Financial Modelers Must Master the Fundamentals Before Trusting AI with Chris Reilly
In this episode of Financial Modeler’s Corner, host Paul Barnhurst welcomes back Chris Reilly for a deep and practical conversation on financial modeling fundamentals in the age of AI. Chris shares why mastering the basics still matters more than ever, how AI is changing modeling workflows in reality (not hype), and where automation genuinely adds value versus where human judgment is still essential.
Chris Reilly is a former private equity professional and financial modeling expert with experience spanning restructuring, FP&A, treasury, and middle market private equity. He began his career at FTI Consulting during the financial crisis, working on major bankruptcies, including Lehman Brothers, before moving to Hilton Worldwide and later into private equity. Chris is the founder of Financial Modeling Education, where he has trained more than 90,000 professionals worldwide, teaching real-life models used to acquire and manage private equity-backed businesses.
Expect to Learn
Why financial modeling fundamentals matter more than ever in an AI-driven world
How Chris actually uses AI in real client models versus online hype
Why simple, well-built models outperform overly complex ones
How to balance technical modeling skills with business decision-making
Here are a few quotes from the episode:
“AI is a great accelerator if you already understand the fundamentals. If you don’t, it just magnifies the problem.” - Chris Reilly
“There’s a missing middle right now where people are skipping the fundamentals and jumping straight to the results.” - Chris Reilly
Chris emphasizes that while AI and automation can improve efficiency, they do not replace the need for strong accounting knowledge, modeling fundamentals, and business understanding. He explains that the best modelers focus less on tools and functions and more on clarity, reliability, and helping decision makers understand what to do next.
Follow Chris:
LinkedIn: https://www.linkedin.com/in/chris-reilly-mission-capital/
Homepage: https://www.financialmodelingeducation.com/
Email list: https://financialmodelingeducator.com/
Follow Financial Modeler's Corner:
LinkedIn Page- https://www.linkedin.com/company/financial-modeler-s-corner/
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/497oAqW and use the code “Podcast” to save 15% when you register.
In today’s episode:
[02:35] - Fundamentals Before AI
[05:55] - From PE to Teaching
[09:17] - Dynamic Arrays in Practice
[13:25] - Limits of Dynamic Models
[18:09] - Modeling vs Technical Debate
[21:04] - AI Hype vs Reality
[26:45] - Avoiding Bad Models
[33:00] - Advice for Modelers
[35:38] - Future of AI in Excel
[38:02] - Closing Thoughts
Full Show Transcript
Host: Paul Barnhurst (00:00):
Financial Modeler's Corner is the world's Premier modeling podcast. It is brought to you by the Financial Modeling Institute, the world's leading financial modeling accreditation organisation. Welcome to Financial Modeler's Corner. I am your host, Paul Barnhurst, aka the FP&A guy. In this podcast, we talk all about the art and science of financial modeling with distinguished financial modelers from around the globe. The Financial Modeler’s Corner podcast is brought to you by the Financial Modeling Institute. Financial Modeling Institute offers the most respected accreditations in financial modeling, and that is why I completed the Advanced Financial Modeler. And then I'm a big fan of the programme this week, I'm thrilled to welcome to the show, Chris Reilly. Chris, welcome to Financial Modeler’s Corner.
Guest: Chris Reilly (00:52):
Paul, thanks for having me. It's actually my third time on the show, and I'm thrilled to be back, so thanks so much for inviting me.
Host: Paul Barnhurst (00:58):
No, I'm really excited to have you back. I'm glad you said yes. I feel like it's by honour and normally I read a guest bio. We're going to go a little off script and see how I can give a high-level bio of Chris, and he can correct me at the end. So I'm going freeform today, Chris.
Guest: Chris Reilly (01:14):
Love it.
Host: Paul Barnhurst (01:15):
All right. So Chris Reilly runs financial modeling education. He's trained thousands, I think, what about 90,000 now in financial modeling or something like that?
Guest: Chris Reilly (01:25):
Yeah, just over 91,000.
Host: Paul Barnhurst (01:27):
Just over 91,000. He has courses through Wall Street Prep. They're private equity. He has ones in the Wharton FP&A programme. He has LinkedIn learning and he has his own course library that he sells through financial modeling education. So he is one of the most respected modelers out there in the world. Does a lot of great work. He worked in private equity, worked in some FP&A, and then started drone business. What's it been four years ago now? It's
Guest: Chris Reilly (01:51):
Actually been six
Host: Paul Barnhurst (01:53):
Years ago now. You started with consulting and then got into education a couple years in,
Guest: Chris Reilly (01:57):
Correct.
