How to Choose Finance Software Without Waiting for AI to Settle with Glenn Hopper
In this episode of Future Finance, Paul Barnhurst and Glenn Hopper discuss how AI is changing the finance technology stack and what it means for CFOs deciding whether to buy new software, build custom solutions, or improve their existing systems. They explore whether AI could disrupt traditional SaaS platforms, where agentic AI is delivering value, and why core financial systems remain essential.
In this episode, you will discover:
Why AI is changing finance technology without replacing core systems.
What the Finance Stack Survey reveals about AI adoption.
The difference between agentic AI and traditional automation.
When finance teams should buy software or build custom solutions.
Why data quality, integration, and security matter for AI success.
Paul and Glenn emphasize that finance leaders should focus on solving real business problems rather than chasing new technology. The key is choosing reliable tools, improving data connectivity, and using AI where it delivers meaningful 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 alikea.
In Today's Episode:
[00:00] - Trailer
[04:21] - Is SaaS at Risk?
[10:27] - Buy vs. Build: Filling the Gaps
[11:09] - Finance Stack Survey Findings
[14:00] - What Is Agentic AI?
[18:20] - AI Adoption and Software Satisfaction
[23:36] - Data and Integration Challenges
[29:41] - Custom AI Solutions
[36:17] - The Future Finance Stack
[42:51] - Closing Thoughts
Full Show Transcript:
Host: Paul Barnhurst (00:59):
Welcome to another episode of Future Finance. I'm your host, Paul Barnhurst, and I have with me the sultry sounds of Glenn Hopper. Glenn, how are you doing today?
Co-host: Glenn Hopper (01:42):
I'm doing fine, Paul.
Host: Paul Barnhurst (01:44):
Could you do
Co-host: Glenn Hopper (01:45):
A little more
Host: Paul Barnhurst (01:45):
Like Casey Cam Top 40?
Co-host: Glenn Hopper (01:51):
The letter this week is - Oh, heck yeah, every week.
Host: Paul Barnhurst (01:54):
Yeah, Saturday morning, listening to what are the top 40 songs? Now the real question is, did you have your tape recorder by it and were you capturing the song?
Co-host: Glenn Hopper (02:04):
Oh, you had to, and it always drove me crazy when Casey or whatever announcer would talk over the intro to the song. It's like, no, I'm trying to get a clean version here. And then you had to have the dual cassette deck so you could make your own mix tapes later. You're just recording them in order. But if you wanted to be a producer, you had to be able to make mix tape for your girl.
Host: Paul Barnhurst (02:24):
How many people in our audience do you think even understand what we're talking about?
Co-host: Glenn Hopper (02:28):
Yeah, what a cassette tape is, who Casey Casem is.
Host: Paul Barnhurst (02:32):
Oh, do you ever sometimes feel like Back to the Future?
Co-host: Glenn Hopper (02:35):
Yes. And I still want to go back in time with my sports almanack and go earn my riches that way. That was Back to the Future, too, though, wasn't it? Or was it part two when he had the almanack? Yeah, yeah.
Host: Paul Barnhurst (02:50):
Yeah, I think it was two that Biff went back and made a bunch of money.
Co-host: Glenn Hopper (02:54):
That's right. Yep, yep.
Host: Paul Barnhurst (02:55):
Yeah, yeah, yeah. Yeah, that one wasn't very good of the three. One and three were the best, but we digress. This is a finance podcast, but they made a bunch of money. We can make a finance relationship. We can tie it into corporate finance, I'm sure Biff and his empire. No. So thank you for joining us. Today, Glenn and I, what we wanted to talk about is something I think that will be practical and helpful. So we want to dig into what does all this AI mean for the finance tech stack. How should you, the end user, be thinking about this? If you think of the typical finance tech stack, it's going to consist of every company's going to have this general ledger, whether you want to call it an ERP accounting software or whatever. Then everyone's going to have a spreadsheet in there.
(03:44):
From there, a lot of it depends on company size. Do you have a treasury tool? Do you have an FP&A tool? You're going to have a data warehouse, you're going to have BI, and then you're going to have all those operational metrics as well. And there's other tools that you often need or have, whether it's close management software, reconciliation software, list goes on and on. And I feel like everybody's kind of struggling to, "Hey, should I overhaul my tech stack? Should I wait? Should I buy something? How do I think about data? How do I think about getting the most out of AI? Is SaaS dying?" There's so much uncertainty. I mean, if you were a CFO right now, would you probably have a little bit of pause?
