In this episode of Future Finance, hosts Glenn Hopper and Paul Barnhurst reflect on the evolution of AI in finance, the biggest lessons from their podcast, and what lies ahead in 2027. They discuss AI adoption, productivity, governance, and the risks of using AI to build business-critical finance applications.
Glenn and Paul also explore the growing importance of human judgment, the challenges of AI-generated content, and the evolution of AI-powered finance software. They share their perspectives on AI agents, AI-native platforms, and why Excel and financial modeling skills remain essential as technology evolves.
Expect to Learn
Why increased AI productivity doesn't always create business value.
How finance teams can balance AI innovation with security and governance.
The risks of vibe coding critical finance systems.
Why human judgment remains essential in AI-driven workflows.
How AI agents and finance software are evolving.
Why Excel and financial modeling expertise still matter.
The discussion highlights why finance professionals must balance AI innovation with practical expertise, strong controls, and critical thinking to create meaningful business value.
Follow Glenn:
LinkedIn: https://www.linkedin.com/in/gbhopperiii
Follow Paul:
LinkedIn: https://www.linkedin.com/in/thefpandaguy
Disclosure: Portions of this episode (such as the introduction or promotional segments) use AI-generated voice narration produced under human editorial review.
Future Finance is sponsored by QFlow.ai, the strategic finance platform solving the toughest part of planning and analysis: B2B revenue. Align sales, marketing, and finance, speed up decision-making, and lock in accountability with QFlow.ai.
Stay tuned for a deeper understanding of how AI is shaping the future of finance and what it means for businesses and individuals alike.
In Today's Episode
[00:00] - Introduction
[03:17] - Future Finance: A Year in Review
[04:07] - AI Productivity vs. Business Value
[05:49] - AI Adoption, Governance & Security
[10:22] - The Importance of Human Judgment
[11:45] - Challenges with AI-Generated Content
[19:40] - The Evolution of AI in Finance
[22:27] - Build vs. Buy & Vibe Coding Risks
[29:46] - Looking Ahead to 2027
[32:28] - AI Agents, Models & Industry Trends
[37:07] - Is Excel Becoming Obsolete?
[42:04] - Closing Remarks & Thanks
Full Show Transcript:
Paul Barnhurst (00:00):
Welcome to the Future Finance Show, where we talk about treasury management on bars. Future Finance is brought to you by QFlow.ai, the strategic finance platform solving the toughest part of planning and analysis. B2B revenue, align sales, marketing, and finance seamlessly, speed up decision-making, and lock in accountability with QFlow.ai. Welcome to another episode of Future Finance. I'm your co-host, Paul Barnhurst, and I'm joined by my color commentator, Glenn. How are you doing, Glenn?
Glenn Hopper (00:54):
I'm good. Let's not get into an Aikman-Buck thing, though, if we're going with sports announcer... They've got the whole thing with the contracts being renewed or not being renewed and that 15, 20 year relationship between them, supposedly a little bit on the rocks. Hopefully, Paul, you and I aren't headed. Did you get your multimillion-dollar contract this year already?
Paul Barnhurst (01:14):
Oh, I don't know. I mean, I've heard how much money you're making and it's on the rocks, baby. No, I'm just kidding. You AI people are getting all the money these days.
Glenn Hopper (01:25):
I guess I'd rather be Aikman and Buck than Romo and Nantz since Romo DUI'd and got fired. I'm already way off the rails. What did we talk about on this show, Paul?
Paul Barnhurst (01:34):
Sanchez and who? I don't know if Sanchez has a partner, but he got kicked for Mark Sanchez, the old USC quarterback. He got himself fired. For some reason, assault doesn't go over well when you're a sports commentator. We can probably spend this whole episode on sports because we're both big sports fans. Maybe that will be our next podcast.
Glenn Hopper (01:51):
I'd love to hear what you're thinking about. Jaren at the Jazz. I'd like to talk about post-jaw grizzlies.
Paul Barnhurst (01:57):
He strained his MCL.
Glenn Hopper (01:58):
Oh, I didn't see that.
Paul Barnhurst (02:00):
We
Glenn Hopper (02:00):
Were
Paul Barnhurst (02:01):
Pretty sure it was going to be a torn ACL, so we'll take eight weeks. He did tear it lightly. So he's going to be out eight weeks. So he's out, our center is out, and we traded our other center. Lauri Markkanen's already missed a few games. We're 1-7. Not that anyone will care about that. And your Grizzlies are like 5-1. They've already beat us twice.
Glenn Hopper (02:20):
Yeah. Well, they started out 6-0 and they dropped their last two and I love it. I picture Paul, if this is somebody's first time tuning into the show and they're like, "What is future what?"You're
Paul Barnhurst (02:30):
Probably
Glenn Hopper (02:30):
Like,
Paul Barnhurst (02:30):
"What
Glenn Hopper (02:31):
The hell?
Paul Barnhurst (02:31):
Let's be honest."
