The Diary of a CFO · 2026-04-09 · 36 min
Key moments - from our scoring
Substance score
67 / 100
Five dimensions, 20 points each
Paul Barnhurst, known as 'the FP&A guy,' breaks down the Excel-versus-FP&A-software decision through a practical lens. Spreadsheets remain irreplaceable for flexibility and edge cases - M&A modeling, one-off analyses - but FP&A tools become imperative when pain points emerge: consolidation errors, privacy concerns, scalability issues, and the need for audit trails and database-level row permissions. The conversation moves into a three-phase tool selection process: research using market maps and free courses to narrow options, conducting a rigorous RFP (which forces comparison against requirements rather than features), and critically, choosing the right implementation partner and allocating sufficient internal resources. Barnhurst emphasizes that most planning tools are functionally similar - they all model 12-month revenue, COGS, operating expenses, and capex - so implementation quality matters more than vendor differentiation. The episode explores AI's actual value in FP&A: variance commentary, anomaly detection, Claude-powered modeling (which achieved 80-90% accuracy on Financial Modeling Institute cases in 10 minutes), and formula automation via Excel agents. Hype centers on the 'set it and forget it' promise and deterministic confusion: generative AI is probabilistic (may return different answers each time) while Excel formulas are deterministic (always the same result). For teams with zero AI adoption, Barnhurst recommends starting with LLM experimentation, then process automation via Cloud Cowork or Zapier, before testing spreadsheet agents like Claude Opus or Microsoft Copilot.
When pain points become critical: consolidation errors, privacy/data security issues, scalability problems, or the need for audit trails and database-level permissions. This typically occurs when spreadsheet complexity and team size make manual management unsustainable, though spreadsheets remain useful alongside FP&A tools for edge cases like M&A modeling.
The quality of your implementation partner and your ability to allocate sufficient internal resources - ideally 50% of a dedicated person's time for 3+ months to gather requirements properly. Most FP&A tools are functionally similar, so implementation quality and clear requirement-gathering determine success more than vendor choice.
Claude Opus 4.6 demonstrated 80-90% accuracy on professional Financial Modeling Institute certification cases in 10 minutes, handling complex tasks like debt structuring. It produces fewer hard-coded numbers and more traceable formulas than earlier AI models, making it suitable for building model sections with human review.
Deterministic tools (Excel formulas, Power Query) always return the same answer for the same inputs; generative AI is probabilistic and may produce different formulas or outputs each time, even for identical prompts, so it requires human validation and works best for variable-tolerance tasks like variance commentary.
Vendors claiming uniqueness, repeatedly citing roadmap items for critical features, making 'whiz bang' claims about exclusive functionality, or demonstrating insufficient knowledge of your business model. Request proof-of-concept pilots and interview proposed implementation team members before committing.
Our reviewer’s read on each dimension, with quotes from the episode.
The episode delivers solid, practical frameworks for FP&A tool selection and AI implementation that most finance operators would find useful. Paul articulates clear decision criteria (pain points, deterministic vs. probabilistic tasks, implementation partner quality) and concrete red flags to watch. However, the conversation includes moderate filler - personal anecdotes, softball follow-ups, and extended tangents about running marathons and Cheetos that dilute the substance-to-time ratio.
when the paying starts to become big, where you start to see errors. You have problems consolidating everything. You have privacy issues. As those start to grow, you have real issues with bringing in all your data. It's time to look at something else.
is this a task I'm going to repeat? Because if it's a repeatable task, GenAI is probably not the right place. You're going to need something deterministic
Paul presents a thoughtful deterministic-vs.-probabilistic framework for AI selection and a pain-point quadrant model that is useful but not particularly novel. The core insight - that FP&A tools are fundamentally similar and implementation matters more than tool choice - is sensible but well-established in the practitioner community. The discussion of AI agents in Excel and recent Claude performance is timely but lacks contrarian or first-principles depth.
These tools aren't all that different. I mean, how many ways can you plan? At the end of the day, they're 12 months in a year...we all have revenue. We all have some income stream.
the more stuff we can train it on, the more likely it will be right...There is no guarantee anytime it's going to output the right answer.
Paul Barnhurst is a legitimately credentialed FP&A practitioner with hands-on software selection experience, modelling expertise, and established presence (112k LinkedIn followers, multiple podcasts). He has implemented FP&A systems, trained corporate teams, and conducted comparative testing of tools. His background spans contracts, financial modeling, and corporate finance roles. This is a high-caliber guest with real operational experience, though not a Fortune 500 CFO or mega-deal operator.
