Will AI Replace Enterprise Performance Management? | David den Boer

We've been hearing a lot of FP&A software, SaaS software, EPM software, Excel. They're all dead.

I've heard Excel is dead really for over 20 years now. This tool's the next Excel, and yet Excel persists as the common language. I've yet to see a tool that has a larger function library than Excel. There's really no benefit to avoiding Excel. It continues to grow.

Meet David DenBourg, founder and CEO of Darwin Analytics Incorporation. He spent over 2 decades in enterprise performance Management, and he's here to make the case for why it still matters. The traditional EPMs, if they don't adjust, are in trouble.

EPM platforms have to do more to own the accountability part of it, not rest on their laurels of being kind of the broadly accepted way of uploading data for plans.

I think we're both on the same page that there's no question AI is going to change things. It's a question of how much does it disrupt the market? Welcome to another episode of FP&A Unlocked, where finance meets strategy. I'm your host, Paul Barnhurst, AKA the FP&A Guy, and each week we bring you thought leaders, industry experts, and practitioners who are reshaping the way we think about FP&A. Today I have with me David DenBore. David, welcome to the show.

Hi, Paul. Thanks for having me.

Excited to have you. So just so everybody knows, in this episode, David's been dealing with some voice challenges. So if his voice sounds a little soft or a little scratchy, go with it. We think the conversation will be worth it. So we're going to record anyway, but just so everybody's aware upfront and you're not wondering what's going on. So David, let's start with, how about you give us a little bit of your background, tell our audience about yourself, And then I'll set up the topic we're going to cover today.

Sure. Long time ago, I started out as an EPM administrator at a Fortune 500 company, working with a tool that has been bought and sold numerous times now. And I started out working with a software vendor, with a company called OutlookSoft, which was acquired by SAP to form BPC. one of the leading EPM tools for large enterprise. BPC has come to the end of its life. And over 20 years ago, I started a consulting practice called Column Five. And I also have a software business called Darwin Analytics. The services side implements EPM primarily for large enterprise. We have a global customer base. And on the Software side, we build practical solutions that help customers get more value much easily, much more easily out of EPM.

All right, perfect. Well, I appreciate that intro. So our audience knows we're going to do a little different today. We're going to discuss enterprise performance management. David's going to kind of take the position of the traditional players in the space, you know, what's historically been out there. That, that's what we need. That's where we're going. I'm going to argue a little bit toward the AI side and why, why do we need these tools anymore as we see AI continuing to grow. And what we're hoping is you listen to us take each position. You'll be able to form your own opinions of where we're going. We're not saying one opinion is right or wrong, but we're going to take these 2 sides and have a little fun in this debate. And the right answer is everything I say, just so people know.

Just kidding.

Um, but with that, before we get into, uh, the EPM discussion, I like to ask every guest this question, so it's kind of my go-to. What does great FP&A look like?

I think great FP&A gives you, uh, transparency between the narrative and the details. I feel like some users produce a narrative and it doesn't necessarily tie directly to the details. I like to see a seamless cascade from the story down into the numbers, as low as you want to go, with everything corroborating that narrative that you've produced.

So narrative with the details so you can understand the full buildup.

That's right. And especially if you're looking at predictions or the future, you know, EPM focuses, there's a lot of consolidation, of course, which has to be 100%. accurate. We also talk about the future plans. Where did you get these assumptions from? What is the thinking behind it? And that has to corroborate as well.

Got it. All right. That's helpful. And I can see where the EPM background comes in because EPM often, you know, the detail has to match the top line number.

Right. At least be directionally accurate. And you have to have— show your work behind the story.

Yeah. Alrighty. So let's jump into this EPM conversation. So we've been hearing a lot of FP&A software, SaaS software, EPM software, Excel, they're all dead.

Yeah, I think their death is a bit premature. Yeah.

All right. So I'm going to ask, kind of share some of your thoughts there. Why Why do you think we hear that? Why are they not dead? Why is that a premature statement? Are they dying? Give your take here.