Co-host: Glenn Hopper (04:27):
Yes and no. You mentioned SaaS dying and we're at a phase right now. Satya Nadella said a couple years ago, this is a problem for SaaS because if SaaS is really just an interface sitting over a bunch of tables, then an AI can read the tables directly, why do you need that interface? You could just custom pull the data however you want. And also we saw evidence of it this year. Was it when Opus 5.1 came out or maybe it was 5.0 and all the SaaS companies stock just went in the tank because they thought, oh, look what AI can do now. And then they've kind of recovered. But here's with clients right now, the first question that they always have, especially in the private equity space, they always, it's what software does this? They don't want to pay a consultant. They don't want a bespoke thing built out for their companies.
(05:20):
They just want to buy some software, put it across their portcos and use that. And there's an argument for that in that, okay, you've got a bunch of tables out there. Are you really going to build your own ERP? I mean, if you need to be audit ready and the defensible software I think is, or the defensible parts of software are having a system of record, being that the source of truth of GL. If you've just got all your GL entries just sitting in a Postgres table out there or something and it's like, ask the AI, it'll go get them for you. A
Host: Paul Barnhurst (05:56):
Database, probably a bad idea.
Co-host: Glenn Hopper (05:59):
Yeah. And also think about what all these software companies have done over the years. They have built in controls and auditability. They have deep integrations. Think about ERPs tying into banks, I guess is the first example that comes to mind. And there's all the things that, think about manufacturing, bill of materials and all the really detailed stuff, three-way invoice matching, all the things that they've built in. Is Kenny in accounting really going to build Vibecode his own three-way match app that's going to be better than what you have? I don't know. I mean, maybe in a couple years if agents keep getting better, but I don't know. If I'm the CFO right now, especially if I'm publicly traded or need to be audit ready or I'm pre - IPO trying to get ready for that, I'm not leaving it all in the hands of homegrown whatever tools yet.
Host: Paul Barnhurst (06:56):
I fully agree. So I think we're both on the same page that you still need a tech stack and your core systems need to be purchased. But I think CFOs, many are still kind of struggling. Hey, does that mean I need to go with a modern tool? Do I go with establish? There used to be the phrase, and we've talked about this before, nobody got fired for selecting Oracle or IBM or insert 50-year-old established finance company here, whatever name you want to use. That used to often be the joke. I think that for many companies that ship is sell, not all. There's still some that have that philosophy. And so I just feel like that opens up more opportunities for companies and they're struggling with, do I do something now? Do I wait? Just with the speed of change. I don't see many CFOs, and maybe I'm wrong, you tell me, but I don't know many leaders that are really, or almost any that are saying, I can just vibe code my entire finance stack.
Co-host: Glenn Hopper (07:56):
Yeah. But I guess I think the question is, does this still need to be software? And I'm not going to bring up any specific examples, but there's a lot of point solutions out there, startups that have spun up in the last few years that are, I mean, if you think about what they're doing, maybe they're just a thin interface over a workflow that if you don't have a good moat as a SaaS company, then not having the moat is going to be even more dangerous when somebody can put together a little agent that does what you did. So a thin startup that just does a single task, that's probably a problem. And I was looking at, I think about all the software out there that even the lower end, maybe it's like a 25 or $30,000 a year subscription for a smaller software package, something like that.
(08:51):
If I could think, well, if I could save that 25 grand and we just build it and we get the same thing that has the audit trail and whatever I need from it, I would think about that a little bit. But as far as core systems go, CRM, ERP, ERM, I'm going to be a little hesitant before I'm ripping those out and letting the bots just run over a bunch of databases.
Host: Paul Barnhurst (09:10):
And I agree with you. Simple workflows for a point solution, especially ones that maybe they never quite met your needs, you can customise a little bit. I definitely think there's an argument to say, "Hey, you could build it, workflow type things." And a lot of these tools are now starting to offer areas where you can build it within the tool. We're starting to see more and more of that where they're secure. So I think I'll get to my report here in a minute, but I think the first message is to finance is look, your core tools are still going to be, today you need to go out and buy them, don't build them. Certain workflows, manual tasks, some one-off things, consider it. You have to make the decision of what's best for you. I don't think it's a one size fits all. Things that you were doing that are manual, you were doing in Excel, look to see if there's ways you can either automate that in Excel with agents and deterministic solutions or move it over to some kind of AI agent.
(10:08):
Is that a kind of a fair, would you say, across the spectrum?