I had the time. It had been three months later when I had the new jobs they'd given me, I wouldn't have had the time...That's what you need is not you need everything documented, but you really need to think through it
I wrote RFPs for the government before I went back to grad school and got into FP&A
Paul provides some concrete specifics - Claude's performance on AFM and CFM cases (80-90% accuracy in 10 minutes), Excel agent formula variation rates (40% variance on repeated builds vs. 80% similarity for Opus), pricing ranges for tools ($20-500/month), and named tools (Zapier, n8n, Power Query, TM1). However, much advice remains abstract or anecdotal: broad statements about AI architects, future org structures, and business partnering lack hard numbers, case studies, or named customer examples beyond vague references.
we tested Claude about a week and a half ago. And we gave it an AFM case...And it was 80, 90 % of the way there in 10 minutes
I ask it to build the exact same model two minutes later, and 40 % of the formulas could be different than it uses to solve it the next time
Host Wassia Kamon asks competent setup questions but rarely pushes back, challenge claims, or dig deeper. She follows Paul's leads rather than steering with sharp follow-ups. The exchange on quadrants is clarified because Paul prompted himself, not because of incisive hosting. Late-stage questions veer into personal territory (running, snacks, family) rather than pressing on unresolved FP&A substance. Paul dominates airtime with long monologues. A stronger host would have challenged the deterministic-vs-probabilistic framework, pressed for real customer failures, or demanded specifics on future org structures.
Gosh, can you tell us a bit more about one? I'm curious to hear about your Claude experiment, but then also the difference between deterministic and probabilistic
And I'm also curious, like a lot of teams are stuck in Excel because they're overwhelmed by the FP&A software options
Computed from the transcript - who did the talking, and the words that came up most.
Most finance teams stay on Excel longer than they should. The hard part is knowing when it is actually time to switch. In this episode, I sit down with Paul Barnhurst, also known as The FP&A Guy. Paul has trained thousands of finance professionals, hosts three podcasts including FP&A Unlocked, and has tested almost every major FP&A tool on the market. He is known for his clear, independent reviews and helping finance teams make smarter choices about the tools they use. We break down when Excel stops being enough and how to know it is time to move to a real FP&A tool. The biggest mistakes companies make when buying FP&A software, including the implementation traps that quietly kill projects. How to spot red flags in software demos and what to ask before signing anything. Where AI actually fits into the modern finance team and where it is just hype. What Paul found when his team tested Claude on real financial modeling cases. The difference between deterministic and generative AI and why finance leaders need to understand it. How FP&A teams will look different in the next few years. And the one soft skill Paul says is now more important than the technical ones.
Transcribed and scored by The B2B Podcast Index.
Welcome to the Diary of a CFO podcast. I'm your host, Wassia Kamon, and each week we explore how today's finance top leaders build high -performing teams, partner with CEOs and boards, and lead through growth and transformation without burning out in the process. Today, I'm super delighted to have with me Paul Bonhurst, aka the FP &A guy. Paul is a leading voice in financial planning and analysis.
He has delivered corporate training and virtual courses to thousands of finance professionals, covering FP &A best practices, storytelling, business partnering in Excel, and data visualization. He also owns the FP &A Unlocked podcast, co -host Financial Modeler's Corner, and Future Finance. He has built an audience over 112 ,000 followers on LinkedIn. He is known for his clear independent reviews of FP &A.
any software and for helping finance team make smarter choices about the tools they use. Welcome to the show Paul. Thank you. And after that bio that you wrote, can I just hire you to be my marketing person?
Absolutely. I accept all forms of payment. I didn't say anything about payment. All right.
Let's dive in. So you are an Excel MVP and you also review a lot of FPNA platforms. So where should Excel remain the hero and where does it become a liability when you think about it within an FPNA function? Yeah.
So, I mean, the reality, whether it's Excel, Google Sheets, whatever the fundamental spreadsheet is not going away. I don't see it going away anytime soon. Even if Excel went away tomorrow, we would still use a spreadsheet. So the way I think about it is there comes a point When the need for a database, for privacy, for integration, for your model, for greater calculation speed, many of those things becomes imperative.
They outweigh the flexibility, because the reality is a spreadsheet, Excel, Google Sheets, whatever, all these new ones coming out, is always going to be more flexible than an FPNA tool. Just by the nature of unstructured data, structured data, it's much easier to be more flexible. For that flexibility, you give up auditability at the level you would get with a true tool, the database level at a row permission, collaboration, the scalability, the calculation engine, some of the multi -dimensional modeling.