I think I've heard Excel is dead really for over 20 years now. I think, you know, that this tool's the next Excel or Excel on the web, or, and yet Excel persists as the common language. I think that's really the need is to have this common portable format that everyone can understand. It's kind of like this universal language where one analyst, maybe from a partner, a bank, a customer, produces an output in some format, hands it off to another group, and they can look at it and see exactly how they got to the numbers, how it works, where the calculations are, and even inside a company, It's that same universal acceptance of the same model that makes the most sense. I don't see anything else coming close to challenging that. And, you know, when I look at, from my perspective, which might be a bit unique, when we're implementing a tool, so many of these tools have kind of a proprietary syntax or interface. It's, if you've been around a long time, You know, you have to ask yourself, is it worth learning this unique interface, or should I just do enough to get it into Excel, and then I can work in my familiar environment? And I think that the greater the Excel compatibility, lower the bar gets for more people to participate as developers. The more esoteric and specialized and rare The skill set is it just alienates users, cuts down on portability, and I don't see it as adding rival functionality. I've yet to see a tool that has a larger function library than Excel. There's really no benefit to avoiding Excel. It continues to grow, and if that's the main output format, then that's pretty strong. Obviously, there's BI tools that have a different purpose that are also strong. And, you know, Power BI is becoming ubiquitous there somewhat for the same reasons as Excel. So those tools are really the toolkit of your average FP&A worker these days.

Okay, so get the argument on Excel. What about the argument that FP&A EPM software is dead? What's your position on that?

I think that with AI, the cost to build, in the past, you would buy any software package because the cost to build was prohibitive. You would rather just buy something off the shelf that just works for a specific purpose. It's supported and you can involve the vendor in configuration and building out new features and all of that. And you knew that By partnering with the right vendor, you were going on this journey and they were going to be investing in their tool. With AI, the bar, the cost of that is lowered to nearly zero. And you can build your own. You can add your own features. You can do all these, these great things that are really specially tuned to your industry, to your models, to your culture even, and have that be exactly what you need. So it's very attractive to say, maybe building my own is something that is better than buying off the shelf. We don't have to make as many compromises, but on the other hand, you don't get support and all that. So I think that some companies are concluding that building a custom solution where it might have been prohibitive before, they're not a software shop, that's not their expertise. They don't have a huge group of developers sitting around that would've made them lean towards buying a product. Now they might be concluding, we'll just make our own. So I would say that's why they would say it's dead. I know that Microsoft CEO said a couple years ago that SaaS is dead. Add that to the pile of quotes that we've seen, but I'm not so sure. I definitely think there's a place for it.

I'm going to take the different argument here. Let's, let's break down Excel first, then I'll get to the FP&A software. The spreadsheet is not dead. I, I think Excel is dying. And here, here's the argument. I mean, we, we, we make the spreadsheet ubiquitous with Excel, but it's a 40-year-old architecture. It has limitations. I think we're going to see Claude. Or ChatGPT or whoever the big Anthropic, these big model winners are, release their own spreadsheet. They have all the training data. They can copy every formula. Nothing in the syntax of that formula is, you know, like proprietary where somebody can't just create that. We see it all the time in other tools. They can create a better IDE. They can add like HTML dashboards. Better connectors because they can build it all native versus having to go to a separate application like Power Query. So I think it's just a matter of time till, especially as we get more used to it, it's all right in that Claude desktop and the Claude code. How about a 3rd button that's spreadsheet and all your connectors come in and it can allow you to do things switching back and forth easier. So I don't think the spreadsheet goes away. I think it's a fabulous use case, but I could easily see one of these AI tools, and I wouldn't say easy, it would take time, but I could see them, I could see them displacing the current version of Excel we have today.

Well, I think there's going to have to be a lot of work done on trust and transparency, you know, being able to trace precedence and things like that, where you really trust exactly what's happening with the familiar Excel environment, even though it can be challenging, is That, that level of familiarity and just swapping it out for anything new is going to meet some resistance.