Co-host: Glenn Hopper (10:11):
Yeah. And what I'm seeing in the consulting world is it's where people realise now that you can come up with either an agentic solution or a software solution a lot more affordably than you used to be able to, whether that's bringing in a consultant to do it or having somebody in-house build it. It's those gaps between, okay, our CRM has this information and does these things, our ERP does this, but here in between the two, we have this gap of something that we're trying to get done where maybe there was another software subscription you could get that would handle it and you think, well, those are my systems of record. They're fine. They're going to stay like they are. I just need to fill that gap. I think that's where if I were in that space with the software tool right now, I'd be a little bit worried.
Host: Paul Barnhurst (10:57):
That's where you could just say with a couple connectors, a workflow and a little bit of patience, you could probably build something that gives you what you need.
Co-host: Glenn Hopper (11:07):
Yeah.
Host: Paul Barnhurst (11:09):
No, I agree. So I think that kind of covers the levels at least a little bit of how to think about it. Next, I want to go a little bit about the survey I just released today. I called it the finance stack survey, and then I just want to get some thoughts on a few different categories from you. So let me start with the report. I'm going to bring it up here on the screen. So if you're watching video, you can see it. You can go to my website, the FP&A guy and download it. So here were the key takeaways, and I'm going to make this a little bit not smaller. If I hit the right button, I'm supposed to make it bigger. So I'm just going to read, these were the five highlights, then we'll run through with a few things. So we interviewed 176 leaders, roughly 40% were CFOs, majority of the rest were FP&A, and then 10% of other.
(11:56):
So call it 50, 40, 10. So the first one is we asked how many believed AI was changing the finance tech stack. So we asked, "Hey, has it already changed it? It's changing it, but it's kind of too early or no, no real difference." 65% basically said, "Yeah, AI is changing the finance tech stack. It's already happening." Then you had a percentage that said, "It's too early," and I think only 3% said no. So those are just your not believers, right? There's always a few that don't believe and whatever. That's like the example of your guy that still did his ledgers when he had a computer.
(12:34):
We're never going to move him to the 97%, so to speak. But what we found interesting, but only 14% run Agentic AI in their workflows. One of the biggest predictors we found, and some people have made some really good points, and I'll share that in a minute, is adoption of Agentic AI, they almost always had tools that offered Agentic AI. So we asked if their tools offered it. Their FP&A tool almost always offered it if they've adopted it or piloted it. And so some people brought up a question, is it they have good tools because they have good data? And really those that are adopting have good data, not that their software's providing Agentic AI, or is it most people that are adopting it because their software providers provide it? Do you know what I'm saying? It's kind of an interesting discussion on LinkedIn today.
(13:25):
Get your thoughts there.
Co-host: Glenn Hopper (13:26):
Yeah. Well, I mean, I'm a little biassed here. What do you think reflective of?
Host: Paul Barnhurst (13:35):
I think it's a mix. My first thought was, hey, probably the quickest way for people to adopt and turn something on is if their software provides good stuff. But most of the companies that have good software, good planning tools usually have better data. So I feel like they go hand in hand. I don't know that the better data is the reason they have better tools. It's probably because they have to really accomplish things. I don't know.
Co-host: Glenn Hopper (14:00):
Yeah. Here's where, this is my old, I'm getting -
Host: Paul Barnhurst (14:02):
And you can disagree with me 100%, please.
Co-host: Glenn Hopper (14:04):
No, no. So this is why I can't get my head right with the question even. I'm not sure that people, I don't know, and this is where, maybe tell me if I'm wrong, I still don't think people understand generative AI that does something cool versus what's truly agentic AI. I
Host: Paul Barnhurst (14:21):
Knew you were going to go there.
Co-host: Glenn Hopper (14:22):
Because I was working with a team yesterday that they're a Copilot shop, and I don't really do a lot in Copilot. And I was talking to them about projects and about ways to set things up, and they were calling it an agent, which is what I guess in Microsoft in Copilot, they were asking if an agent and a project were the same thing. And I was like, no, unless Microsoft is calling something else an agent that's not actually an agent. So when I see these numbers, and it is still a little small on my screen, but give me the numbers again.
Host: Paul Barnhurst (14:53):
65% said AI is already changing finance work, but only 14% run agentic AI. So if I look at, let me go down and see if I can find, here's what we got. Where agentic AI adoption actually stands, 55% said they're not using it, 31% said they're piloting, 14% said it lived in workflows. And piloting or live with Agentic AI, 70% said their tools offered basically agents.
Co-host: Glenn Hopper (15:25):
What do you think those are?
Host: Paul Barnhurst (15:27):
What do you mean when you say what do I think those are?