Glenn Hopper (02:36):
I think this is indicative, and I don't know, maybe we'll edit this out. Maybe we'll let it fly like it is, but I think this is indicative of the crazy year that we've both had that we're just off the rails by November here.
Paul Barnhurst (02:48):
Yeah, and I might tell the guys who are editing this that they could even leave this in if they want, but if you want to put a little part at the beginning and then this part, go ahead, have fun with it. It's the last one of the year. So we'll let our audience know we're taking December off. Glenn and I are burned out. You burned out, Glenn?
Glenn Hopper (03:06):
Yes. I need
Paul Barnhurst (03:08):
A
Glenn Hopper (03:08):
Few weeks just to get regrounded. I've got
Paul Barnhurst (03:10):
A
Glenn Hopper (03:10):
Couple more short trips ahead of me in the next couple of weeks. And then I used to work with chief of staff actually who he always said, "We all spend time in the business," and this is a big leadership thing he said, "But we never step back and spend time on the business." So I'm going to mandate and make myself spend a couple of weeks just strategizing and trying... I know it won't really work, but I'm optimistic now, but I'm going to try to strategize and get out of whack-a-mole and fighting fire mode for 2027. I'm already working on my knee exactly. Well,
Paul Barnhurst (03:43):
If you want to have a call or two over the holidays, we can both strategize because I have a feeling you and I both do too much whack-a-mole.
Glenn Hopper (03:50):
Yep, yep, yep. And it's just with AI, it's like just throwing gasoline on it. It's just the whack-a-mole suddenly means I'm just pointing my agent and my bot to go do the work.
Paul Barnhurst (04:01):
Yeah, so speaking of AI, so today the goal of this episode was not sports, was not talk about us. We are going to do kind of a recap of the year, some things we talked about, but something you said of how AI pours gas on it made me think of something I want to get thoughts on. So I just got off a call, a tool that's been around for 10 years. They started as what I'll call a product. They became a platform. Now they consider themselves kind of a full solution, that journey. They're saying in the last year how much AI has increased what they're able to do. They go, "When we first turned it on, it was like 20x. It took us a month or two to figure it out. We said, No, we want about 4x the productivity and be able to test better, really make sure our clients love it versus just pumping out stuff because we can't.
(04:42):
I'd love to get a little bit of your thought on just that general idea within software and then maybe how does that apply to finance? Because I could pump out twice as much content as I pump and I pump a crap load of content. I always have. Everybody's always like, "How do you put out so much content?" I mean with AI, I could pump a lot more, but that doesn't mean it's good content. It doesn't mean it's value. So how do you balance that, the tendency to want to just pour gas on it because you can?
Glenn Hopper (05:12):
Yeah. I guess to bring it back to the finance side and really across business, every department is having this issue. So four years ago we could come in and just do neat things with AI and we were out on the bleeding edge and we could impress people with our dropping public company financials in and having it do charts and graphs and everybody was like, "Ooh, wow, you're a mad scientist. This is crazy."
Paul Barnhurst (05:38):
I remember a fate two years ago when you built a graph, it was like...
Glenn Hopper (05:46):
Yeah, but now my mom is using ChatGPT on her phone to plan recipes and vacations. So everybody's using it. Nobody really understands how it's working. Some people are using it effectively. Some people have lost their minds and are just firing the AI slop cannon. But when you look at a business, I'm talking to a lot of CFOs and really the full C-suite now, and they are all... First it was, "Okay, we've got to do some AI." So their idea of doing AI, and I'm not faulting anyone in leadership on this, it was, "Okay, well here's Copilot, here's Gemini, here's Claude, here's ChatGPT, whatever, go do some AI." But then when they're sitting at dinner or they're asked by their board, "What are you doing with AI?" They're like, "I don't know. I gave everybody access to the tools and they're doing some AI." And people are using it at different levels.
(06:46):
We are in the big part of the adoption curve, and people are figuring out ways to make it their own RPA, smartly or not smartly, but it's like there's not a clearing house, there's not something that they're all going through where they're figuring out is it okay to do this? Am I being safe with our data? Is this repeatable? Is this auditable and all that? So there's a bunch of noise out there. Everybody's kind of playing whack-a-mole with AI, but it's where businesses need help right now is yes, encourage your people, have them go out and vibe code and build with AI, figure out how to automate their work, but don't let them just push things into production. These are not software engineers. They need some way to evaluate all that and not just evaluate it. I mean, one, from the controls level, two, from a security level, three, from a cost level.
(07:41):
I mean, there's so many things that need to be looked at, and that's the problem with this whole bottom up. It's great, everybody has the tool, but it's fully just who's watching this? It's wild west out there.
Paul Barnhurst (07:53):
Well, it's funny, and this is what came to mind, and some people like terrible comparison, but I'm going there. When you think of revolutions, downfalls of countries, it always starts bottom up. How often is it top down where the vice president goes in or the assistant and takes control and really fundamentally changes things? Yes, have we seen that happen? There's always fighting among leadership. But I mean, when there's a fundamental change, it almost always starts from the bottom and it's a groundswell, and it's usually quite chaotic. When the or the peasants or whatever you want to call it overtake the ruling class, it's never a smooth transition. We can look at history. There's
Glenn Hopper (08:41):
Going to be beheadings.