And I'm not denying that, but I think there's definitely a case to be made where we could see it being replaced. You know, when it comes to EPM software, where the future is going in my mind is, again, as long as we can have a governed data layer, we can have that context, that schema, we can start to build within tools. There's more and more tools coming out with saying, hey, We're going to manage your vibe coding. I have a friend who built an application that's designed for finance people to help with the governed and the compliance and the security and the OAuth, all while vibe coding. So kind of checks for all those things and helps make sure in that environment you're building something. So, you know, the EPMs are limited in this new world. I think many of them, and we're seeing them try to adjust that. They're all racing to change, saying, hey, we're the decision layer. We're the intelligence layer. We have all that finance context. The problem is the typical EPM doesn't have the operational context. So I think, you know, you're going to see a governor, a governed layer and more and more AI and some vibe coding. And I think EPM as it sits today is in trouble. Now, could it adjust? Sure. Will it change? Yes. Are we seeing, I think, a growth in things like Fabric Planning because they are BI tool with that semantic layer, kind of the warehouse, that whole thing. That all the context is there to be the decision layer and they're giving you planning on top. I think that grows along with the AI and the vibe coding. And I think the traditional EPMs, if they don't adjust, are in trouble.

I have to agree with that. I do think that there is also a system of accountability where, as you're making a plan, you're kind of establishing a contract, often compensation. is tied to it. And all these variables that are being controlled, like, these are the assumptions as of the time that you're committing to deliver whatever KPIs you own as part of your role. That has to be sort of frozen in time so that I can go back and look at what those assumptions are to be able to assign who is not necessarily delivering. I'm not saying these other tools can't do it, but as of now, if you view the EPM function as primarily uploading data or, you know, just getting data into the database rather than managing those contracts for performance with management and assigning accountability, I think there's a lot more rigor that needs to be added to The data wrangling that some of these data platforms like Fabric and even others are making progress. Snowflake is moving in that direction, Databricks as well. All of them are blurring the lines for sure, but there's more to it than that. I think EPM platforms have to do more to own the accountability part of it and not rest on their laurels of being kind of the broadly accepted way of uploading data for plans. So whoever can expand and own that area better, and I agree, operational data as well is something that has not traditionally been a strength of EPM. And I know that some of the products are expanding in that area, supply chain, S&OP, other things that they're doing. Whoever can get that one-stop shop where you can get a full view of what's driving the enterprise, will be very appealing to the market.

Today's sponsor is Lineos. Most finance teams I've talked to have built something impressive in Excel. Real models with years of institutional knowledge sitting in the formulas. Then the data underneath goes stale the moment someone posts a journal entry. And somebody spends a morning rebuilding it. Lineos provides reporting intelligence by connecting your ERP data directly to Excel. Reports refresh with current data, your formulas stay exactly as you built them, and you can trace any number back to the transaction behind it. Everyone works from one version of the numbers, including whoever is building the board deck. Your team moves from assembling reports to analyzing them when you leave it to Lineos. Learn more at www.lineos.com. www.insightssoftware.com/lineos/solutions/reporting-intelligence. That's www.insightssoftware.com/lineos/solutions/reporting-intelligence. And my argument here kind of on that is, right, so I get it, it'd be great for the EPMs to do that. But these are all bespoke tools. You've got to implement them. If they're already bespoke tools and we can build so cheaply with software now, why not just build a custom solution internally? You got great Excel, you got BI, you got data warehouses, you can get that governance and security. Why the additional cost and why a dedicated tool? With AI and with where we're heading, I think it becomes much easier for companies to say, It's bespoke. It takes months anyway. I'd rather just put— it costs me $100,000, $200,000. Why don't I just put that toward a developer, write my own requirements, and build something internally that I own that's customized to my needs?

Well, I definitely think that there's, you know, it's possible to vibe code anything, as you know, but I do have considerable respect for the rigor in the system, the way that auditability is handled. I think there's something about deterministic engines that are not just logic, but it's also the mapping and the workflow and other things. You can't just vibe your way through a planning process that has up to 1,000 users collaborating, working together. Some of our customers have that many in different countries with currencies and different GAAP rules. There's a lot of sophistication there. I'm not saying AI couldn't do it, but when you have transparency questions about how AI gets there and a little bit of hallucination, making a lot of inference-based assumptions along the way rather than working on static plumbing that has been established, I think I have way more confidence in the EPM tool getting to a more precise result that I can rely on than the vibe-coded. And to answer your question about the implementation costs, yes, that is a problem. That's— it's not only the cost for the initial implementation, but it's the cost of ownership and expansion and even switching. It tends to lock customers in that they don't appreciate. But huge strides have been made adding AI as a configuring agent that configures these robust platforms that give you the precision, the reliability, the transparency that the EPM tools are known for. But they give you that AI agility on the initial configuration, but on the ownership and extension and perhaps switching if you want to. You know, my own market that I know very well, BPC, is kind of end of life right now. And a lot of my customers are switching. By being able to apply AI to that, we can lower costs and build that independence so that you can configure something cheaply initially and then adjust it very efficiently over time. So AI is influencing every step of that process.