Co-host: Glenn Hopper (15:29):
I'm going to have to let it go, Paul, but what is it actually to -
Host: Paul Barnhurst (15:33):
I think we asked about FP&A others, so I think some of these are budgeting agents, forecasting. Yeah. Some of those are nothing more than a fancy way of saying, "Hey, I built this tool that will give you answers related to certain topics." But I think more and more what I'm seeing from these EPM vendors, and so the way I think about it when they say agent, it's a tool that they can talk to and have it actually go do work within the system, set things up. It's performing tasks. Is it 100% autonomous? Probably not. In some cases, maybe. It's monitoring whatever. So is it agentic in the way you think about it? And probably not. It's more agents that are, but it's AI doing work is probably the way I think with limited human involvement, put it that way.
Co-host: Glenn Hopper (16:26):
Yeah, and that's fair. And I just think about this is an agentic planning tool and it's like, what's agentic? I mean, it's a dashboard and it's showing me a forecast and it's redoing a forecast and there's a lot of deterministic stuff. And there can be. In an agent harness, you can have plenty of code and deterministic parts too.
Host: Paul Barnhurst (16:46):
Oh yeah, yeah, yeah. And you should in finance, let's be honest. If your harness has nothing deterministic, you don't want that software. Hint, spoiler. If they haven't built some deterministic solutions, select a different vendor.
Co-host: Glenn Hopper (17:01):
Yeah. And I guess to me, if it's a drill down and another way to look at something and it's just another click on a dashboard, that doesn't really feel like an agent. Now, if I can come in and ask it a question, it has to go and pull this from the GL, this from whatever economic data and do this, redo the calculation and everything. And I would say in that case, who's really using Agentic? I guess that's my whole reason I got hung up on this is like you said, if it's in the software and they call it an agent, I'm not going to blame the user on that. But companies who are out there running their own agents, unless they're in the tech space and/or they're big enough that they have true AI engineers that are building these harnesses, I think a lot of it is just cooler stuff.
(17:47):
Some SaaS platforms have cooler AI chatbots and pieces to them than others, because I don't think outside of developers right now, and I know there's some sort of drag and drop, and you could have a pseudo-agentic workflow using an ADAN or make or Zapier, you'd have your workflow that's pretty deterministic. You drop an AI bot in there to do something with it, but that's not a true agent to me. So I'm sorry, I can't let it go on my semantic thing on this, but let's get back to the... Yeah.
Host: Paul Barnhurst (18:20):
So let's just talk more AI, a little bigger picture. I get it. Semantics, we're going to get you hung up. But those that have real AI running in their software are much more likely to believe it's changing their work. So they're seeing the benefit of AI, regardless of whether it's... We can wordsmith that the rest of the episode, but we can at least agree on that part. So what I found interesting is those who say they have Agentic AI live in their tools, whether it's really Agentic AI or not, it's AI. They're using it, they're using probably some agents. Their FP&A tooling, they rated it at 8.2 versus those who didn't rate it a 5.9 on satisfaction. So what's saying is those who are using AI, regardless of how we define that, are much happier. Does that surprise you? More satisfied?
Co-host: Glenn Hopper (19:06):
No. And again, from the consulting side, everything I build is bespoke and a special snowflake, but I have a rapidly growing multi-location business as one of my clients, and I built them a dashboard that was cool and great and fine, but it just kept getting more and more questions. And because of the nature of the work that I do, I thought, well, the first thing I did was just had AI go and just read the results and give, and I flagged it and said, "This is an AI opinion. It may contain errors, whatever." But it would look at the trends and analysis and it would say -
Host: Paul Barnhurst (19:44):
I have meetings.
Co-host: Glenn Hopper (19:47):
It would look at the trends and analysis and just give its sort of commentary on it. But I recently added, this is one of my proudest builds I've ever done. It's a chatbot on the dashboard that is tied in to their ERP. So yeah, we just have the KPIs on the front and I've also, I've guardrailed it, so it won't let them get into... Non-finance and accounting people don't need to get into things like intercompany eliminations and all that stuff. So I block all that from it. That's the thing with dashboards is you could drill down one, two, maybe three levels and get some more information, but when they can't get it from the dashboard, that leads to another ad hoc question that just gets in the queue for the FP&A team or whichever team it goes to. This chatbot, I don't hear anything from them anymore on it.
(20:37):
They ask the chatbot and it gives them the questions or it gives them the answers without me having to go do ad hoc reports and I've got it set up. It'll give them tables. They can output it in a little Excel file if they want to. It cites everything and they are loving it. They're understanding their financials more than they had because before they brought me in, they had a very small internal team and no dashboard, and now CEO and the chief of staff and all the people that need it can get these answers immediately. I put guardrails, it won't forecast and it won't try to do anything crazy that's beyond what it can do other than just the descriptive analytics of what's in there.