Yes, some of that can be done in Excel. Almost everything can be done in Excel or with the Microsoft tech stack on your own. It's a question of, do you really want to invest all that time and have to keep those people? So I think what you have to look at is when the paying starts to become big, where you start to see errors.
You have problems consolidating everything. You have privacy issues. As those start to grow, you have real issues with bringing in all your data. It's time to look at something else.
That doesn't mean you throw out the spreadsheet. You may select a tool that very well integrates really well with the spreadsheet. You may select a tool that doesn't. but you're still going to use the spreadsheet.
There's always going to be those edge cases. You're doing M &A and you're not going to put all their data in your planning tool and build a model for it to decide if you're going to buy the company. Yes. At least I don't think anyone is.
Maybe there's a tool out there doing it with, but not yet. So that's kind of how I think about it is when you're big enough and complex enough that there's that real pain. to where things are going wrong, it's taking too long. And I think everybody knows, we've all seen them, there's plenty I can list.
That's when it's time to look and say, what more could I add? And that could be as simple now as an Excel agent with some of the things they're building there that are many ways the light planning tool, all the way up to a big, huge enterprise tool, depending on your company. Well, thank you so much for sharing. And I'm also curious, like a lot of teams are stuck in Excel because they're overwhelmed by the FP &A software options.
I remember when I started in FP &A, there's only like couple tools, but now I feel as there's a new tool every month. And so how should a finance leader logically choose an FP &A tool for their team, especially when they're moving from Excel to that first FP &A tool? Call me. On a serious note, I think there's a few things.
I look at it as there's really three phases to this process. There's the narrowing down the tools you want to look at. And that's the research part. Talk to other people, check out the market maps I've done.
I have even a course that's free that you can go take on FP &A software selection. And so I think first is figuring out the market because the reality is people know of tools. But rarely do they really know, hey, are these good mid -market or are these good enterprise or the good small business? Is this one good in a certain industry?
What type of tools? And so someone, I can take some questions and narrow down their list to five, 10 tools quite easily. And so that's the first step is kind of narrowing it down. And then it's really going through a selection process.
I encourage people as much as we all hate it, do a proper RFP. you'll thank yourself in the end. It doesn't mean you have to be super strict. I started my career in contracts, so that's where that comes from.
I wrote RFPs for the government before I went back to grad school and got into FP &A. And I think there's a lot of value in going through that rigorous process because it forces you to compare the tools against your requirements, not against each other or against the whiz -bang features they have. Nice. And then the third thing, which is more important than the tool you select, is do you have the right implementation partner?
Have you done the work on your data? Do you clearly know what your requirements are? And can you implement it correctly? Because these tools aren't all that different.
I mean, how many ways can you plan? At the end of the day, they're 12 months in a year. There's 52 weeks. We all, with few exceptions, have revenue.
We all have some income stream. We all have a cost of goods sold and we all have operating expenses. And then we have capex, right? It's not...
It's not that different. Yes, there's a lot of intricacies in between that, but as much as the planning tools all want you to believe they're unique and there are better ones than others, they're pretty similar. Okay. And so it's in the implementation part that I think we often get in trouble, right?
Because we either ask someone, like people on our team, and I've seen it, to work on it in their spare time, right? Who has spare time? And then if you didn't do, like you said, the right research on implementation partner, that's another problem as well. Yeah.
And so I'll share an example. I was asked to select a new FP &A tool for us and the project kind of changed that for time and it came where, hey, how about you implement? We're going to... implement a US version we'd had at TM1 was the tool.
And I went through and wrote all the initial requirements. But at the time that was about 80 % of my job. I'd come in new to a company, they're still trying to figure out where I'd fit. Things had changed a little bit later.
And I spent a bunch of time with all the businesses gathered all the requirements. And I still remember one of the vendors implementation partners going, this is probably the most detailed requirements we ever seen this early. Nice. And that's when you, that's what you need is not you need everything documented, but you really need to think through it and you need time to do that.
I had the time. It had been three months later when I had the new jobs they'd given me, I wouldn't have had the time. It would have been nowhere near that good. And so without the time, you just really can't do it right.
And so if you're going to invest 60 ,000 a year and a hundred thousand in implementation, invest half of somebody's time or hire another person if you have to. Cause the savings will pay for themselves. Nobody wants to hear that. I get it.