I think we're both on the same page that there's no question AI is going to change things. It's a question of how much does it disrupt the market? And so we'll let our audience decide what they think, but I want to keep going on some other themes around this. We're starting to see more and more, you know, previously we saw some people doing statistical modeling, machine learning modeling. Now we see, hey, I can give inputs to AI and it can really quickly build a model. And Even before all this, I still remember going to an AFP conference where Microsoft was sharing they had done a bunch of statistical modeling on their revenue. I think they were using it for real estate. It was more accurate than the human. So why do we continue to need the human to forecast it? Can't we just go to a default where AI does all the forecasting, maybe human reviews it versus it being human-led? What would your argument be to that?

Well, I'm certainly not going to argue that there's no place for AI. I think the most prudent place for AI right now, and I'll go back to that accountability, as long as we have human managers, and unless you're arguing that we don't need people in companies, as long as we have people owning the data and the responsibility to drive these numbers and being held accountable, they have to own the numbers. If you have a black box generating what your targets are, And it's landing in your lap. Hey, this is what you have to deliver, Paul. You're not going to feel that ownership. If that number came from you where you said proactively, I can deliver this number, now you're on the hook. And that accountability is at risk of being lost if we go completely AI. Now, where I would put AI is to challenge those human numbers. Because part of, I think, you know, when you say people are involved and it's more accurate, there's 2 sides to that coin. You have some people that are not incompetent, but these are complex workbooks and concepts and they make an oversight. Something doesn't get updated. There's just an error. There's a darker side. The darker side is the motivation. You have the ownership. We have people that sandbag forecasts. They could deliver a higher number if they stretched, but they try to obscure what they can do, set expectations low, overdeliver so that they get a higher bonus or have some other motivation. And I expect AI to uncover some of those things to work saying, hey, you're only putting in 2/3 of what I think you can do. That'll flag to leadership where to press for better performance. It will help. That partnership between AI and human motivation elevate the company's performance. That would be a more positive use case, I think.

So I'll take the devil's advocate position here. You know, a lot of these, a lot of studies have shown 50% of people, sometimes even more, are unhappy with their jobs. 20% do 80% of the work. We've all seen the million-dollar Excel errors. Every budget and forecast has a mistake. of any, in any company of real size. So why is it not, you know, AI creates it all, you give it the inputs, you still need that human, they review it at the end, but you really remove the human doing probably 80% of the work. I don't think we're quite autonomous. I'm not going to say, hey, it should be 100%. I see no reason AI can't lead it with kind of having what you call the human in the loop and us moving much more toward an AI. There are some exceptions. I'm not going to I'm not ready to say the robot overlord should run the whole company. I'd love to take that position, but I don't think I could argue it. I'd be laughing. But I'm trying to go as close as I can. And I think that's, I think that's where we can go, especially in stable, mature companies where there's not a lot of change.

Well, when you have a million-dollar error on a spreadsheet today, You know who to hold accountable for. Who do you hold accountable when AI hallucinates?

And I think you still have to own the finance department, still has to be accountable.

So what kind of control and transparency do they need to accept that accountability? It's not zero, but they might need their own.

No, it's not zero. I would agree with that. I couldn't argue that one.

Yeah. They might need their own AI that they trust To double-check the numbers on their side. It comes down to that transparency. Correct.

I would recommend that every model that's built, you're going to have some kind of adversarial agent that goes through and argues against everything that's been built. And somebody is going to need to review that, but I still think you can have AI lead the process in the sense of doing, doing the forecast, doing the first round, giving you everything. There still needs to be a human in the loop. I'm not ready to say we see a lot of companies claiming, hey, it's autonomous. You don't need anyone involved in a lot of things right now, right? We've all heard the autonomous finance. I'm not to that point yet. I'm not ready to make that argument for sure.