Host: Paul Barnhurst (01:59):
Yep. Okay. That's right. So that's a little bit about his background, and I think it's been what, probably eight months since you've been on the show?
Guest: Chris Reilly (02:05):
I think so, yeah. Last time we did a deep dive on three statement modeling. We got pretty technical on some steps and kind of just helping people a little hard to visualise via audio, but we just talked through a couple of mechanics that people can do if they're building out that kind of model.
Host: Paul Barnhurst (02:20):
Yep. I remember we talked about indirect and direct method and how you like to tie those out and some different things of how to think about that model as you're building it up. Well, let's start with what have you been up to in the last eight months? Tell us a little bit about your business, what you're doing.
Guest: Chris Reilly (02:35):
So I've really doubled down on education and in the emergence of how AI is changing our industry. While I'm embracing the technology a lot, I'm also finding that there's a gap of understanding when it comes to mastering the fundamentals. And I feel like now is a very important time to grasp these accounting and finance fundamentals so that you're in a position to accurately interpret some of the outputs that AI might give to you. And I feel like a lot of people are skipping ahead to this result and there's this missing middle layer of people who've done work the hard way that know the fundamentals. And so I'm just refocusing my messaging around that because I actually think it's really important for people to know that stuff before they're using this new technology to make their work more efficient. So I'm very focused on the education space still, and then I'm also still working with clients pretty regularly just to keep my skills sharp and stay in the market. And my clients are mostly what I call middle market type companies, like 200 million revenue or less. So it's a combination of those two things.
Host: Paul Barnhurst (03:44):
Got it. I think I really relate to something you said there on the whole AI thing of helping people realise you have to understand the fundamentals. If you want to use ai, well, you have to understand modeling the accounting, the Excel. I think I've shared some of your quotes. One of the presentations I did something put on LinkedIn kind of about this one time of, I think it was a little bit how many of us have inherited somebody else's model and start it over now inherit a model from AI that uses more complex formulas and that we can't talk to in the way we can a human. Do you really think you're going to take that without making any changes?
Guest: Chris Reilly (04:18):
It's an interesting time right now because everything is evolving and I'm sure what I say here today could be totally different in a month, but it is sort of interesting to see the different branches that this technology is taking. And some of what I see online doesn't quite match what I'm seeing in the real world, but that my sample size is small. So it's all kind of different and I think people are going to embrace this technology in different ways, but right now I'm still feeling like it is a doubling down on these fundamentals is still the most important skill. I think that just so you can set yourself up to leverage this technology in the future and eventually make yourself more efficient.
Host: Paul Barnhurst (04:59):
Yeah, a hundred percent agree with you. And we'll get into the ai. I have a few questions I'd love to kind of get through there, but what I always like to say on the AI side is AI is a magnifier, and I think it's true of a lot of professions, but financial modeling, if you know what you're doing really well, it will magnify that and help you be more efficient. If you don't know what you're doing or you don't know it very well, we'll magnify that as well. It's just a matter of time. What I mean by that is you've seen it, you, it's good to build a model. It doesn't balance. You try to figure out why it gives you a wrong answer or has a sign wrong, or you have to know things really well to troubleshoot it and to validate those assumptions. Even if it all balances and looks pretty, there's still a lot of things you need to understand.
Guest: Chris Reilly (05:39):
Yeah, absolutely.
Host: Paul Barnhurst (05:40):
And that's where your education comes in. So let's start with backstory. How did you get into training? You started a consulting business, now you're very much focused on training. You've trained almost a hundred thousand people. Give us a little insight of how that came about.
Guest: Chris Reilly (05:55):
So it came from my private equity days, honestly, when I was working on doing buyout deals, lower middle market, and then also managing the FP&A function for those portfolio companies and the models that I was building, they were just insanely complicated. They weren't so template oriented. It was multiple scenarios, multiple tranches of ebitda, multiple EBITDA adjustments and all kinds of wild structures. And I said to myself, the training that I received before this just does not compare to what I'm doing now. And I said, I wish I had some way of knowing this in advance. And so I started by recording myself doing an LBO model, frankly just to help me in my own deals. And as I got through that, a few people just started reaching out, Hey, can I see this recording that you built? And that's when I got this idea for, oh, maybe this could be a product, maybe this could be a course.