Paul Barnhurst (08:42):
There's going to be... And in this case, there's going to be eggs that are broken and fillings that are hurt and processes to break is probably the right term here because there isn't a clear playbook, there isn't an easy way, and this isn't like the computer or the internet where it was easy to control. It's very easy to turn on or off the internet for everybody when it first came out or choose who gets a computer. Yeah, you occasionally had the rogue person who brought their laptop to work and okay, yeah, they could have shared some data, but that compared to today, it's just a whole order of magnitude bigger. And so the frameworks, it's just going to take time to settle to the new normal. The challenge is, are we moving so fast that we're never really going to have a new normal? When do we hit that?
Glenn Hopper (09:29):
Yeah, I mean, well, we're going to hit it when people start breaking things. You're right, in a revolution things they're going to get, but there's going to be something -
Paul Barnhurst (09:36):
A big company has a couple billion dollar mistake because of AI.
Glenn Hopper (09:39):
Yeah. And just people forget that one part -
Paul Barnhurst (09:43):
Or somebody presses a certain launch code and we have a disaster that we talked about last week with S-1.
Glenn Hopper (09:48):
Yeah. I mean, the problem is AI is getting so good that the default mode would be to get sort of, I don't know, flippant about, I don't know, it was right last time, let me run it again. I'll assume it's right and we stop checking, we take out our discernment and judgment. AI makes some recommendations and we think, "Yeah, that looks pretty good. I'll just pass that on." And we're all just firing around AI generated.
Paul Barnhurst (10:10):
I find myself getting a little lazy. Haven't you found yourself a few times checking stuff probably a little less than you should, if you're honest? No, I imagine you balance risk. If it's high risk, you're not doing that, but I'm sure there's some things where you're like, "I've done this enough times. I know it's going to be good enough and you don't do the check you did two years ago or three years ago."
Glenn Hopper (10:29):
Well, you pick what you would delegate in the past and you're like, "Okay, you know what? If this is something I would delegate, let me just do that." I got burned. Lauren Pearl and I were doing a demo on cybersecurity with AI and we were both traveling and we had a deck and I told Claude, I was like, "Go through and make..." And we had a transcript from a call and we're making sure we were aligned on everything. And I told Claude, "Go through and look at the transcript and make sure what I'm saying in the deck is aligned with what Lauren wants to say and fact check the things I have in the deck." Fortunately, it didn't mess up the deck itself. It went insane in the speaker notes. It was like arguing with itself about Lauren's intentions and there was all these links and it was trying to cite everything in the speaker notes.
(11:24):
And I was traveling, I didn't even look at it. I just was like, "Cool," because I flipped through the slides themselves and Lauren texted and she was like, "I don't know what's going on in your speaker notes, buddy, but that's not anything we can use." And this was the day before the presentation. It all worked out fine. She's more of a professional than I am, but that's an example of you get used to doing a really good job and somehow it went off the rail there and difference between a presentation and doing real client work, the stuff that I wouldn't delegate, but got burned.
Paul Barnhurst (11:56):
I'll share another example. We do a lot of training and now we've started, because the goal of storytelling, we started to encourage people to just have AI create the decks. May even start just giving them the decks and just focus on the presentation. But what I've noticed, the decks look way more polished than they did in half an hour when we first started this, so we give them 40 minutes and halftime they're presenting in Excel, they barely got a deck together. Now they come with these fancy decks and sometimes the fair is they just read them. But the other thing I've noticed is AI has a tendency to want to be verbose. We all know that. And it thinks you need three different levels of titles on every slide and then words in 20 different boxes, and it puts them all in individual boxes. It doesn't use the typical PowerPoint boxes, and then it always has a couple lines of footnotes at the bottom.
(12:46):
I'm like, "What happened to brevity? Nobody wants to read... I mean, it looks great, and often it's right on point about the message and the numbers are right, but I'm like," We're outsourcing the aesthetic and the design, and it doesn't understand that human connection. Yes, it knows the rules of, Hey, you shouldn't have more than three things on all things that have been... But it doesn't follow those because you give it this message and it's taking probability saying, "Okay, to cover what you've told me, I got to put all that on this one page." And it doesn't have that discernment without a lot of back and forth to say, "What's the core message? How do I just give that to a way a human will take it and I can influence a decision?" Do you know what I'm talking about? That's almost like me with decks.
(13:37):
And I use it all the time, don't get me wrong, but I try to control that as much as I can.