I do think, you know, you want agility, but what is your risk tolerance for that level of agility? Because you, I mean, look at the headlines from this week alone. And next week, spoiler alert, AI is going to be doing some crazy things with a high degree of autonomy. You know, whether it's hacking a rival AI company or it's making huge errors, we're going to continue to see that. So that human in the loop, that adversarial agent, I think for some of the things we're building, building an LLM council that constantly is debating and fact-checking and, you know, looking for errors and checking for risk. That's prudent. And I think you can add that to an EPM solution and get the combination of the ownership, the transparency, the boost from AI for agility with the fact-checking, get that AI Hallucination as low as possible, get the human approvals as high as possible, and move with agility. I think it's a combination.

Yeah, I was just thinking we should have, uh, agreed to run the transcript for each of our different phases of our argument through like 3 different LLMs and let them grade and see who the winner is. Might have to do that on the back end just for fun.

Okay. Do I get to pick which one I want?

You can pick the LLMs. I'll run the transcript.

All right. I want to, I want to provide my own prompt. If that's all right.

Oh, I'm not sure I'm ready for you to get to do the prompt.

Look, when you're talking about planning, there's gamesmanship. Do you think that AI will wring out gamesmanship? I don't think so. I think there's—

I was planning on gaming it with my prompt. What are you kidding me?

Well, exactly.

All right. So let's talk, I think, some of the other areas that we're seeing a lot about. So we've obviously covered AI versus human. Everything is dead arguments. Let's talk reproducibility. This is a big thing we hear. We hear the audit trail, all these discussions around it, and it's one of the biggest arguments you hear for the traditional players. Hey, there's governance and there's auditability and traceability. So kind of what's your position there? Talk to that.

Well, I think it comes down to this deterministic versus probabilistic. If you rely 100% on inference, then you have a problem. You can have 2 users asking the same prompt, getting different answers, because, you know, it's opaque what the assumptions are on either side. You have to control it. And I think you have to not leave to inference things that are known. So whether it's A formula, that's the most obvious example, where A times B equals C. You don't ask AI, hey, what happens if I change B? Well, if you can run through a calculation engine, you should do that. That's reproducible. But then you get into other areas where deterministic concepts also apply. Mapping of dimension members, workflow validation, certain things that need to be done that can be rules-based and reproducible precisely every time. I think what you have to have is removing as much of that from the inference engine and doing it in a very deterministic way for that reproducibility and precision. And I, I think that over time, you're gonna see AI generating that deterministic rule on the fly. And it'll make the decision that you've asked me like 5 questions and, you know, 4 of them are the same. I'm just gonna create a rule. I'm gonna run that rule for reproducible and transparent execution of that same question. In the 5th case, I don't have, you know, it's softer. I don't have all the data. that I need. I'm going to run it through the inference engine, but I'm going to tell you that, and I'm going to tell you that I used AI to generate this one. So it might not be exactly correct.

Yeah, I think on this reproducibility, I 100% agree with you that AI is going to write deterministic. We're seeing it today. You know, I'll share an example. We did some training this morning with a company in Germany. And shared an example of, I asked Claude to help me build something and it went through and said, hey, for this I'm gonna build Python script. That's basically a deterministic rule. It's running the exact same thing every time. It said, that's how I should do it. It created a, you know, its skills file, but both reference files that it asked to use were scripts and they were Python scripts. And so I think we're already starting to see that. So I think the reproducibility, you know, the audit trail, all this argument that it can't be transparent. quickly going away. I think the key is if you want to have AI that's reproducible, that's transparent, the logs, you need to build that into how you build your agents. And more importantly, well, I'd say 2, there's 3 things. You need to build that into how you build your agents. You need to build in deterministic to it. And you, you need a good data and schema layer. If you do those 3 things, I think you can get the reproducibility out of AI and you can get it close enough that the cost savings and other things are probably worth not having to have maybe all the tools you have today. Time will tell.

Sure. You can, you can reduce the tools, but when you get into multi-user, how do I know that your Python script is going to be called by some user in Australia? When they need the same answer. I need— that schema is going to have to know where to find that script. And I think that's where the EPM tools have a strength. They're designed for multi-user, and it's regardless of how many users you have or where they're located or what the assumptions, the same script's going to run in a very consistent way. I think unless you have your AI It's going to work great for one person, but when you have it set up for multi-user, you're going to have to have more sophistication than is averagely available today.