(06:55):
So I built my first course in my early days of going out on my own in kind of 20 20, 20 21 and released it. And much to my surprise, it did quite well. I didn't really do market research at the time. I knew about Wall Street Prep and these other players, but frankly, I just sort of went for it. I needed it for myself. And then I got a lot of positive feedback on it. And of course as a business owner, I realised the digital product is quite scalable and a nice income stream to have. And so then people started to ask for more of them. And so I started to just make a few that were simpler, like a 13 week cashflow and then a three statement model in the FP&A space. And then software models and SaaS modeling became very popular because you're modeling contracts, you're not really doing inventory based modeling.
(07:42):
So I built a prebuilt template for that as well and kind of rolled them all up into this. Now what is a comprehensive training package? And so it basically came from something that I needed that would help me, other people wanted it as well. And so then I kind of got this idea, oh, this could be a business. And so I've been doing that pretty much ever since. It's my primary focus, like I mentioned earlier, but then I just make sure I'm still doing deals with actual clients so that I'm staying sharp in the market.
Host: Paul Barnhurst (08:11):
The deals you're doing is really mostly to help you stay current and make sure you're doing some real work modeling from time to time, but the core business now is training? That's
Guest: Chris Reilly (08:21):
Correct. Yep.
Host: Paul Barnhurst (08:22):
Alrighty, got it. And so you offer, it sounds like it's primarily a comprehensive course that has all those things. You're not offering a bunch of different courses. It's really here's the whole package.
Guest: Chris Reilly (08:32):
Well, there's just sort of bundled together into one package that goes from beginner to advanced. So that's just the progression of the way I thought it'd be helpful to build things. People can get some introductory stuff and then they can get into the advanced content. So that's how it's structured and keeping it that way for now until I can see how the industry changes and then see if I'm going to make a course maybe in maybe next year possibly. But I'm just kind of watching what's happening right now and how the landscape is changing a little bit
Host: Paul Barnhurst (09:00):
And it's a good time to watch. It's definitely changing a little bit. I'm curious, is there anything that has you really excited that you've seen out there with these changes like AI tool or even not ai, but anything in modeling or education at the moment that you're really excited about?
Guest: Chris Reilly (09:17):
So to take a step back from ai, the thing that I really like the most right now, even though it's been out for a few years now, is dynamic arrays. And I think that's been one of the best changes to the Excel interface in years because you're now making automated formulas that scale with your model and they're much less error prone. And especially when you integrate those with tables, you can update your models really quickly, very efficiently. You make a lot of dynamic dashboards or whatever you have to make. What I'm really liking now is just integrating that technology into the models that I'm building and then to just weave AI in a little bit. I will sometimes use AI to help me write the dynamic array formulas because they can be a little bit more complicated, especially if you grew up on what I'll call old school Excel, which is a late two thousands sort of mid two thousands kind of thing. The dynamic array, it's a new syntax and stuff that I still am learning. So the AI helps me speed up that portion of it, but I'm really liking the automation that I can get from those dynamic arrays for refreshing financials or refreshing headcount data. And so I've really leaned into that a lot when it comes to building client models.
Host: Paul Barnhurst (10:31):
Have you brought that into your training yet or is it mostly client models or how are you thinking about that? Because a lot of debate around dynamic arrays and should we build fully dynamic models and how complex it is. And so I'd love just a little bit kind of dig on this a little further.
Guest: Chris Reilly (10:47):
Yeah, I haven't pulled it into my training yet because I still feel like I know sort of the what'll call the 2080 of it, not the 80 20. I've messed around with a couple formulas that I find are really impactful. They help me in the moment get the job done, but I don't feel as though I'm in a position to educate on them just yet. And I know Carl Seman has a really nice dynamic array course on LinkedIn learning and he's extremely skilled at that. And so right now I'd almost be pointing people there. I may look into that. I've
Host: Paul Barnhurst (11:17):
Taken the course, it's really good. Great. I haven't any problem with it, but I took a couple of his just to see how he thinks about it.
Guest: Chris Reilly (11:22):
So I view him as one of the most skilled modelers in the space. And so I may get into that in the future, but right now I'm just using them as a get the job done on my current client models. And then if I feel like I get to the place of mastery, which I'm always very slow to admit to myself, then I may start to build a much smaller course around that. But nothing yet.
Host: Paul Barnhurst (11:44):
Do you have a favourite dynamic array? Is there one you're really enjoying right now or one you're digging into? Because there's what, I think there's about 42 of them now.
Guest: Chris Reilly (11:52):
Yeah, there's a bunch. I mean I use unique the most. I think that one's just nice to clean up things really quickly, but if I'm being really creative, I'll use a combination of filter unique and sort by altogether and I can make a really nice list of whatever I need. And the example that comes to mind would be a headcount schedule. So I had two separate headcount schedules in a models one for terminated employees, one for current, and the dynamic array would sync them both together into a master headcount schedule and then sort it by higher date and by department. And then of course that dynamic array is sort of this living breathing piece of the model, so it's easy to reference with other formulas, and that was mostly built by using the combination of those three dynamic array functions.