Glenn Hopper (13:42):
Yeah. And like everyone else, well, I'm colorblind and I'm not a good visual person, and I love data and I love data
Paul Barnhurst (13:50):
Visuals. That
Glenn Hopper (13:51):
Explains a lot, Glenn. But my brain does not work that way. And so being able, starting going back to Gamma, which has now been... I mean, they're still around, but their Claude code is kicking their butt or Claude design is kicking their butt. And ChatGPT does a nice job too. It's funny, they have their default sort of designs they do, but I can now put together really nice looking decks, but I'm doing an apparel company I'm working with. I'm doing the final management readout with them and it put together a beautiful deck, but just like you, I have to fight with it and say, "I mean, here's the content. Just say what I'm saying. Stop adding more words or changing words or rephrasing." And it doesn't want to do that. And I think that that's a feature, that's not the AI itself. They're going to get there, I think.
(14:42):
And again, I keep looking for Microsoft to Copilot in PowerPoint would be the perfect way to do this. Just give me what I used to give when I had designers that I could hand stuff over to. Let me just throw some bullet points in a deck, send it off to you, make it pretty, don't change my words and go with that. I mean, I think that it's going to be just a feature and it's going to get better. But yeah, it's hard not to, especially when you're doing a million presentations, not to just say, "Well, I'll let AI handle this." But then if you don't go through and read it and use your judgment, it's going to say weird stuff in there and just phrase things weird.
Paul Barnhurst (15:18):
Okay. Yeah. We said we're going to do a recap. We're 15 minutes in and we haven't started our recap, so let's spend 15 minutes on that. And then I'm going to surprise you with the little game I've started playing on one of my shows we're going to do here at the end since we don't have our typical guests where we used to do our AI questions. I will do AI with the surprise thing though, because I think I can pull it up real quick and pull that off even while we're chatting. So we'll see if I can do two things at once through this episode. So recap, let's go back. We've now been doing the show for about two and a half years. I think when it first started, we brought on a lot of guests that were early adopters of AI. I remember we brought on David 14.
(15:59):
We brought on the guy from New York University, NY.
Glenn Hopper (16:03):
Connor Brennan,
Paul Barnhurst (16:03):
Yeah, head of strategy. Yes. We brought on people that had been in machine learning or really technical. We brought on Nicholas Boucher. We brought on Christian Martinez. Then we're kind of like, okay, we've gone through that list. What do we do now? And it kind of turned into, probably for a good year there, we were interviewing cutting edge AI software vendors. We had a long stretch of those. We did AR tools, we did billing tools, we had tabs, we had sequence. Congratulations to them. They were just sold to Intuit. A couple weeks ago, if you remember Ria Grover, we did what? Seven or eight ERP tools. We covered for a while there the gamut. In fact, we kept getting more. I remember one time you're like, "Please, can we just have something other than an ERP tool? I'm done." That was your note back to me like, "No." And I was kind of feeling the same way, so I was okay with that.
(16:50):
But we had on Rillet, which is now a billion dollar company. We had their CEO on, so you can listen to that episode. His name's slipping my mind right now. I can picture him, but I can't think of his name.
Glenn Hopper (17:01):
Nicholas.
Paul Barnhurst (17:02):
Nicolas Kopp. Thank you.
Glenn Hopper (17:03):
Yeah.
Paul Barnhurst (17:04):
We had Campfire, John Glasgow. We had Santiago from Dual Entry. We had Digits, CEO, the guy that was on the social media show, the famous... I can't even think of the name now, social network. We had him on Digits.
Glenn Hopper (17:19):
Siqi Chen or no.
Paul Barnhurst (17:21):
Siqi is Runway. We had Runway on the FP&A side. We had very first episode we ever did was the founder of Puzzle.
Glenn Hopper (17:30):
That's right. Yeah.
Paul Barnhurst (17:31):
Which has now switched. We had Basis, the accounting tool. So we had a few of those early, and then we went through a long stretch of that and learned a lot of just how AI is changing the way tools work. The OneStream CEO, Farseer, some small companies, PRP billing. And then we had some on cybersecurity and managing it and that kind of area. And so I think it was really interesting. And then we kind of pivoted. What we found when you and I talked and we looked at the numbers is the episodes that generally did best were the two of us. So we pivoted the last few months to sharing our thoughts. We covered what we thought about Anthropic's S-1, how to think about budgeting and forecasting. We went through the benchmark survey I did. We did your book, The AI-Ready CFO. We talked a little bit about how to budget and forecast it, how to deal with fatigue.
(18:26):
And then we had the one episode I thought was really fascinating. We just talked about the rate of change. If you remember, you did all your artifacts and the cost has come down 60 times from three and a half years ago on what a model cost for a million tokens. And it's like the stuff was three, four months announcements when they started. Now it's like hours sometimes between the tools. And so I'd love to get your thoughts as you look back and we talk about that journey, maybe a little bit thoughts on the software side, what people should take away, any memories or things that stuck out to you from guests, anywhere you want to go there.