I mean, you could have org-level skills and that could be a challenge, but I think we'll get there. I think it's where we're heading.

You're getting dangerously close to building your own platform then. And it's, it needs a lot of thought to, to get that level of accuracy. And I think the EPM tools Deliver it today.

But I like building.

That's you. I think some customers would prefer for it to just work. But yeah, it's— that's building.

I can't even argue with that one. I'll give you that one.

Okay.

Score one for David there. So the reproducibility is an interesting one. All right. So a couple other areas and then we'll kind of We're going to wrap this up with some thoughts in general. So next position we're going to argue about here is, you know, this whole idea of cost of change, you know, kind of moving fast versus slow is, you know, my position here is companies can't afford not to be moving fast in this environment. They're going to get left behind. EPMs are a lot of big companies that often move slow. So I think the more and more you adopt AI, the better off you're going to be to prepare yourself for the future. And sure, there might be a little bit of pain along the way, but there always is with adoption. That's just life. So what's your argument to that? Today's episode is brought to you by our sponsor, Lineos. Lineos brings live ERP data directly into Excel, allowing your team to build reports with Fresh data, no stale files in sight. More on that in a few minutes.

Well, I'm not going to argue for pain. I will, I agree. I think there's more to it when you talk about large enterprise, and I don't want to alienate any of my partners, but I think that companies that have traditionally relied on a high cost of change as a moat to lock people into a particular stack are gonna be most under attack. Customers have never appreciated that, never felt that being held hostage was to their advantage. And the more that companies have policies that penalize companies for experimenting on the outside and moving data that they should own outside of that stack to experiment, I think they're going to get backlash. And I think that's already started to happen. I've seen AI definitely play a role with MCPs and other connections that allow for flexibility. But here again, the openness plus AI inevitably will lower those switching costs and will enable what I would call innovation agility, which is a new skill that companies need to embrace via AI. In the old days, it would take 6 months or a year, sometimes longer, to implement EPM. And forget EPM, it's just, let's say, implement a change in in your finance process. And you'd have this long process to implement it. And I would argue that there's a long tail effect where, because it's change, all the people with their manual process had years to adapt and get to the level of proficiency from wherever they came from, whether it's Excel or a different version of an EPM tool, 3-year adoption and proficiency cycle is intolerable now. They need to have the ability to change much faster and get to a level of proficiency within months or even faster. And I think that companies that in the past would measure value based on operational agility, you know, how automated am I, how many manual hours are needed to execute a process, are going to shift and look at innovation agility, where they say, I want to experiment with a new AI that maybe I built. I want to replace part of the process with something and see if I can't speed up a painful aspect of our process that handles it with more autonomy, more agility and moves faster. And I think once you understand that, there's no argument for standing still and staying on the sideline. I think what gets in the way is the people. You said 20% to 80% of the work. I don't know that organizations are ready to start calling the 80% of the people that are delivering only 20% of the work and relying heavily on AI, but inevitably that's coming.

So I say to that 80%, just keep pushing AI, force them along. I mean, change is about technology. And the reality is the cost to do nothing is too great. Everybody hated the lock-in. Everybody hated when you felt you had, you were strong-armed by the vendor. I worked for a company where we had a I won't name the tool, but put insert big finance tool here mitigation strategy. The goal was to get ourselves completely off their tool. We didn't quite accomplish it, but it was because of the way they worked in this case. And so just, just, you know, deal with a little bit of the challenges and the pain and start adopting more and more and as much AI as you can. Make sure you have that good data layer. You have a good context layer and, you know, start automating. So that, that, that, that's what I'll say there on, on the change. And I think in the long run, it will pay for yourself, but I can't, I'd be naive to say there won't be bumps in the road. There's going to be bumps in the road if you go slow. So why not just break a few eggs along the way and get there?

That's kind of like up to each organization, what their risk tolerance is. And, you know, I don't think the answer is make no change, but the pace of change has to be in alignment with their culture. And, but change is inevitable. AI is inevitable. And moving in that direction, relying on highly governed AI, highly, you know, embedded in maybe a familiar EPM tool that provides the audit and all that, that's right for some personalities. an organization. And, you know, the Wild West with vibe coding is maybe more acceptable in other organizations.