Host: Paul Barnhurst (12:39):
I know filter unique and sort or sort by you can get a lot done. I just, in fact this week just finished a training session where we spent one day showing how to build an entire revenue dashboard using dynamic arrays. And my partner that I trained with, he did it with pivot tables so they could see the difference. And like you said, I don't feel like I'm a master yet on it. There were a lot of questions from one of the people where I'm like, I didn't even think about that. I'm like, oh, here's how I think I'd handle that or I'd have to come back to you. And so you really got to see that messy middle of dynamic arrays are amazing, but they can get really complex in a hurry if you're trying to do all dynamic arrays. It's extremely difficult to build in complete model that's completely dynamic arrays for a financial model.
Guest: Chris Reilly (13:25):
I don't know that I see myself ever going there. I almost use them as these efficient bridges from point A to point B, a data source that can be cleaned up and then it's now in this consolidated place that I can then reference for a more traditional model. Fully dynamic scares me a little bit, especially if something breaks, nobody's going to know how to trace that. Especially we talk as advanced modelers and we get into these dynamic array functions. Somebody who doesn't have that deep Excel or modeling experience and is more of an A plus B plus C kind of person, they're going to say, this is way over my head and I have to keep that in mind. We always have to remember the audience or who's going to be the end user of this file. And most of the time it's somebody who doesn't know this Excel system the way that I do. And so I try to keep some of those automations on the backend where they won't have to touch it. And where they interact with the interface is something that's just as much more understandable and a little bit more universal.
Host: Paul Barnhurst (14:30):
And I think that makes a lot of sense. You got a lot of different opinions, but the reality is to build fully dynamic models, you're getting complex, you got custom lambdas and all kinds of things that the average person isn't going to understand. Now, depending on what they have to do with the model, you could argue maybe they don't need to understand it, but if you're going to train it, that's a much higher level than, Hey, you need these 10 functions to do 95% of what you're doing now let me show you this big huge lambda that manages your cork cur. Why can't I just do equals beginning plus minus
Guest: Chris Reilly (15:03):
This
Host: Paul Barnhurst (15:03):
Equals end and just link it. Then learning this big huge lambda.
Guest: Chris Reilly (15:08):
Yeah, I'm kind of the same. And actually Lambda freaks me out. I haven't really messed with it too much. So that's maybe, again down the road, I don't want to educate on something if I'm too scared to use it myself.
Host: Paul Barnhurst (15:17):
Yeah, I've started to use it a little bit more recently, but I'm with you. I've always kind of Lambda and LED have scared me a little bit, but I'm getting there.
Guest: Chris Reilly (15:26):
That's good. It's a slow process,
Host: Paul Barnhurst (15:28):
So I'm glad we're not alone on that. Sometimes I think people think we're always on the cutting edge of everything as trainers, I imagine you feel like that sometimes people ask you something and you're like, nah, I haven't even used it.
Guest: Chris Reilly (15:39):
Sometimes I put that pressure on myself too. It's like, oh, Microsoft just released this new thing. I should be all over it. And then I think I kind of need to be honest. I know what I know and what I don't, and I got to teach people what I'm good at and double down on the pieces that I can help people with and just be clear that there's things that I don't know and admit there's things that I'm never going to know and I'm just going to figure that out through completing these projects that have to get done in a certain way. And so long as I've met that result and then I say, okay, that's good enough for now. So I don't feel like it's a great use of my time to always just be on this cutting edge of this new stuff. I kind of have to sit back and observe and see, okay, well what can I use to solve the problems for clients or even my own business right now and not always be on the cutting edge of something brand new.
Host: Paul Barnhurst (16:29):
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.
Guest: Chris Reilly (17:23):
Yeah, it's too overwhelming. I mean, I couldn't even keep up with it. So like I said, I focus on what I can do to get the job done and I keep it there. And at some point if I feel like I've achieved a modest level of mastery, then maybe I'll package that into some kind of educational thing, but that's not for a while. I would say
Host: Paul Barnhurst (17:40):
No, I like the way you're thinking there. So next, I want to go to, you recently talked on LinkedIn. You posted about the kind of whole VLOOKUP versus X lookup debate AI and said, look, what we have to keep in mind is the real question is we're trying to help the business better understand things. So why do you think there's so much focus on the technical versus getting to helping the business make better decisions, which is really the goal of the model at the end of the day?