Host: Paul Barnhurst (21:18):
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. Q flow 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 air-tight alignment, improve decision latency, and ensure accountability across the team.
(22:25):
Yeah, and they're much happier, and that's a great example of what I'm talking about. That person probably rates it a nine and before they were a six or whatever, or a five. And so there's a huge difference because AI has allowed them to be meaningfully more productive, get answers in a shorter time, solve what some call that Monday morning problem. If you remember when we had N Wong on from summation, and then by next Monday, everybody had forgot it. Well, finance was rushing to find the answer and never found it. And so I think that's a great example of the happier adoption. So I'm not going to go through all five of these. Excel still runs FP&A, no surprise there. Claude clearly leads the LLM race, and if you're a small company, it was like 88% of small companies had Claude and only 10% had Copilot.
(23:15):
And if you're over 5,000, almost every company had Copilot. And so what we're clearly seeing is preferences Claude, size of company influences whether you can use Claude or not.
Co-host: Glenn Hopper (23:26):
Here's what I'd be curious about if Copilot users had a different opinion of the impact of AI than Claude.
Host: Paul Barnhurst (23:36):
There wasn't enough in the survey to really get down to that level of detail. If we could have got six, 700 people for the survey, I would've redone some of the questions. Hopefully for next year, we can get further on some of that because there's a few things like that. The other one, and this won't surprise you, then we'll dig a little deeper and then we'll bring it back around. But the real pain is data plumbing. So we did open-ended question with text answers about their biggest pain. Data appeared 60 times. Cost as their biggest pain appeared twice. I mean, so it tells us, is it surprising that data, data plumbing, data silos is still one of the biggest problems? That's why a lot of people were saying, "Well, is it really because their tools have AI or is it because they have good data is why their tools have AI, because they know they can get benefit." Where a lot of those who haven't looked for tools with AI is because their data sucks.
(24:26):
It's a mix. It's some of both, I'm sure. Yeah.
Co-host: Glenn Hopper (24:29):
What are you hearing on the data side? So think about, yeah, your data may be a mess, but if you're using the AI that's built into the platform you're using, think about, I may be at a company where my data is a mess, but I've got NetSuite next and it is confined to what my data is in NetSuite. Now, I know there's a lot of stuff that won't... Transaction level detail for e-commerce transactions or whatever, you're just going to have the rolled up journal entry and you can't drill down to that, so you'd have to go into another system and whether that's in Stripe or whatever it is, but I wonder how... And this is going to be different based on the size of the company too, but if you're just... I guess my point that I feel like I'm really rambling today, but my point is if you are using the AI that's built into your system -
Host: Paul Barnhurst (25:25):
You didn't see that. That's good. How about like this rambling?
Co-host: Glenn Hopper (25:30):
Yeah, I was staring off in a
Host: Paul Barnhurst (25:32):
Space. And he just kept talking, so it was classic. It will be a great... You have to watch that part on video, everybody, to appreciate it. Keep going.
Co-host: Glenn Hopper (25:41):
If you're just using the AI that's built into your software platform and you have to get something from your CRM, something from your ERP, something from your AP tool, the data that it's going to look at there is going to be pretty good because it's built into the tool, it knows where it is and all that. The problem is going to be when you're trying to get to tie it across the systems, and that's that gap that sort of consultants fill. But I'd like to know from the people that were surveyed in this, from their point of view, what was the problem with their data? Because how many of them even have a data warehouse versus -
Host: Paul Barnhurst (26:18):
We can go through and look at a couple of the answers when we get to the end. We shared a few here. Let me share a couple other things from the report. We'll do that, and then I want to bring it around to different kind of categories within this. So this is what I was highlighting earlier. If you look right here, Copilot, small company, bigger company, bigger company, bigger company, bigger company. Claude, contracts, company size, bureaucracy has limited clauded option. Wouldn't you take that as a takeaway there?
Co-host: Glenn Hopper (26:48):
Yeah. I wonder though, I mean, I wonder where ChatGPT enterprise would fall.
Host: Paul Barnhurst (26:55):
So ChatGPT, 38% said they had it, so they could select multiple. So it's going to be more than 100%. On average, it was almost two tools per company. We did count agent inside Excel, which kind of is double counting because hey, you have Copilot, but what we want to do is get an idea of how many were using AI in Excel, and it came out at 20%. My guess is because some said, "Well, I'm using it with Claude, so is that really an agent?" There's semantics here of how clear you have to be in interpretation and in a short survey, how much word can you add? So definitely some learning on how to build a survey, but that's what we got there. Let me share a few other... No surprise, Excel still runs FP&A. I think we'll skip there. Where they were most disappointed, this was interesting, close and consolidation is the biggest area that they had the most disappointment with.