Or implement AI and find some savings, whatever, whatever it might be, but don't skimp on the person's time. Cause they're your expert. Yes. And it makes buy -in at the end also easier when you had someone championing the solution throughout.
So you'll burn out the person if you give them a hundred percent of their regular job plus that. Yes. Yes, very true. So what are some of the red flags that you've seen in FP &A software demos that buyers should watch for?
I think, one, if they're always kind of telling you they're unique, two, especially the further you get. That first demo, not so much, but what I always recommend is you should do some kind of proof of concept. Once you've narrowed it down to a few tools, you're really into that process. If it's clear they don't know your business, they're not paying attention, they're just selling you something standard.
or you're worried about implementation partner, that's an area where I'd consider running. And I saw that with, I helped the company through the process. And when we started, I expected one tool to win. And we got in the meetings, it was clear one of them.
had done their homework and understood the business better. And the other was just trying to sell something standard and wasn't listening. So that's, that's the first increase of a concept in the first demo. The big thing watch for one promises, Hey, this is on the roadmap.
If you keep hearing it's on the roadmap, unless you're, you're also really in, in mature and you're confident in the company, probably a good idea to run, you know, or because the reality, if you pick any tool with only a few years, you're going to have to deal with some roadmap because it takes a long time to get there. So, you know, just keep that in mind. Don't, don't buy into just promises. So if a lot of promises, that's an area where I'd be concerned.
And then if it's a lot of whiz bang or we have this and nobody else does those types of things always make me nervous when I'm like, I've seen so many tools, people like, nobody else is doing this. There's nobody in the space. And what are you talking about? I'm not quite that direct.
I want to be sometimes like, And I said to one of them, I can give you a list of like 10 or 15 tools. No, no, none of them are the same as us. Okay. Like I'm not going to argue it with you, but you're wrong.
And I never heard anything from that tool after that phone call. So they obviously didn't scale too far. And so those are a few things. What about you?
What do you think there? What's your thoughts? For me, like you said, When it comes to demos, I like to see that a company did their homework because I work right now in a not -for -profit space. So not -for -profit, I don't, I'm not like a food bank to have like inventory.
I'm closer to a bank, but I also getting grants. So my financials are quite complicated because I have to follow, um, the financial institution audit, but also the not -for -profit governance. And so I was shopping for FP &A. software last year.
And so, yes, the one that I eventually narrowed down were the one that could show me how did they put the financials of a not -for -profit in their system? What is grand tracking going to look like? Because you show me cost of goods when I'm like, it doesn't fit my business model. So - Yeah.
They're showing you a manufacturing example and you're like, not quite. And that leads to one other thing, as you mentioned that, that just came to mind a little different, but - Make sure your implementation partner knows FP &A and ideally has some experience in your industry. How do you surface that? Because, you know, oftentimes the people that are doing your demo and other people that are actually going to implement, like you're dealing with a sales team and then the implementation is like...
The smaller you are, the more likely they're doing the implementation. The larger they are, the more options you have for implementation. Vendors may have their preferred implementer. but less control they have over who you choose.
So I think a couple of things, ask for resumes of the implementer. Ask for references. Say, look, I want someone that has this experience I have. I don't think this makes sense.
Or let me interview them. If I'm not comfortable, I'm not signing the deal until I'm comfortable. So if it's one where they're doing it in -house, would ask. What's the worst?
They say no. And then you're like, oh, that's a red flag. Maybe I'll go elsewhere. Yes.
Because they're not transparent for sure. Right. And then if they're one that uses a hundred implementation partners, and that's a pretty big name tool, then go talk to other people that have implemented it. You know, do a little bit of your networking versus just trusting the company.
Yes. I mean, networking has worked really well for me because I'm part of the CF leadership console. And I just went in, I'm looking at these two solutions. What would you have?
And I had like 20 CF was coming back to no, no, no, stay away. Yes. Yes. So it definitely helps a lot.
trying to narrow the right solution. Yeah, they're a great organization. I'm big fans of them. I'm getting ready to interview their head, Jack McCullough.
Oh, yeah. I had it on my podcast last year. Yeah. He's a great interview.
Great guy. Yes, absolutely. So I'm curious to hear, you know, when we look. out at the future of FB &A and we hear a lot about AI, our favorite topic these days.
I don't know if you've heard about AI, but where is AI already genuinely useful in FB &A and where do you think is mostly hype? I mean, I think variance commentary, it can be very useful. I think it's even more and more that anomaly detection, machine learning AI is already very valuable. And you can combine that with GenAI to help you learn some of that, to build better forecasts, to bring in external data.