We all romanticize the Wild West, so why not just continue?

I can think of a lot of million-dollar spreadsheet errors that are the downside of that Wild West. All right.

So, you know, we've both taken a different position here. And so we're going to kind of wrap this up by talking about a little more what I say, a little more nuance. I've been very much AI, AI, AI. David's probably been a little more nuanced in his article, in his arguments here than I have. So where does the crossover sit? Let's just talk for a minute. Ignore the positions we've taken. I'd love to get your take. I'll share mine and we'll wrap up here. How should people, from your view, how should they be thinking about AI EPM software? How do you think that sweet spot comes together?

Well, I think, I think that there have always been something that bothered me about EPM is when you lay out an ideal, let's say on the services side, here's the ideal project with finance transformation and implementing the tool with maximum automation, training, support, all this. You present that to the customer and the customer gets out their red pen and says, okay, let's take this out. Let's take that out. Let's cut these corners. And inevitably, that would reduce the value that they would get. I think at a minimum, AI allows us to not cut corners and implement things the way they always should have been. And I think that if you accept that, you would say that EPM by itself in its current state is not something to be protected and unquestioned that I want to replicate what I'm doing today in AI and replace it. I think that there's a much higher level of proficiency that is available now because of AI, even within EPM tools. You know, you mentioned statistical modeling and you talk about the depth of modeling. Most companies don't do that. Most companies are just doing basic financial statement production even now, even with EPM tools that are capable of a lot more sophistication. So let's use AI to expand the value with the tools that you have as a starting point before we start talking about switching. Switching to AI to me is often replicating the status quo. The status quo is a product of corners that have been cut and opportunities that have been missed. So I think you gotta look at what is truly possible, where to apply AI to get to a higher value state, more agility. I mentioned operational and innovation agility, and then decide, how am I going to get there best? Is it by more fully exploiting the tools I already own or could easily acquire? Or is it something I need to build? Or do I add something to the layered on, on the outside, maybe with an MCP working to configure and connect and orchestrate multiple components? And I think that's what each organization needs to understand.

Great points. And, you know, as I think about this, obviously I've argued all AI I think the EPMs are adjusting. I think we're going through a changing time. I think that data layer becomes more and more important. I think small businesses, in all honesty, they'll be able to push out when they need a tool longer and longer with AI, with Excel agents, with ERPs that are much better at consolidation with these newer ERPs and accounting tools that are coming out. I think it gives you more of a runway before you need one. Really large companies, the calculation, the context, the The governability I think is still needed, but the tools are going to have to figure out how to be more operational. That's why we're seeing things like Fabric Planning, K4 Analytics, a lot of BI tools that sit in the warehouse. I'm seeing more and more tools designed to be that governance layer, and we're seeing things get combined. And if the tools don't adjust, they're going to die. I know one tool we just interviewed, a company comes out on Future Finance, will actually be out before this one. So if you want to listen to it, go listen to the CEO of Farseer where their goal is Hey, we want our customers to be able to build their own software within our tool. And it's an EPM. They're basically saying, we're going to open it up to JavaScript. We're opening it up to SQL and the technical people, you go build what you want. We're going to give you the modular capability to have people have built ESG solutions and other things. And I think we'll see more and more of that. So I think it's a blend and that context and schema is going to be more important than ever.

Yeah.

So I think we're in a similar spot. I don't think any of us, you know, either of us honestly think it's all AI or that the EPMs are dead. If the EPMs don't adjust, they will die. I think we could both agree on that. I think we both agree AI is not going anywhere. And I think the final thing that's critical to all this, the technology is moving much faster than the people. You know, you go to LinkedIn and you get this idea that everybody's automating everything and that's just BS. That's a bunch of marketing hype. I do training, and the reality is a lot of people, oh, you could do that with AI? Oh, I should be doing more than prompting? You know, like agents, skills. Most people are here at the beginner level. The marketing's over here, almost beyond what the tool could do, and it takes time to move. The prime example I give Excel's the default spreadsheet in the world, has been for 30-plus years now. And yet Power Query has been out for 15 years and maybe 10% of people use it. Fabulous tool. The adoption rate's been horrible. Yes, AI is quicker, but this idea that I've seen people, we're going to fire 50% of all people and everybody's going to be autonomous AI in 18 months— nothing moves that fast. Are you kidding me? These are big, huge organizations. So I think we can both agree there's time, but don't just sleep on AI and think, I don't need to change my tech stack.