Co-host: Glenn Hopper (27:44):
With the FP&A, not with AI. FP&A
Host: Paul Barnhurst (27:47):
Tooling. Yeah. We asked them to rate it on five areas, how satisfied they were from a one to 10. This is the percentage that said five or lower. So 38% were five or lower. Vendor relationship was next. Reporting, then overall tooling. They're okay on the forecasting side. Everybody complains about being able to model, but that's not necessarily... We didn't ask that specific question. And then requirements fit was where they were least dissatisfied in the sense of unhappy, the least percentage unhappy.
Co-host: Glenn Hopper (28:22):
There's a boatload of close and consolidation software out there. Why do you think dissatisfaction is so high with that one?
Host: Paul Barnhurst (28:28):
Well, because about 50% of the survey was still using Excel, so we'd probably have to split it out. I think that's part of it. And just given the size, it gets really hard as you start splitting it further and further to have any confidence.
Co-host: Glenn Hopper (28:42):
Fair. Yeah.
Host: Paul Barnhurst (28:43):
So I think that's part of it. I also think close just gets really complex. It can sometimes. And then here's, I think, was my main point. So this was 124 open text pain points that people could talk about. So this is how many times they mentioned a keyword. 60 said data, 20 times Excel or spreadsheet came up. 17 is integration, which goes back to data in many ways. Time, manual work, and then consolidation. And so these were some comments. Data, data, data is what one CFO said on their biggest pain point. Another said very disconnected and lots of manual steps. So I looked through it. It was a lot of manual, "Hey, we have a bunch of systems. They don't talk to each other, silos." Quality data came up only a few times. Missing or late data wasn't an issue in what they listed, but that's not surprising giving it was a tech stack.
(29:41):
So those are probably the highlights from the report. I mean, there's other areas we could go in it. A pretty big split between small and big companies. None of the companies over 5,000 who were surveyed said they were live with the Gentiq AI. Now, I get the wording, but what it shows is smaller companies in general were using AI more. So bureaucracy from everything I could see here, Claude, bureaucracy slows AI adoption. Not surprising. Like you mentioned, if you're getting ready for a public company, are you going to want to go out and add Numeric as your ERP that just launched this financial data platform and kind of implied that, hey, journal entry should go away and we should rethink the entire way we're doing accounting?
Co-host: Glenn Hopper (30:28):
Yeah. I'd like to explain that to the auditors.
Host: Paul Barnhurst (30:32):
And it's not to say that it's not a great tool and there's some really good points they made, but it's not a risk, a risk adverse company's going to make right now. A year from now, two years from now, somebody will, but not today. It's going to be a little while. They just barely launched. You
Co-host: Glenn Hopper (30:47):
Know what drives me crazy, and again, not to get hung up on the wording, but I've been doing a lot of work with, well, one of them in particular, a pre - IPO, but still very large company, PE backed, like 60 something people in their finance department. And they've been healthcare, so they have limitations on what they can do and they had to turn off a lot of things in Claude because of with Claude acceptable use. But
(31:12):
A lot of their audit concerns, we're piloting an agentic harness to do work for them. And what I'm trying to show them is if you have the controls and you've already got your documented controls, the same controls that your people use, that's what the agentic harness is going to use. So all those things you were talking about, the data complaints, all those gaps and pulling it from different places, that area is ripe for AI. And I don't mean point your chatbot at it. I mean setting up, having the MCP server, API connections, what you can get to the data, pull it in, whether you're pulling it into some kind of little sidecard data mart or whatever, transform the data, do whatever you need to it outside of the systems, saving the audit log all there. There's a wealth of opportunity in that. And I don't mean for a startup to come in and do anything.
(32:06):
I mean for companies to figure it out themselves as long as they have access to someone who can build these harnesses. And I think that that's where if you could start getting wins in that area, then your satisfaction talking about how much AI has improved your close process and your consolidation and whatever part of finance or accounting you're looking at, that's where you're going to start seeing real results. And I think now we're kind of out on the bleeding edge of that right now, but as fast as everything's moving, I feel like if we do this, you do that survey next year and we're talking about it, I think we're going to see a lot more companies doing those bespoke agent harnesses to solve some of that.
Host: Paul Barnhurst (32:43):
Yeah. So I think trying to bring this all around and I'm trying to figure out where to go. I think as we talk to leaders, one, you got a couple options with your ERPs. You could say, "Hey, I want better AI." You can add a layer on top. You can build yourself. The biggest risk if you're going to build on your own and you're not bringing external consultants is you're asking your day job employees to basically take a new job and do their old job more than likely and learn a skill that they may not naturally have.