I think where AI is most rapidly improving with Claude coming out for us is modeling. What it can do in Excel now compared to what it could do three, four, five, six months ago is night and day. I got an episode coming up Tuesday where we tested Claude. And we were extremely impressed.
It did some very complex stuff extremely well. Wow. And so, and then I think the hype is in this idea that you can almost fix it and forget it. It's the promise of, oh, it can do all, it's going to save you hundreds of hours and it can do everything.
I don't know many people that are working 20 hours because they have AI. We're still pretty much all working 40 hours. As Inch Noir said on my podcast, he goes, I guarantee you the lights are not going off at five at the investment banking houses because they're all using AI. It just means they're all doing more work than they were doing before because they can get more done now.
And so I don't know if that makes sense, but I think that's the hype, this whole idea of, oh, it will just do everything. And the second one is you don't need to review it. You can just trust it. No, it's still human led.
If it's deterministic, just like if I automate a process with Power Query, Power Automate, machine learning, and I've run it a few times. Okay, fine. I can trust that. I don't check my Excel formula every month to make sure it's still working the same, but it's never going to return me a different answer unless I change the inputs.
That's not true with generative AI. Gosh, can you tell us a bit more about one? I'm curious to hear about your Claude experiment, but then also the difference between deterministic and probabilistic, like those two different models and how they come together. Yeah.
So let's start there. So something that's deterministic always has the exact same answer in the end, right? Two plus two is always four. If I do a sum is formula and I give it my sum range and my criteria, my criteria ranges, I do all that.
It's going to return the same answer every time unless the inputs change. If I add more rows to it, I change a number, it will change. I use generative AI. Gen AI's whole idea is the more stuff we can train it on, the more likely it will be right, the more we can fine tune the model.
There is no guarantee anytime it's going to output the right answer. Where a deterministic I know if I put two plus two in, I'm getting four every single time. And so the probabilistic is it's using probabilities to give you what it thinks it wants. That's actually one of the biggest weaknesses of Excel agents.
I ask it to build the exact same model two minutes later, and 40 % of the formulas could be different than it uses to solve it the next time. I've run that test multiple times. Opus 4 .6 was almost the same.
It was probably 80 % similar. But this is only a model that needed five different types of formulas. When I ran it before that, I'd get completely different results in almost every formula and how they solve them. And that's because they're taking probabilities to get to the solution.
And then give me the second part of it again, because I forgot the other part of the question. Sure. Claude, tell us more about your experiment with Claude and how he's helping with modeling. Yeah.
So what we've done is we have this Mod Squad series. You can find it on YouTube. It has his own playlist with Inch Noor, who's the executive director of the Financial Modeling Institute, which does credentialing for modeling. He's a world -class trainer and Giles Mel, who runs a modeling house.
So the three of us have been testing all the tools. We've tested eight or nine tools now. done 12 episodes, and we tested Claude about a week and a half ago. And we gave it an AFM case, one of his cases that you do in four hours with instruction.
And it was 80, 90 % of the way there in 10 minutes. I mean, it was really good. It picked up on a lot of things. We gave it a CFM case, which is the harder case.
And it was probably 90, maybe even a little higher. It was a debt structure, layer debt structuring model that you had to build just that section, just the debt section. And it did. very, very well.
Like it's at the point now where I would trust it to build something with my inputs, with my review, with proper checking, where before I might use it on little sections and a ton of prompting. Now we're starting to get to that point where people are going to use it to build a model regardless. They were before. But I mean, good modelers that know what they're doing are going to be much more comfortable having a build sections and parts and models.
Yeah. And you can see your formulas and you can trace them. Cause that was always my biggest thing. Like when AI came in, I would test to see how it does with modeling.
I'm like, I need to trace my formulas. I need to know. I don't want to see hard coded number. I need to know where it's coming from.
You still get the occasional hard codes. That is an issue, but getting much better. You still get more complex formulas than you would write. in many situations.
And that's why anyone that's an expert in this space is saying, look, AI is a magnifier. You got to know the fundamentals. And that's true in data viz. It's true in financial modeling.
It's true in most places. The people who are going to get the most out of it have taken the time to learn what they're doing. True. And so if a finance team has zero AI right now, what would you say is one low risk but high ROI place to start?
If they have zero AI, let's just say, I'll start with the simplest, they haven't used it yet in professional. Get an LLM and start experimenting, get comfortable with it. I think 90 % of people have at least done that either professionally or personally. Hopefully they have.