Glenn Hopper (18:59):
Yeah, I mean, I think if you look at the arc of what we've done with the show so far, early adopters were really more the engineer types or just the tech evangelists or whatever. And when we saw the potential around AI, the tools were amazing to us, but truthfully, we were out there on the leading edge and companies didn't know what to do with it yet. So it was everybody wanted to hear about the potential. So we're talking to the people who do all that. And then we got through a really good group of smart people who were doing interesting things with AI. And then companies still weren't really ready to pull the trigger on building any sort of custom work or anything, but they were ready to hear from the software people who understood this stuff, "What are you doing with AI?" And then I think the timing of all the AI first ERPs that came out, it made sense and it was great to talk to the SaaS companies and see how they're using AI.
(20:07):
And then that actually, if you think about it from a, I'm a finance or accounting professional, I'm not a machine learning engineer or a developer, I'm going to trust the AI in the platform that I'm using more than I'm going to trust Rita in accounting to go build her own reconciliation bot or whatever.
Paul Barnhurst (20:27):
Ever feel like you go to market teams and finance speak different languages? This misalignment is a breeding ground for failure in pairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet QFlow.ai, the strategic finance platform purpose-built to solve the toughest part of planning and analysis of B2B revenue. QFlow quickly integrates key data from your go-to-market stack and accounting platform, then handles all the data prep and normalization. Under the hood, it automatically assembles your go-to-market stats, makes segmented scenario planning a breeze, and closes the planning loop. Create airtight alignment, improve decision latency, and ensure accountability across the team.
(21:35):
We had a big talk on the buy versus build because all of a sudden it came out you could vibe code anything and it's like, well, should you? Can I say something on that? I think you'll find this fascinating. In fact, I should have pulled it up. I got lit up on TikTok for one of our videos that they did. We do shorts and I basically said nobody should vibe code, remember we said a major application. Oh, the knives came out. People were telling me, "I'm vibe coding our ERP for our healthcare company right now."
Glenn Hopper (22:00):
I'm ready for a TikTok fight, Paul. I mean, if it's an engineer or a group of engineers or they have the right people behind, great. But just -
Paul Barnhurst (22:07):
One person came back and said they were an accountant doing it. Another said they had got rid of their HubSpot. They coded. Now, many of them were engineers. And I said, "Look, I'm talking in a finance situation about finance people who don't have development skills." And then a few of them backed off, but a few of them still fought me and said, "A lot of them said," This take isn't going to age well. You guys are out to lunch. "I'm like," I'll fight you on the core tools for somebody that doesn't understand development. If they have a development background or they're doing the prototyping and bringing in AI, which as some people said, I mean, bringing in IT, fine.
Glenn Hopper (22:43):
But
Paul Barnhurst (22:43):
There were several
Glenn Hopper (22:44):
That
Paul Barnhurst (22:44):
I got beat up pretty good.
Glenn Hopper (22:46):
But the hubris to think that your controller is going to just vibe code an ERP, it's like, come on, man.
Paul Barnhurst (22:53):
I'm with you. Like I said, I almost said, "I need to get Glenn on TikTok so he could take some of this abuse." All right, so back to where you were kind of going, sorry, I had to share that. I thought our audience would find that kind of hubris, but that was probably one of, if not the most popular short we had from this show, was that one because everybody was disagreeing with me.
Glenn Hopper (23:19):
That's funny. I mean, I get it. And if I were an acting CFO right now, I would have some kind of sandbox or lab set up. I would have clear guardrails, data rules, ways that people could do it. And I would say, "Yes, this is your imaginarium. Go over here, build cool stuff, do whatever you want. Nothing's going into production until IT, till development, till security. We need actual engineers looking at this because this is actually talking about Lauren Pearl, this is what we were talking about, cybersecurity and data security around vibe coded apps. If you've never done a pull request in GitHub before, you don't just automatic... Yeah, the AI can go do it and you can say," Make this secure. "But it could be something dumb like you didn't realize you had to make your GitHub repository private, so you're just publicly just dumping or maybe you're putting your API keys out in the open and all.
(24:21):
I mean, people, if you don't know and you look at code, it looks like...
Paul Barnhurst (24:28):
And I'll admit early on, one of the companies that work with marketing agency, they gave me an API key. And by mistake, I shared it not thinking with Claude. Then I accidentally shared it later when I copied something I didn't mean to. It's like," You shared this with me three times now. It's clear my words, you're being stupid. "I had to go get a new one at that point, and I did. They gave me a new one and they were good with it. There wasn't a big risk there, but it's just a reminder when you're not used to working with these kind of things and you don't think for a second, it's easy to expose something like that. It was an eye-opening learning for me. Fortunately, there was no risk to it. I mean, okay, somebody saw my podcast data. The API key was just my cables for my podcast stuff, so pretty low risk, and they switched it within a day to a different API key.
Glenn Hopper (25:17):
I don't know. Don't you think though, I feel like we've had our finger on the pulse of what's going on and at times getting electrocuted by just trying to keep up with all this. But I think if you look at the arc that we've taken with the show, it's sort of reflective of what's happening in the broader environment around AI.