Co-host: Glenn Hopper (33:17):
Well, I was about to say, oh, and by the way, learn how to be a developer.
Host: Paul Barnhurst (33:22):
So as a general rule, I would caution any company to really think they're going to build everything themselves. Doesn't mean they shouldn't build some stuff, doesn't mean some vibe coding, but there's a reason many of the AI companies we all say services the new SaaS in some ways. You've heard that term with AI is really think about where's your competitive advantage? And I would say in a finance department, very few finance departments, their competitive advantage is building agentic harnesses. It's well less than 1%. So that's the advice I'd give there regardless of whether it's an accounting tool, whatever tool you're looking at, simple things where it makes sense, doing some buying, building, doing some consultants, fine. But general, most companies will be more satisfied by going commercial off the shelf. And that could be commercial off the shelf service, so to speak, or going with a software that brings in agentic AI.
(34:22):
There are different situations. And it's weird saying commercial off the shelf service, but with AI, it's so much more is than it used to. That's my new buzzword. I just created it. What do you think?
Co-host: Glenn Hopper (34:32):
I get what you do internally, where you invest internally. What about this? What about services as software?
Host: Paul Barnhurst (34:40):
I used to always joke a company we had was a service as a software, not a software as a service.
Co-host: Glenn Hopper (34:45):
Yeah. So think about if I'm a consultant and I work with a lot of small accounting and fractional CFO firms, but if you're selling bookkeeping and basic accounting and maybe some fractional CFO services, if you can tell a company, we'll process your AP, we'll do your close reconciliations, whatever, and then you as the provider have these apps. So this is your competitive advantage if you learn how to do this more efficiently. So the company doesn't have to build it themselves. It's almost like offshoring if you think about it. It's, okay, now instead of paying time and materials for outside accountants, bookkeepers to come in and do this work, they're going to sell it to me as if it is software and it's just going to get done. And I don't know what the margin is. I don't care how long it takes them. I just know they're giving me a good deal on doing that.
(35:35):
That's going to be, I don't know, I think you just evaluate it the same way you do outsourcing and the spend kind of migrates from headcount or old time and materials billing to just it's a vendor line. It's like, this is our retainer to do your AP every month or whatever the case is. And then they're the ones who are building these more complex agent harnesses and doing it. I don't know.
Host: Paul Barnhurst (35:56):
Makes a lot of sense. I think don't be afraid to bring in offshore, bring in companies to help you build. Core tools, probably treasury, large companies, EPM, business intelligence. The larger the company, the more you're probably going to want to go with an established software. Exactly.
Co-host: Glenn Hopper (36:16):
Yep. Yep.
Host: Paul Barnhurst (36:17):
Smaller the company, the more you can go homegrown. The more it's a point solution, especially if you're unique, the more you can consider building or bringing in a contractor to build for you. And you need to do that evaluation like you would with anything else. Remember, it's about competitive advantage. I would encourage and love to get your thoughts, I don't think people should hold off. I think if you need software, you've expanded, look at it. You really need to think about how this connects, how easy is it to connect with other software with AI in the equation, which makes it easier. MCP servers, I would encourage you to make sure something like that's there or having somebody that can technically build you an API if you need that, you can figure it out with Claude. But again, that's time and skill. And I'd also say, Core in here, the spreadsheet isn't going away.
(37:06):
That doesn't mean you may not use it less. And an AI tool. At a minimum, every company, I think ERP spreadsheet AI tool, the bigger you get, then EPM, treasury, other things come into play. Data's more important than ever. Really thinking about where your competitive advantage is more important than ever, because you can waste a tonne of time on AI and not get anywhere near the benefit you would by just hiring somebody or just picking the software. So don't be afraid to hold. Go ahead and look. Be willing to look at the new players, the older players and pull the trigger if it makes sense. Ask them about AI, ask them about data, understand their roadmap and recognise that, yeah, you may rip something out a little earlier than you did before, but if it makes your life easier for several years, is that a bad thing?
(37:58):
I would argue no. So your thoughts. I mean, that's kind of my general advice if I was to boil it all down for people trying to figure out how to think about software in this day and age. And then if you don't have good data, don't waste your money on a bunch of AI tools until you fix
Co-host: Glenn Hopper (38:16):
It. Yeah. Well, yeah.
Host: Paul Barnhurst (38:18):
A bunch. I'm not saying any. Don't try to add a big huge AI layer without having at least a thought about your data and making sure you have a plan to fix it. And AI could be part of that plan.