So I think the next two areas is one, use AI to document a process and then try either with some of these tools out there, Cloud Cowork and N8n, a Zapier, depending on how deterministic versus depending on the task and what you need, try automating it. Try building out with the help of Gen .ai that process. I think that's a great area.
Pick something that's really manual and think about why it's manual. And I always like to use the framework is make a list of your pain points, look at the highest ones, say how much could AI help on these. and then try some. Now for that first task, maybe pick something that's low time, fairly simple to get your hands wet or your hands dirty, so to speak.
So that'd probably be that. The third one, I would say, if not that I think everybody should at least install and be testing a spreadsheet agent, not saying turn your five -year financial model over to it and let it build it, but start using it. Have it audit some of your files. When you got to do a next investment, have it maybe build And you can still build on your own or, you know, review it, but don't, don't just do it all on your own.
Start using it to assist you writing formulas, whatever it may be. And where do we start with those Excel agents? Cause some people till this day, when I speak with a lot of people, all they know about AI is chat GPT. They don't even know about fraud.
There's about 30 tools out there. I just did a post today, a market map. If you're in Google. You know, check out what Google has.
There's a few other tools. If you're Excel, I think there's a couple places to go right now. You could do the copilot route and it's good. I wouldn't say it's the best in the market, but it's good.
Claude right now is one of the best I've seen. Opus 46 is the best model they use. So those are your first two options, probably your cheapest and the ones that will make the most sense for most companies, because almost everybody has Microsoft Office. And if you don't have Office, many people have Claude or can get Claude within the company so you can use it in Excel.
If not one of those, then you got to look at a third party add -in. And there it's deciding what you really want. It's a little more of a process. But again, it's nothing like an FP &A.
You don't need to do a formal RFP and FP &A tool. I mean, we're talking their prices are anywhere from 10. I think the highest I've seen is 500 a month for a user, but most for small companies are going to be somewhere in the 20 to $50 a month range. Not a big investment.
Okay. So I remember when we spoke about AI in finance, you gave a very nice framework on how people should use AI first. You mind sharing it again? It was like a quadrant thing.
Yeah. I'm trying to remember what my quadrant was now that I shared. So I know, I know there's a pain point of how you select something, but I think beyond that, you know, something you have to think about is one, is this a task I'm going to repeat? Because if it's a repeatable task, GenAI is probably not the right place.
You're going to need something deterministic in the sense of, it's repeatable. I need the exact same answer every time. Then it's look and say, Hey, should I automate this with Power Query or Power Automate? So I think you first, you need to understand, is this the one -off?
Is this a repeat task? And then is what's the level of variability? What's the level of tolerability? So can I have it vary every time?
Like variance commentary. I doubt you look to say last month, well, you used a different word here with the same variant. Yeah. Nobody cares.
And if they do, they have too much time on their hands. Right. And so you have to understand that. What's the toleration for error?
And that will help you decide, is this somewhere AI makes sense? Then it's a matter of understanding, okay, is it gen AI that makes mistakes? Is it machine learning? Is it just general automation?
That's kind of the framework and how I really think about it. is kind of asking yourself those questions, then deciding where to tackle first is that whole framework of where's the biggest pain point, where is AI the best? That's the ideal quadrant. If it's a low pain point and low AI is not good, that's human.
All right, if it's a high pain point and AI is not good, that's a wait or look to non -AI. Because even if it's a high pain point, if you can automate it, great, you may be able to do it with non -AI. And then the last one is, all right, well, if it's a low pain point and AI is good at it, I'll put that toward the end. Or I'll use that to experiment on because it's easy if it messes it up because I know it's a low pain point anyway.
Yes. Yes. Thank you so much for sharing. And I'm also curious as we are...
embedding, like you said, kind of a roadmap. I look at my high pain point, I look at AI can really help. And I have this whole plan of how I'm going to implement and embed AI in my workflows. So what do you think an FPNA team will look like if AI is fully embedded into the workflows and processes?
It's a good question. I'm still, to a certain extent, trying to figure it out a little bit. I think one, you see tools. You know, there's confluence payload, I think is others and others are these AI analysts are coming out a lot more.
So I think what you'll see in every tool, every team is off some kind of AI analyst. That thing can work 24 seven. And then within any large FP and 18, you're going to have an AI kind of architect specialist that can help with agent stuff. Cause no matter what every company wants to make it simple and, and do that.
There's going to come a day where nobody's going to outsource all their agentic AI to have someone else build that workflow or that agent. They're going to start doing that in -house, just like nobody has all their modeling outsourced. You build some of that. So I think there will be an AI architect that will help bridge that gap.