I agree with a lot of that. I do think you're touching on something about ripping and replacing. And are you better served by a monolithic tool or approach, or are you better served by phasing in, starting out hybrid, Phasing it in and kind of orchestrating or connecting these various components. I think that's what's more likely to happen. And over time, AI will continue to expand and the tools will get better, obviously. So I think everyone has to have an AI strategy as far out as they can see, which might be 6 months. And things are changing very quickly.

No question. I think everybody has an AI strategy. I think AI adoption should be owned by the CFO. I agree with Glenn Hopper. His new book comes out arguing that. We won't get into that today, but another episode we just did. And so I think that's a fascinating subject. But as we wrap up here, we just have a few minutes left. I ask every guest these questions, so just looking for some short answers, you know, kind of 1, 2 sentences wide. What do you think is the most important technical skill in FP&A I think you have to understand the data.

The, uh, something about almost like SQL skills, but understand where data comes from and how to manipulate it.

A foundation of data, how it works, what good data normalization is. Do you have to be an expert in SQL? No. Do I think it's helpful to learn? I think everybody should understand the basics. I do. So if I had to give an answer, I, I'm at, that's where I'm at now. It's, it's data. I would agree with you. My data background's been incredibly helpful. I usually don't give my opinion, but I figure I'll agree with you on one here. All right.

Take it.

What is the most important soft skill?

Well, I have to agree with you on this one. I think you hit this often, storytelling, where you have to create a narrative out of the numbers. You can't start out by reading every line of a report. You have to make it say something and have a point and call out what is this data telling me to do? What action is this telling us to take?

Storytelling is always, always an important one. It's definitely at the, uh, near the top of my list. All right. This one I thought I'd ask you since we have an EPM specialist here. What's the number one thing an FP&A professional should do in order to have a successful implementation?

I think they should gather a wide variety of opinions, and that could be AI. That could be interviewing a number of vendors, could be interviewing customers, could be attending a conference, for example, like our EPM Summit. But the more you know, the more equipped a buyer you can be. A lot of people have agendas. That should not be a surprise. And vendors want to lock you in. They want to convince you their tool is the end-all be-all. And the more breadth of opinions that you have, the more your aperture opens up. And that's only going to serve you better.

All right. And where I want to finish is you have a big conference coming up in November. We're recording this in September. It'll be released in October, but tell people what the conference is, how they could learn more about it if they want to attend. You have, uh, in Vegas, an EPM conference. So go ahead and take a moment and tell us about that.

Sure. It's the EPM Summit. You can read about it at epmsummit.com. And it's the largest gathering of EPM tools with the most in-depth agenda. We have hands-on, we have panels with multiple vendors on the stage at the same time. We have over 120 sessions. So if you want that combination of breadth and depth for this technology, which is obviously a hot topic, and to see how it's evolving with AI, whether you're moving away from an older tool like BPC or Hyperion, or you just want to know what's available out there, it's the best place to come to get educated and inspired in this area.

Yeah. And anyone who's listened to this episode or been listening to prior ones, you probably heard some of my ads about that. So go ahead and check out EPM Summit. I think it'll be a great event. You'll learn a lot there. And David, thank you for joining me. I've enjoyed our chat and getting to spend some time with you.

Thank you so much, Paul. Thanks for having me.

That's it for today's episode of FP&A Unlocked. If you enjoy FP&A Unlocked, please take a moment to leave a 5-star rating and review. It's the best way to support The FP&A Guy and help more FP&A professionals discover the show. Remember, you can earn CPE credit for this episode by visiting earmarkcpe.com. Downloading the app and completing the quiz. If you need continuing education credits for the FPAC certification, complete the quiz and reach out to me directly. Thanks for listening. I'm Paul Barnhurst, the FP&A Guy, and I'll see you next time.