Guest: Chris Reilly (18:09):
So to go against my own post a little bit, I think there is fulfilment in the craft of building the models themselves. And I think we enjoy to put them together and I think we like to see are there more efficient ways to get this done? There's a pride in completing the project, and so if you believe that one function does it differently or better than another, I think there's a lot of debate on getting the technical piece done. At the same time when I walk into a boardroom or a managing partner meeting, they don't care that I've used X lookup or some product. They care about the revenue, the margin, the ebitda. Is this company profitable? What is it worth? Can we take it out to market? Or something like that. So there's always this juxtaposition of technical mastery, but with the ultimate purpose, which is the model is just a tool to help us figure out what to do next with the company.
Host: Paul Barnhurst (19:02):
That makes a lot of sense to me. And I was going to ask, how high was your joy? How good you feel the first time you had a really tough model and the balance sheet didn't balance and then you got it to balance?
Guest: Chris Reilly (19:14):
Yeah, I mean it's almost painful at that point. I'm just ready to leave, go home, frankly. But I think it's more about once you figure out how it clicks and then that it's working and it's reliable and some of that stress frankly goes away and okay, I now have this reliable engine that I can trust because I know how it works and I can use it for its intended purpose, which is to communicate to the higher ups of what to do next with this company. So there's a relief in getting that done, but there's a fulfilment in the build process in the first place.
Host: Paul Barnhurst (19:50):
Correct. Right. There's kind of that joy. There's the relief of, oh, I can now present this to my boss. It ties out, but there's a certain amount of joy, and I understand this now when you finally master index X match or whatever the function might be that at first you're like, I'm never going to get this function. Then there's just a day or just a time when all of a sudden it clicks. I just did all that and I didn't even have to think about it. Yep,
Guest: Chris Reilly (20:13):
Exactly. And that comes with time, but grinding through it and earning it and learning that skill, there's a fulfilment there.
Host: Paul Barnhurst (20:21):
A hundred percent agree. And definitely time. I think it was just a week ago, I finally feel like I'm starting to get comfortable with a number of different things in combinations with dynamic arrays beyond where I was before. And it's kind of cool like, oh wow, I can do this now. And I think back when I started it was like, I'd never doing a function like that.
Guest: Chris Reilly (20:42):
Right. Yeah, of course.
Host: Paul Barnhurst (20:45):
And so let's shift gears here a little bit. I want to spend the rest of our time on ai.
Guest: Chris Reilly (20:50):
Sure.
Host: Paul Barnhurst (20:50):
That's just a big topic. So let's start with your take. How do you see AI in financial modeling? Where do you think we're at with it? We'd love to just get your broad thoughts of how you're thinking about it.
Guest: Chris Reilly (21:04):
Yeah, I'm always wary about talking about it because I feel like I will eat my words like three weeks from now and some new technology will come out and make me look like a moron. Okay, so I'll just say as I sit here today, that's my qualifier right now, I'm finding the AI hype, exhausting, and I find that what I'm seeing online is different than what I'm seeing in the real world. Nonetheless, that doesn't mean I'm anti ai. In fact, I'm all for it. I use it pretty much every day for different parts of my business. And more specifically, I use it on modeling, but I don't use it in the way that I think I see it presented online, which is building a DCF model in 14 seconds or Wow, look at this thing, put together this amazing multiple scenario, boom, boom, boom. And it's really cool. And I say that because I tested agent mode earlier this year and I said, here's a great use case. I'm going to give it the prompt from one of my courses because my course starts out as an email from A CFO. Hey, build the model. Here's a bunch of sheets and do it. And I'm like, this is basically a prompt, so lemme try this.
Host: Paul Barnhurst (22:04):
What was your LinkedIn course? This one?
Guest: Chris Reilly (22:06):
No, this is from my website, a three statement modeling course. And so I'm like, this is a prompt. Lemme try it. And it was a total mess. I tinkered with it for hours. I basically got nowhere broken model unusable. I tried to re-prompt it over and over and over again, make it more and more detailed and basically got nowhere. And so I'm saying to myself, well, this was a massive waste of time. I've built nothing. And it led me to some weird realisations about just handing over the keys. One is, let's say it gets it perfectly, which I think there will be a future where I do snap my fingers and boom, the perfect model appears. Weirdly, I'm almost creating the worst case scenario for me personally, which is getting a model from somebody else, which I hate. Now it'd be perfectly built. And so as somebody who can understand what I'm looking at that maybe it's not so bad, but I think all of us get a model and we're like, oh, I don't want to go through this thing.