Kind of like you've seen in the last 10, 15 years, 20 years, many more finances. Sometimes FP &A transformation people are kind of systems people. I think you'll see the equivalent of an architect. I think structures, especially bigger companies will become a little more flat.
You won't need as many analysts. Now that's not to say other jobs won't come about because you'll have the AI architect. You may even have an AI implementation person, depending on the size. Someone who trains, they may be company -wide, they may be...
you know, FP &A or finance, the whole office specific. But those are some of my thoughts right now. And I think you'll continue to see more and more of the collaborator style in FP &A. I'll give an example of that.
My last job before I started my own business, my job title was director of finance, operations and data analysis, something like that. I didn't manage a single person. I didn't even actually build the budget, funny enough. I coordinated between FP &A, operations, and our data analytics team on what all our reporting was going to look like and how we were going to make the transition to a SaaS business with leadership.
And so it was much more of a coordination and collaboration role where they're looking from experience because you're managing and having conversations. OK, let's clarify what bookings means across the company. And so I think you'll see more and more of that collaboration. And strategic, not just tactical, but a lot of strategic collaboration.
I think that will become a more prominent role. Even more than it is today. Okay. And as you're training corporate teams and you know, the curriculum is probably changing.
Maybe you seeing CF was already asking you to emphasize more things than others. Like what have you seen lately? So, you know, I'd say over the last few years and starting to see more and more of AI training, but soft skills, and Excel. Those are the two biggest that I see quite a bit.
On the soft skills side, a lot of storytelling, sometimes around managing time or conflict, but I say storytelling and influencing. So many are like, I need my FP &A to be a better business partner. I need them to be curious. I need them to understand operations.
I can't tell you how many times I get people saying some combination of what's most important for FP &A is, as far as being a better business partners, understanding operations, being curious. I've had many people say the most important soft skill is empathy. Now those aren't the things you think of when you think of your FP &A person. When you're hiring, you don't think, I wonder if they have a lot of empathy.
I never did. I'll admit it. I never once thought, hmm, wonder what their empathy's like. Yeah, that's an area that I think FP &As they're wanting to see more and more of.
And then recently, obviously, the one everybody's asking about is AI and not getting left behind and how can I use it? And I just trained a large makeup company recently in cosmetics. There are staff of about 40 people on copilot. And so when you think about how CFOs will need to reskill their team to thrive in this environment, because like you said, soft skills, it's kind of hard to teach.
So I'm curious to get maybe a one -on -one business partnering from the FP &A guy right now. Yeah. So, I mean, definitely there are aspects that are hard to teach. For me, if I'm telling someone to be a better business partner, the things they can easily control and where they can start.
One. You can learn the business, regardless of how good your communication skills are, regardless of all your other soft skills. You can go out and ask for those meetings. You can really come to understand the operations.
Two, you can learn to be humbly curious. And when I say humble curiosity, the reason I use that term, it was someone, it was coined by a guest I interviewed, is because we all have the person who's curious just because they want to prove you wrong, or they're curious because they want to be right or look smart. You need to be curious because you deeply want to understand your business and your business partner. I think those are things everybody can easily work on.
Three, which will tremendously help your soft skills, is learn to be good at data storytelling, but also data visualization. And data visualization, there's a lot of science to that. By learning what a good visual is like, you can go a long way on the technical. That allows you to then tell a better story.
which is going to eventually allow you to be a better influencer. So sometimes you can start with those easier things that really are about effort, then work on fine tuning with your bosses and hey, where are my biggest weaknesses and really tackle those. Okay. Your boss like, well, you're really bad at pushing back.
Okay. What's the plan to do that? Or I'll give a personal example. I was way too detail oriented.
I literally had a director who wasn't my director, but he kind of was because the other team was all elsewhere. And so we talked quite a bit. He told me, he goes, look, every time you get too detailed, I'm going to tell you. And he would just, I am in a meeting and he's like, bring it back up.
You've lost everybody. And you know, that really helped a lot. And then I had a general manager that looked at me one time and he said, Paul, I'm going to teach you the bluff principle. I'm like, I don't play cards.
What are you talking about? And he goes, bottom line up front. He's like, just get to the point in your messages. Then you can add whatever you want to the bottom.
Cause I was trying to give the entire background and all of them. And so it stalled my career for a while. And so that's something that I would recommend. Okay.