Paul Barnhurst (25:35):
Definitely. I mean, I think a year ago we had a good period where you were ranting every episode about agentic workflows because we weren't there yet and now you feel like we're there. You've been really good at separating, I jokingly say ranting, but separating the marketing hype from where we really are because there's been a ton of marketing hype, and we've seen it with some of the guests we've had. Some have been a little more marketing than other, and there's times I know both of us kind of like, "Okay, maybe you can do that, but it's a little over promise." But I agree. I think we've been pretty good on the pulse. I think my favorite comment would've been February, "The hottest take we've had yet. We just had elections, so we have to ask how this one's aged in your opinion, if you ever think will come true." If you remember Ali, the CFO at TABS, I think it's Ali Hussain, if I remember right, he said the agents would vote one day.
(26:30):
Do you remember that?
Glenn Hopper (26:32):
I forgot about that. Yeah.
Paul Barnhurst (26:34):
Doesn't that have to be the hottest take we've had on the show?
Glenn Hopper (26:37):
Yeah, yeah, that was good. I forgot because the question was encouraged them to -
Paul Barnhurst (26:41):
We'd asked him for a hot take or something, and we were both like... And he goes, "No, what happened one day?" And we're like, "Okay, well, I guess we'll see."
Glenn Hopper (26:52):
Yeah, I'm not ready for that.
Paul Barnhurst (26:55):
I don't even trust humans to vote. I even trust in AI. I see us making bad choices all the time, at least in my opinion. I mean, half the people think my choice is a bad choice, so who's right? You guys can decide. We're not a politics show.
Glenn Hopper (27:09):
Yeah, maybe that's our next podcast is sports and politics, and we're just going to argue all day. We're going to...
Paul Barnhurst (27:15):
I think we should do sports, politics, and AI, and we could bring in AI to help us argue some of these issues. It could be the voice of reason. We could record it and have it give thoughts, and we could have some pretty good fun. We could mix the show up.
Glenn Hopper (27:29):
We could have our third host could just be a bot.
Paul Barnhurst (27:32):
We could be a digital avatar of one of us. It could be a bot at this point. We could use... That would be fascinating. Should we do an image of Will Smith as our third bot? We'd probably get sued, but...
Glenn Hopper (27:44):
Yeah, if we do Will Smith, I guess he'd have to slap us every now and then if we messed up.
Paul Barnhurst (27:51):
Whoever listens to this episode is going to be like, "What?"
Glenn Hopper (27:55):
What were they smoking? What is going on?
Paul Barnhurst (27:59):
I mean, I think as I look back, the biggest thing is I think we've been a show on the cutting edge. I think we've told people a lot of good advice, especially you, the AI-ready CFO, your different books and how you're using it, and just helping people realize that at the end of the day, with the technology, with all of this, take a deep breath, don't let yourself get over anxious, and just dig in and accept that you have to learn it. I think we said that three years ago when we started, that this is going to be something people have to learn, and that's only become more true, not less. I mean, not all of our takes and all of our opinions have been right, but I think we've generally been at the edge and given good advice to people that's mostly held up pretty well and then pretty sound.
(28:46):
At least I'd like to think so, at least you have. I mean, I mostly am comic relief, but...
Glenn Hopper (28:51):
Yeah, it's been fun. And I think when we take our break for strategy at the end of the year, we need to figure out what does 2027 look like? What is the area of focus? Where are people going to want info and guidance and guests and all that?
Paul Barnhurst (29:04):
So I'll put that out there to our audience. What would you like from 2027? Glenn and I are trying to figure out what the future of, no pun intended, what the future of future finance is. So we'd love to hear from you, but before we sign off and wish everybody an early happy holidays, by the time you get this, it'll be about a week before Thanksgiving when this is released, I have a surprise section for you. So as everybody remembers, when we first had guests on, we'd have AI create questions, 25 questions, and they'd randomly pick one. All right, so I have a list of 11 words that AI picked. You have one minute to argue if the word is overrated or underrated, since you go on a rant if you want. And how it's going to work, we got two options. I can ask AI to pick the number, your old method, "Hey, which one of these would work best for Glenn?" Or you can randomly pick a number between one and 11.
Glenn Hopper (29:57):
At this point, I'm trusting the bot more than myself. Let's see what AI thinks.
Paul Barnhurst (30:01):
All right. How should I write this chat? We all know prompts are critical, right?
Glenn Hopper (30:05):
Yeah.
Paul Barnhurst (30:06):
Because we've all seen on LinkedIn, everybody has shared their stupid prompt packages. Should I just say what client?
Glenn Hopper (30:12):
I'm talking to a nerd who thinks he understands AI and finance. Which one of these should I throw at him?
Paul Barnhurst (30:21):
I'm writing it exactly you said I have. I am talking to a nerd who thinks, if I can spell that right, he understands AI and finance. Which number should I give him? See what it comes up with? It mulled for a while. It said I should give you four. Oh, and AI gave me the reason I should give it to you and that it's a trap. So let's see here.