(23:01):
I'd rather just rebuild it. So as I'm letting this engine hum in the background, I'm like, man, I got a time I would've spent building and also learning the business. I'm now just waiting to get a model spit out to me that I have to then interpret and learn the business. So am I really gaining any time here? I don't know that I am. And then there's also that emotional piece we just talked about. There is fulfilment in building something. If somebody hands you a prebuilt Lego, that's cool. You like Legos because you built it and you're proud of it and you put it on the shelf, this is this cool thing that I built. We remember doing that as kids or maybe now all good. So I think modeling is that same. It's like you got pieces that you are proud of that you constructed, and the construction of that is also you mastering the business at the same time.
(23:51):
And so as I mess around with ai, I am like, I'm not really feeling like I'm getting the time value out of this in terms of just handing over the keys. Where the value comes from for me is using it to help me with certain aspects. We talked about the dynamic array, having it write dynamic array formulas for me that I sort of just brute force in my model until they work. And I'm not fancy. I go chat GPT, copy, paste, copy paste. I'm not overly integrated yet with some of the plugins, and I'm going to start working on those more now that they are more desktop friendly. But when I was testing this, they weren't. But it's something where it's a predictive machine, so it predicts the logic that I want, but at least I can test, test, test until it works and so then I can trust it.
(24:35):
Same idea with power query. Do I know how to write custom code and power query? No idea. But I can give it the instructions of what I want to say, Hey, can you syphon the information from this CSV and pull it into a table in my model because that's going to make my life a lot easier. I can just export something from a system to CSV, stick it in a folder, press refresh and it's done. And I had AI write that code. So what I'm finding right now is there's this really good balance of if somebody knows what they're doing and they actually know what they want, then the AI in experienced hands is a great accelerator like you mentioned to speed things up. And the model that I just completed for this mid-market company a few weeks ago was examples of that. Now frankly, I spent more time messing around with AI than I wanted to, but the result I got was really pretty slick and some of those dynamic arrays and bringing things together was extremely helpful.
(25:31):
So I worry that we're moving towards this idea that we're going to press two buttons, we get these perfect models and then we're done. You go home for the day or something like that. It's not going to be like that. Your boss is just going to ask you to churn out 10 more models in the same day instead of one. And you're going to be overwhelmed with this trust factor of do I even trust what I'm looking at? Because until the hallucination problem can be solved, AI would rather give you a wrong answer or a bad answer than admit it doesn't know something or is doing something incorrect. And you can only uncover that by sort of reverse prompting it. And we've all been there and it's like, oh, you're completely right. I'm sorry about that. And we're like, well, wait a minute. What were you just telling me in the last couple chats? So I want to be clear, I'm very pro ai. I'm using it all the time, but what I'm seeing on the ground doing these client engagements is very different than the hype that I'm seeing online. And this snap your fingers perfect model comes out and I understand that that hype creates attention and that's fine. And I do think there will be a future where I can snap my fingers and I get that model, but it's a major disconnect from what I'm seeing right now.
Host: Paul Barnhurst (26:45):
Bad financial models can lead to bad decisions or worse. So how do you minimise the risk of a bad model? You make sure the models you build are great Financial modeling Institute developed the Advanced Financial Modeler accreditation programme to help modelers like you. The AFM programme offers a step-by-step approach to building world-class financial models. The programme ensures that you know the best practises in model design and structure and will help you brush up on your Excel and accounting skills too. Be the one on your team to build great models. If you want to impress your boss and your clients, get a FMI accredited podcast listeners, save 15% on the AFM programme. Just use code podcast at http://www.fminstitute.com/podcast.. Alright, you ready for rapid fire?
. I appreciate you saying that. I think I echo what you say. I know you've listened to some of our Mod squad series that we did where we tested all these tools and that's the conclusion we came to is very similar to yours of what they're selling doesn't match reality now, could it tomorrow? Sure. But I think let's say it does say we could snap our fingers tomorrow and it builds a model. You still have to understand the assumptions. You still have to present that to somebody. We're not at the point where an AI robot's going to stand up and present it all. I mean then they don't need us at all. Why do we have a job? But we still have to understand everything that went on in that model because your boss is going to expect you to know it as well as if you built it yourself.