Thank you. And now I'm curious about, you know, now that you are training more and more teams, what do you hear from the professionals that you are training? Not so much the companies, but the actual FP &A analysts and managers. What are some of their concerns or worries right now?
Or what do they want to learn more about? You know, they're all interested about AI right now. Okay. You know, a lot of them are, they often have a certain situation or what do you do with a difficult partner?
That often comes up a lot. We're having conversations like, well, what if the business is not listening or how do I manage that? And so we'll often try to walk through different situations. But I think right now the overarching for most of them is really trying to figure out what the impact AI is going to have because everybody's afraid to change.
And the pace of change in the last, what, six years since COVID has been unbelievable. I think almost everybody, whether they say it verbally or not, is just the constant changes on a lot of people's minds and it creates anxiety. Even though I think in the end it will be a good thing. I think on the whole, we could benefit in society.
There will definitely be impact. Yes, there will be for sure. So what is the best career advice you ever received or usually give to FP &A professionals? I think the best advice was an advice.
It was watching how much my dad was willing to serve and work. And so seeing that is something I've taken into my own life. I'm a big believer in the best way to help others is to serve others. And I had to learn, not everybody shares that.
Sometimes you have to take different approaches in leadership. But I think being willing to work hard and serve. And then the advice I give people, usually they're people wanting to break into FP &A. And so I break it out different depending on where they're at.
But one of the biggest things to say, look, if you want to be good at FP &A, One of the biggest advice that I always give is learn the business. If you're trying to get into FP &A, create that roadmap, figure out your weaknesses, make sure you can do Excel, make sure you can do financial modeling. But remember, because most advice to get are people reaching out wanting jobs and in that area. And one of the biggest things I try to tell them is your job is to make sure the hiring manager doesn't feel risk in hiring you.
If you can reduce that risk, I'll give an example. Let's say you're hiring someone and one looks like, Hey, they have the potential to be a superstar, but they might completely flame out. So there's a little bit of high risk, but you can see there's a really high reward. Are there's one that, okay, they're going to be a good performer.
I'm not sure they're ever going to be a superstar, but I have almost zero risk. Who are you typically going to go with? The second person. Right.
And especially in finance, we tend to be risk adverse. Yes. You're going to take the second. So I think, you know, helping people realize you're really trying to mitigate risks.
So think. understand what their pain points are. And that's more job than career. In career, I think the biggest advice I would give is learn how to serve within whatever you're doing.
Thank you so much for sharing. I can tell, you know, from how you helped me with my podcast and so many others. I know I joked I call you the Oprah of finance podcast. Put that picture of Oprah with a beard on LinkedIn.
Oh, that was That was DHA marketing, yum, out of France. Oh, gosh. But yeah, I can see and I'm super grateful for the work you've done, not just, you know, on LinkedIn, but also behind the scene on LinkedIn. People may not see all the DMs you send to support, you know, the whole community.
So thank you for that. I think I appreciate that. It's been fun. What's it been now?
Three years? Yep. Three years already. Yeah.
We're getting old. No. 39 .99.
Inside joke. Sorry, everybody. Yeah. Inside.
No, completely. We're going to keep it inside. Last question. What is your favorite thing to do outside of work?
Outside of work? What's this you talk about? No, there's a few things. I like spending time with my family.
I have a 12 year old daughter. I get to spend a lot of time with her. I enjoy running. I usually run every day.
And then I like young adult fiction. I'll admit it. There's my guilty pleasure. Oh, okay.
Nice. I'm in Netflix and recently I started lifting more ways than Cheetos. So it's getting there. What brand is heavier than Cheetos?
Yeah. Cheetos. Like I haven't had any Cheetos since the Reno 2026. So to me, this is like a big achievement.
No, I need it. I, I'm continuing trying to eat better. I had a I had some health challenges in 25 and I got to be better with my eating. So I get it.
But you run so much. How many miles a year in a day? Um, usually it varies whether I'm doing a speed or long run, but three to six, although I've missed this week, unfortunately, but probably three to six miles. I'd like to get more of about eight, eight on average a day.
So about 50, 60 a week. I want to run the Boston marathon. Oh, good. Good for you.
I ran my first marathon when I was 17. Fun fact. Well, we'll continue cheering you on for a distance. When I go to Boston, I expect you to be there at the finish line.
Oh, okay. We'll try not to finish. I cannot promise that it won't work. Well, thank you.
Thank you so much for being on the show. Well, thank you for having me. Always a pleasure. Thank you.
Other episodes covering the same guests and topics, from across The B2B Podcast Index.