Glenn Hopper (30:45):
Just walking just happily, gleefully into your trap, Paul. All right, let's go.
Paul Barnhurst (30:49):
All right. Agent.
Glenn Hopper (30:53):
You set this up.
Paul Barnhurst (30:54):
No, I promise I didn't actually have agentic workflows on the list. I was hoping it would pick that, but it picked four, which is agent. You got one minute. Overrated or underrated your argument, I cut you off at a minute.
Glenn Hopper (31:07):
Overused and underrated. True agents are here. They're only going to get better. The meter index that measures how long AI can go out and work is, we're seeing kind of a Moore's Law with agent speeds enhancing. The reason that it is underrated is nobody knows what an agent is because everybody's been calling everything an agent for the last couple years.
Paul Barnhurst (31:33):
All right, good job. You did it in under a minute. Here's what it said. Someone who thinks he knows AI will jump straight to software agents, tool calls and autonomy. That's the trap. The finance, meaning the principal agent problem is older and arguably more useful for deciding who's accountable when an agent books the wrong accord. If he only knows one side, it shows fast. If he knows both, you get the best conversation on the list.
Glenn Hopper (32:02):
The principal agent problem, good Lord.
Paul Barnhurst (32:05):
And then it told me to avoid five.
Glenn Hopper (32:07):
Which was?
Paul Barnhurst (32:08):
We're going to give you a couple more, so I'll tell you five. We'll keep five out of the list. It said avoid five, overfitting. That's home turf for a technical person and he'll just lecture. Great. So this time you can pick a number. Four and five are off the list. Pick a number this time. Three. Between one and 11.
Glenn Hopper (32:25):
All right, let's go three.
Paul Barnhurst (32:27):
Three. Variance. I don't really like that one. Pick a different one. I don't like that one.
Glenn Hopper (32:31):
Yeah, let's go 11 then.
Paul Barnhurst (32:33):
Model.
Glenn Hopper (32:34):
Model. I mean, model is air. I think about the model. I'm working on my fourth book right now on building agent harnesses for model.
Paul Barnhurst (32:42):
You got to give the word first. Is it overrated or underrated? Then you can describe. You got to pick one.
Glenn Hopper (32:47):
Over.
Paul Barnhurst (32:49):
I knew you'd say that. Go ahead. You got your minute now.
Glenn Hopper (32:52):
Yeah. Okay. So the reason model is over is because it really is. We've been talking about these models for four years now and it's certainly an arms race still. The model is the base and we've said for years, even if AI doesn't get any better than it is right now, we will keep finding uses out of it. So the model today is important and it powers all the work that we do, but it's behind what we're figuring out with what we can do with harnesses and by giving the models goals and everything that we put around them. So I hate to say over on model because that's the foundation of this whole thing, but I do this with clients. If something happens and a model changes, I'll swap out ChatGPT for Claude and vice versa, and it really isn't making a difference now. So that's why I'm going over if I have to pick one.
Paul Barnhurst (33:48):
I'm going to pick the next two because I just want to give you the opportunity because I know where you'll go with both these, I think. So we'll see if I'm right. I'm going to give you number nine, AI-native.
Glenn Hopper (34:00):
Oh, I don't know, man. Over, but if you think about it, so over is my answer, but if you think about it, AI-native, there is an advantage to not lugging all of your historical. If you're in software and you've got -
Paul Barnhurst (34:17):
There's a legit reason to be AI-native.
Glenn Hopper (34:20):
Yeah. And you're not going to rebuild the battleship with that in mind. I think if you are new enough to be building in a post-AI world, you're going to have some advantages. But we always talk about data as the foundation, and I think the systems and software that were built before AI hopefully have a really strong data foundation. Nobody wanted to spend any money on data governance and on data warehouses and all that for years. This is kind of the clarion call of it. But yeah, it would be nice to start with a clean slate, but the non-AI-native, I think there's room for them to adapt.
Paul Barnhurst (35:01):
Okay, fair enough. I'm going to ask, I'll go one more, and then if you have anything you want to throw at me before we wrap up, if anything comes to mind, you can. AI CFO.
Glenn Hopper (35:13):
Well, my book is The AI-Ready CFO, which it's an adjective to describe the CFO's readiness around ability to use AI. AI CFO, I will say is over because I don't know what that means. You are the CFO. That is your job. The AI CFO would be a CFO who uses AI. If it's 2026 going into 2027 and you're not using AI, then you're the no CFO. I don't know. I mean, I don't know what an AI CFO is. I know what a chief financial officer is. I know what a chief data officer is. I know what a chief AI officer is, and I know all those people can use AI, but I don't know what an AI CFO.
Paul Barnhurst (36:04):
I hear you. I'm the same. All right, so you got anything for me before we wrap up?
Glenn Hopper (36:10):
Yeah, I've got one for you and it's asked all the time and we see the models keep getting better and better. Is Excel dead or dying? Why or why not?