Guest: Chris Reilly (28:33):
Yeah, exactly. And then what I keep coming back to is there's a layer of nuance that I don't think AI can, I don't want to say ever achieve, I don't want to look too dumb on this episode, but I keep referencing this recent model just because I just closed the engagement. I could see a world where AI builds a perfect three statement template in maybe a couple minutes. That's great. The revenue build that we went through with this client just to get the top line of the revenue was insane. The amount of sub schedules and micro drivers and little incremental pieces combined with all of their custom metrics on how they measure their funnel, there was just so much information that I don't know that an AI could kind of process. It'd be impossible for you to prompt it to even communicate what you want. And so I'm thinking, okay, maybe AI can kind of grit this shell, but the layer two stuff of all the support schedules, the nuance around headcount, the nuance around revenue builds or even contractor builds, whatever they may be, that piece seems so hyper specific that I always feel like there's going to be this end point where you can take what AI has given you, but at some point there needs to be a handoff where you finish the last 30%, call it something like that.
Host: Paul Barnhurst (29:54):
And
Guest: Chris Reilly (29:54):
That was a pretty big realisation for me when we went through this detailed modeling build on this most recent deal I did,
Host: Paul Barnhurst (30:01):
When you say that, it reminded me an FP&A role where I supported a business where we probably had 20 different products and there was a call centre. There were digital products, there were email products, there were subscription with transactions, subscription without transaction. And of all of our billing data was coded incorrectly. So I did build just lookup tables to get all right. First time we built the model, we spent six months and then I realised I really couldn't support the business with it. It wasn't until a year later that I finally really understood the model and the business. I rebuilt the whole thing and all these different sheets like, well, I got to understand this customer does a difference, so I got to build one. And it's just a huge massive revenue model. And I'm sitting there thinking how long it would take me to prompt that and try to get AI to do all that. Would it be any better when I'm done? Probably not.
Guest: Chris Reilly (30:49):
As far as I know, AI takes the information that humans have given it and it does its best to predict what we want next. And since humans run companies and we make human errors, I think it is in our nature to give it imperfect information and information that's not accurate or data that is wrong or whatever. And if we're going to give it something that's messy and then ask it to clean it and then improve it and then give us a future, it just seems like we're setting ourselves up for something that goes in the wrong direction. May not. But that's kind of just what I would think. And also any kind of black swan event, you're not really going to get in a forward looking thing and maybe you can ask it to run a scenario like that and maybe there's enough sophistication in the model to do so.
(31:40):
But I imagine it's mostly going to be trend-based information based on historical forecast. So it's going to unfold very interestingly. I feel like it changes every week. I see a lot of, I guess what I'll call Excel killers out there, a lot of Excel plugins. I kind of think it's all going to come down to Microsoft and integrating agent mode into the desktop at the end of the day could be wrong about that. But that's just sort of my guess as we sit here today. And hopefully it's at a point where advanced modelers or even intermediate modelers that think it's supposed to help out more people, it can help them get what they want to do, just get it done faster as though you were working with a team of two or three simultaneously and you get a little bit more efficiency out of what you're working on.
Host: Paul Barnhurst (32:25):
And I appreciate that. So we just have a few minutes left, so I want to kind of bring this back and first, as an educator, as a modeler, as an influencer, someone whose people look up to in this space, what's the recommendation you give people today? And I get it, it's ai, it's all changing constantly, but if I ask it today, what's the recommendation you're going to give to the average modeler? How should they think about ai? How much time should they spend there versus focusing on the fundamentals, whether that be Excel or that be modeling, telling the story, whatever. What would kind of be your words here to 'em?
Guest: Chris Reilly (33:00):
So I think everybody should take it with a grain of salt, right? Because I'm sort of incentivized to promote my courses in a way. So I'll just put that on the table first. Nonetheless, I actually think the most important thing that somebody could do right now they're starting out, is to build at least just one three statement model from scratch. And I say that because you simultaneously learn accounting and finance in one holistic project and you also understand the business. And if you understand how the three statements work in your head and somebody says, oh hey, we're going to have this capital spend project, what does that do? And that puts you in a position of ultimately where you want to be in the business. You kind of want to be this strategic finance partner that can help out the operations team and you learn those fundamentals by building, in my opinion, a three statement model.
(33:52):
That's been the biggest thing in my career, and it's the thing I talk about the most. So if you're beginner, intermediate, or even advanced, just build one the hard way from scratch. You get a huge library of fundamentals. And then when you're grounded with that, then is a great time to start getting into, well, how do I make this faster? I can use dynamic arrays, I can use AI to make my dynamic arrays better, or I can use AI to help me with accounting concepts. As an example, quick segue. I used AI to teach me, let's see, operating leases and right of use assets and how to model the rent payment. But I knew how to interpret what was right and wrong because I'd built so many three statement models. So I got it to work fine. I was really proud of myself. I used AI as my accounting coach and I didn't have to go to the CFO of this company and be like, Hey, can you teach me operating leases? Right? So that was a great use case, but because I understood the underlying accounting, so I feel like you've got to start with the fundamentals, do it the hard way once or twice, and then you're in a position to really leverage the technology.