Co-host: Glenn Hopper (38:32):
Yeah, absolutely. I think about migrating for another big client, migrating a bunch of data into, in this case, it's old Oracle database to a Snowflake environment. And I think about how much we used AI in the mapping and setting everything up.
Host: Paul Barnhurst (38:49):
Oh yeah. It's fabulous at that kind of stuff.
Co-host: Glenn Hopper (38:52):
I know. And so I even did that for another client on consolidation work where they had every company had the fuzzy matching and SQL to try to, of course
Host: Paul Barnhurst (39:02):
I got as many as I could. Writing your ugly old statements to try to figure out, well, if it's 70%, it's probably the same.
Co-host: Glenn Hopper (39:10):
Yes.
Host: Paul Barnhurst (39:10):
And then manually looking at it and going, nope, screwed that one up.
Co-host: Glenn Hopper (39:13):
But I think about all these mappings and maybe AI, if you have someone that has the ability to use it and understand how to do it, maybe that's what helps people finally get their data dictionary and their ontology and everything that they need that AI will then turn around and use. Maybe it makes it easier for you to get there. But I guess I keep getting hung up on what the nature of the work that the humans are going to do as AI gets better and better. And if I'm managing a team right now, yeah, it's great. If I can buy a tool that's going to do it, set it and forget it, that's great. But where you're spending your time, money, and effort right now while you're trying to, oh, by the way, do your day job of accounting and finance work, but is it capability building, tool buying?
(40:02):
Is it re-skill, upskill? Are we hiring new things? What are the skills we need to have coming in? I understand why there's volatility around SaaS stocks right now, but I mean, like we said through the whole show, I'm not ripping them out and replacing them with an army of agents at this point. It's just we're not there.
Host: Paul Barnhurst (40:22):
No, I think that's the core message. And then the last thing I'll add is beyond buying the software as a company, you always need a thinking about how can we bring the semantic and the context layer together to get the most out of AI? So whatever your roadmap is, however you're building, think about what's the solution long term to make sure I can connect AI to my financial and operational data and give it the context, think like you would an employee so you can get better answers. So keep that in mind regardless of how you're going. I guess that's keep the big picture in mind because ultimately you need an intelligent layer. That's the term we see all the vendors, every tool is an intelligent layer now, but you need that. And that's really just semantic and context for AI.
Co-host: Glenn Hopper (41:12):
Let me throw one more thing out here that I really, I don't know what you just said that kicked me off on this as we're getting close to end of time here, but I guess I was thinking about -
Host: Paul Barnhurst (41:23):
Why am I nervous?
Co-host: Glenn Hopper (41:24):
Yeah. Well, it's a whole other topic, but going back to what
Host: Paul Barnhurst (41:28):
I- Is this a poral moment, Glen? Yeah. You know the show, right? The Pixar up, have you seen that where the movie the dogs chasing the?
Co-host: Glenn Hopper (41:38):
That's right. As I was thinking about AI could help you clean it up and you were talking about setting up the semantics and all the data setup that you need, one thing about using the AI that's built into the SaaS platforms right now, you're logging in with your same role-based access. If you're not a data person and you're trying to now think, well, I'll build my own data warehouse and you don't get the whole role-based access, you could end up exposing by trying to build it yourself, you could end up exposing everybody in the company sees everything instead of the very clear role user.
Host: Paul Barnhurst (42:18):
If you're building and you don't know security, and that's why they're so strict at large companies.
Co-host: Glenn Hopper (42:24):
Yep, yep. Your friend Lauren and I are doing a webinar with Ramp on security with vibe coding and all that coming up here in a couple of weeks.
Host: Paul Barnhurst (42:32):
Not surprised. So there you go. There's your pitch to join him for the Ramp. We'll go ahead and wrap up. We'll also mention, I think one of the great tools out there, they're our sponsor. QFlow does some really good stuff and go to market. They have some really neat things they've done. I know they've got a lot of growth in private equity, so if that's something you're interested in, check them out. There's my plug for our sponsor, but there's lots of great tools. Don't hold off. Just do like you would always do. Do a good evaluation, think through things. If you need to bring in an external to give you some advice, go find somebody. Don't be afraid to spend a little bit extra to save some time. Glenn will be happy to come in. I mean, he is putting his hands up. He does that kind of work.
(43:17):
But I mean, in all seriousness, there is a lot of benefit to that. So we'll go ahead and wrap. If you have questions, reach out to us. We'd love to hear from you, and thanks for joining us.
Co-host: Glenn Hopper (43:26):
Thanks, everyone. Thanks, Paul.
Host: Paul Barnhurst (43:27):
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.