said it. He may have even used a few f bombs in there as he said it. You know, very colorful. And so we had to go to Texas. And one of the first things he wants us to do is we're using this offset to build a ramp. And I had never done it, and I couldn't figure out the formula, and he could tell I was struggling a little bit. So this is only ten years ago. I'd never learned proper model training. I just kinda figured everything else out of my own. I'd call myself a spreadsheeter at that point. I'll be honest. This is ten, twelve years ago. So he called in an investment banker to work with me on this model. We spent six months and got to the point where we had something that was decent, but it could never really support the business because our data was so crappy. We didn't have an understanding. So I spent the next year basically remapping all our data, and then finally, I told myself, I have to rebuild this model. Crap. It's gonna take forever. I'm not even sure I know how all of it works. Still occasionally got circular references in it, which I never figured out in the prior model. I just turned it off and... Oh, and the numbers look good. I shouldn't admit that. So here's all the horror stories. Dim is just going right... Rolling over right now going, what, sir? You idiot. And so I rebuilt it, and and it was still a complex model. It was a really complex business, and this is just revenue. But then I can have real conversations with the business. Like, hey. For this key customer, we're estimating this for this product, and these are the campaigns, and this is the timing. And I could start to make important decisions and take my boss through what was really going on, and we make changes. Here's what happens to margin. And that was the best learning experience in modeling I've had in my career, but it was painful. That model at times was just awful. Yes. We broke many modeling rules.