Host: Paul Barnhurst (34:58):
Great advice there. And so I'm hearing it, it's get the fundamentals down, then figure out how to use ai. If I was kind of summarise that all and just kind of a simple sentence.
Guest: Chris Reilly (35:09):
Yes, you said that much more eloquently than I did.
Host: Paul Barnhurst (35:12):
No, I loved your example and I appreciate all you shared. I just wanted to recap it for everybody and see if I could sum that.
Guest: Chris Reilly (35:18):
Yes, definitely.
Host: Paul Barnhurst (35:19):
So we just got a minute or two left before we need to wrap up here. Let's see if I can come up with a fun question or two for you since we've already done rapid fire. So who do you think is going to win this whole AI agent battle? I'm guessing are you're betting on Microsoft? Is that who you think will be the kind of agent tool of the future?
Guest: Chris Reilly (35:38):
I think so. I think either they'll roll it out themselves or they'll acquire the best third-party player in the field and integrate into Excel. So that's my guess as of today.
Host: Paul Barnhurst (35:47):
Alright. What's one thing you would love for them to fix in Excel? If you were in charge of Excel for today, what's something you'd like?
Guest: Chris Reilly (35:55):
I guess I won't say emergent centre. I've kind of navigated away from that. You know what, here's something I would fix. I would love a way to not, so complicatedly have automatic conditional formatting for a number that was hard coded to turn blue and a number that was linked to another sheet to turn green and a number that was linked to another workbook to turn red or something like that. I do all that manually and I know there are macros that do it that I'll use, and I can even set up conditional formatting to do that. And I've posted about that a few times. I'd almost love that as an on-off toggle. It just makes your model style.
Host: Paul Barnhurst (36:32):
You have a modeling file, you just go click, and I want this modeling conditional formatting style to be turned on.
Guest: Chris Reilly (36:39):
Yeah, so it doesn't create any more computing load on the file, like conditional formatting does. It'd just be an on off toggle. So of all the advances they can do, I just want better colours.
Host: Paul Barnhurst (36:50):
I love that. Better colours. It's like we're in third grade, I just want a colour with red, green, and blue.
Guest: Chris Reilly (36:54):
More colours. Yeah.
Host: Paul Barnhurst (36:56):
Oh, black, green and blue probably. But what is one thing, if it went away tomorrow, would cause you the most panic in Excel? One thing that you'd just be like, I'm done. I might go to Google Sheets.
Guest: Chris Reilly (37:10):
Oh well, let's not say something we don't mean. Oh gosh. I don't know. I had have to think too hard. I'm not sure they've got their hooks in me pretty good. I suppose if the equals sign went away, I'd have to go to Google Sheets.
Host: Paul Barnhurst (37:25):
You guess you could do the plus that the old modelers put the plus in front, that's how it
Guest: Chris Reilly (37:30):
Used to work. You're, that's the sign of a true old school. The plus first. Yep.
Host: Paul Barnhurst (37:35):
Or if you still, instead of hitting control T to build a table, if you do control L, you know you've been using Excel for a long time, back when it was the list in 2010.
Guest: Chris Reilly (37:44):
Oh yeah, those are the good old days.
Host: Paul Barnhurst (37:46):
Yeah, all those fun little quirks. Well, Chris, this has been a lot of fun. I really appreciated you joining me. As always. I love having you on the show and chatting, but as we wrap up, anything you want to share with your audience about your business resources, learning more about you, or just a moment to share whatever you'd like
Guest: Chris Reilly (38:02):
Think. If people are interested and they do want to double down on some of the fundamentals that I mentioned, then of course, please feel free to check out my course website. But there's no pressure to do so. I believe anybody who wants to kind of have a solid position in the future, get the fundamentals first and then you can focus on the high-tech stuff later.
Host: Paul Barnhurst (38:22):
I think. Well said. So I think this episode I might label focus on the fundamentals, worry about AI and high tech later and don't buy the hype.
Guest: Chris Reilly (38:30):
I like it.
Host: Paul Barnhurst (38:30):
Alright, perfect. Well, thank yo,u Chris. Really appreciate it.
Guest: Chris Reilly (38:34):
Yeah, thanks so much, Paul. Thanks again.
Host: Paul Barnhurst (38:36):
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 www.FMInstitute.com/podcast and use Code Podcast to save 15% when you enroll in one of the accreditations today.