Paul Barnhurst (36:20):
No, it's not dying. I've asked twice in the last year in polls. I asked three questions. People use Excel more, less, or the same. 80% say the same or more. And Claude, they spend a ton of time developing for Excel. Somebody argued, no, they develop all that stuff so they could basically release their own programming tools. Not seeing that yet for modeling from Claude. Now they might do that and they might be thinking about that in addition to Excel, but Excel, ChatGPT, Excel. How much has Microsoft spent on Excel and AI? The reality is the spreadsheet is fundamental to the way we think at this point and work. The spreadsheet ledger spreadsheet that came from way before computers and then they deported it to computers and being able to look at something in rows and columns, yes, you could do multidimensional stuff, you can do that spreadsheet, you can do that with other tools, but that ability to look at it, that transparency, that tracing, that prototyping, that's not going away.
(37:23):
Is the way we work with Excel changing? If it hasn't, you're behind. Does that mean you should turn everything over to AI? No. There's a medium here in working in Excel. You still need to know what you're doing. You still got to be able to trace things. As one modeler put it, we trade in trust. I talk on the financial modeler's corner to some of the best modelers in the world. Just this last week, I interviewed the only woman to ever win the financial modeling World Cup. And she talked very much about, "Hey, we are doing billion, hundred million, billion dollar transactions. There can be no mistakes." Now you could disagree with the final conclusion, but we got to make sure these models are clean and they're free of material mistakes. And so we're not ready to just turn that over to AI even though we're using it in areas.
(38:12):
And so they're still using Excel. So I don't think it's going anywhere at this point. I don't think spreadsheets go away, but yes, it's fundamentally changing. Your take, agree or disagree?
Glenn Hopper (38:21):
So this goes right along with what we were saying about vibe coding. Engineers vibe coding, awesome because you know what you're looking at, you are going in and reading the code and you can just 10x, 100x, whatever your output by letting AI read the code. And then you're going through and you're QCing it. An analyst who is never really... I mean, there wouldn't really be an analyst who didn't know Excel, but an analyst who -
Paul Barnhurst (38:48):
I don't know. I've seen a few that are pretty close.
Glenn Hopper (38:50):
Okay, fair. Yeah. But an analyst who doesn't know Excel and is letting a model do all of its work and using LET and LAMBDA or crazy functions that AI can build, when you present numbers and have to explain where that came from, you're going to be in a bit of a problem. So I think there still is a reason to have that expertise and to have the tools and to have the judgment and know what to look for and to think an Excel expert with AI, what they can do versus give my mom, who is pretty good at using ChatGPT for vacation planning, give her the same assignment, you're not going to get the same results. So it's the same thing with coding and with any sort of subject matter expert in whatever area it is. If you are a professional race car driver and you drive a Ferrari versus you hand it to a 16-year-old, you get a different result from the car.
Paul Barnhurst (39:50):
No, exactly. I mean, I think you nailed it. Get one last thought. I'm going to keep you to about two minutes here and I know we need to end, but when you mentioned that, I heard someone recently say, "Hey, the coding problem, so to speak, AI's pretty much solve coding in a sense they can do a really good job now. You need an expert. I don't think it's solved coding for everybody." Whereas modeling, I don't think it's as far along as coding because coding is very structured. A spreadsheet, C12 can be 5,000 different things in 4,900 times you look at it because somebody tried to make it three things in a couple of those. And so it doesn't have the structure that modeling has. Would you agree with that? Because I've had some people kind of disagree. Modeling in the spreadsheet is a little harder for AI than coding.
Glenn Hopper (40:39):
Yeah, I think so. And it's funny to me because early on when we were one of my last full-time CFO roles, we were doing a lot of automation and forecasting with no planning tools or whatever. We were taking basically our Excel modeling that we did and trying to turn it into code and put it on a dashboard. So my brain for years has been trying to switch more away from modeling in Excel to using whether it's machine learning or just straight whatever modeling tool.
Paul Barnhurst (41:07):
Python or whatever.
Glenn Hopper (41:08):
Yeah. So yeah, interesting.
Paul Barnhurst (41:11):
All right. Well, I think that's probably a good place to end for the year. We'll wish everybody a happy holiday. We'll remind you. Hey, reach out to us. Let us know what you want the show to look like next year. We're trying to evolve, not transform, because I don't believe in finance transformations, but we're trying to evolve the show just as AI moves along. So let us hear your ideas. Thanks for joining us. Glenn, have a great holiday season. Enjoy your December off working in the business. Remember, not on the business. All right. Well, thank you everyone. Last words, Glenn, you get to send us out.
Glenn Hopper (41:45):
Great season, Paul. Looking forward to a break and looking forward to where we go from here. And thank you to all of our listeners and great holidays coming up for everyone.
Paul Barnhurst (41:53):
Signing off, and if you're an AI CFO, congratulations.All righty. Good night. Thanks for listening to the Future Finance Show. And thanks to our sponsor, QFlow.ai. If you enjoyed this episode, please leave a rating and review on your podcast platform of choice, and may your robot overlords be with you.