Mike Dion works at a Fortune 100 company. He has previously worked as a finance leader at Verizon-as well as startups and as a mentor to enterprise giants to scrappy startups—unlocking tens of millions of dollars in value across industries like Entertainment and Telecom. He does this through Mike’s F9 Finance: a no nonsense website and newsletter (with 20k subscribers) passing on the skills that have accelerated his career, providing a guide to new tools (based on his experience automating 100,000 hours of labour) and secrets to promotion. He tells Glenn Hopper: “ Three things, increasing revenue, decreasing expenses, and making your leaders look good. Those are the three things that move your career, not the reports, not the forecast tools to get to that.”
In this episode
- Passion in media and entertainment
- putting in our first consolidated planning system
- At Verizon starting a center of Excellence (after facing a situation where 40 VPs wanted 30 decks based on Excel files
- Lack of approachable finance content: introducing Mike’s F9 Finanance
- Secrets to dynamic modeling and scenario planning
- 3 Ways teams are getting overwhelmed with forecasts
- Better prompting and my relationship with AI
- A completely new answer for Fave Excel Function
F9 Finance: https://www.f9finance.com/
F9 Finance YouTube: https://www.youtube.com/@f9finance
Full transcript
Glenn Hopper:
Today. Welcome to FP&A Today, I’m your host, Glenn Hopper. Today we have with us Mike Dion. Mike is a senior finance leader with over a decade of experience helping businesses from Fortune 100 firms to startups unlock value and streamline operations. As the founder of Fnine finance.com, he empowers finance professionals to save time and excel in their careers through automation. The author of The Financial Storytellers Playbook and the Ultimate Guide to AI Prompting for Finance, Mike Blends humor, expertise, and practical advice to simplify complex concepts. With over a hundred automation projects and over a hundred thousand hours of labor saved, his work has become a go-to resource for finance innovation. When not consulting or advising, Mike enjoys time with his family cruising and following market trends. Mike, welcome to the show. Hey, thanks Glenn. I’m so excited to be here today. Yeah. So I’m gonna shift everything and we’re just gonna talk about cruising for the whole show. Is that all right, <laugh>? I could, I could talk about that for hours. <laugh> <laugh>. Awesome. I get, we’ll, we’ll get to that at the end. I think maybe our, our guests would rather hear about FP&A ’cause we do have some hardcore FP&A a nerds out there who, um, I think would rather maybe sometimes be doing fp and a than Cruising <laugh>.
Mike Dion:
I could do both all the time. Yeah, that’s, that’s life right
Glenn Hopper:
There. Yeah. There you go. Yeah. <laugh>. So, well, let’s dive right into it. I want to hear about your career in FP&A from working with both the Fortune 100 companies all the way down to startups. Very, very different fp and a same concept. Lot of different information, different ways to approach. So we’ll get to that. But I guess maybe walk me through what initially drew you to finance, and you and I have this in common, it sounds like from very early in our career, automation and technology were kind of joined at the hip with finance, and you’ve used it through your career. So walk me through some of that.
Mike Dion:
Absolutely. So when I first started in college, I was a general business major. Didn’t really know at all what I wanted to do with myself. That wasn’t until I took introduction to Finance. I remember sitting in our first exam, I was looking around and there were actually people crying, taking the exam. And I’m like, this is so easy. This is awesome. I love this. What, what’s everyone so upset about? Um, and actually afterwards the professor approached me and said, Hey, based on your score, I want you in my finance program. Um, and from there, the rest was history. I started my career as, uh, an FP&A intern, uh, for a media and entertainment company. Um, really from, from start to finish, I’ve always just been fascinated by technology and what technology can do, getting the most out of Excel. And, and the biggest thing is I’m never happy with the process.
A process is never good enough. I might stop working on it ’cause I’ve gotten enough of the value, but I’m just never happy with the process. Built my career up at that company. From there, I wanted an opportunity to just take my career in a different direction. Moved over to the startup world where we went in no systems. It was, it was literally a brand new company, clean slate. They’d spent a little bit of money but didn’t even have a QuickBooks set up all of our systems, integrating with hr, putting all the processes in place, and you know, being, being the person for everything, which honestly really made me appreciate large companies having tax departments and, uh, payroll departments and all that. ’cause payroll doesn’t go out, I’m fixing it From there, uh, jumped back into the media and entertainment industry. Wanted the opportunity to work a little bit closer with the CFO.
So, took an opportunity at a studio to work, uh, side by side with the CFO, doing all the board decks, all the presentations. And they had just put in their first consolidated planning system and didn’t have any processes around that. Had the opportunity to really stand that up, automate the forecasting, automate the consolidation process, uh, where they literally used to be manually adding files together and didn’t even have them linked from there. Went over to, uh, try something new in the telecom industry. And Verizon helped stand up a center of excellence. The great thing about that was they wanted us to try as much tech as possible. I got my hands on nearly every dashboard program out there. I got my hands on a lot of the new no code or low-code coding tools, and had a lot of leadership support to just tear everything down and rebuild it. Um, and then from there, decided to jump back to my original passion media and entertainment, um, where again, on the cusp of, you know, brand new systems, brand new ledger systems, brand new dashboarding systems, and just a lot of processes ready and ripe for innovation that have been done the same way for a long time. And that’s back where I’m today and just absolutely loving every minute of it.
Glenn Hopper:
In addition to your professional work on your, your side hustle is this F nine finance.com site. The focus of that is on helping finance professionals learn how to automate and optimize their workflows. And tell us a little bit about F nine finance and kind of what inspired you to start it, what kind of things you focus on there and, and maybe from the feedback, what are some of the biggest challenges that you’re seeing fp and a teams facing today?
Mike Dion:
Yeah, I think two things really got me going on. The first was there’s just, there’s a lack of approachable finance content, especially when you get to automation tools. It’s almost like there’s kind of some gatekeeping either because people, you know, want the job security or because people have something to to sell you, right? And they make it sound more complicated than it really is. And even sitting in fp and a teams in large companies with a lot of resources, people didn’t know about these basic tools, if I go even today and ask 10 people if they know how to use Power Query, which is a free tool built right into Excel, maybe one or two of them out of 10 will know, will know how to use Power Query. And it is a free tool that every fp and a professional could benefit from. So that message is just not getting out there, whether it’s just from how busy all of our teams are, or from, uh, you know, people being given the resources or feeling like they know where to go for those things, or thinking that they need to go ask for budget when they can get started for free.
And for fp and a, you know, especially since covid, our teams are getting smaller, but business is also getting a lot more volatile. And I’m seeing the need for dynamic modeling and scenario planning more than ever before. Our scenarios are getting wider fp a’s insights on that, you know, the businesses is demanding them now. Whereas when I started my career, some people didn’t even like it when we showed up. We would, you know, kind of send them a report and that was our interaction with them. And now our operations teams are demanding. We have a seat on the table and come to them to tell them what’s the range of possibilities and how do I move it to the best case scenario And away from the worst case, that’s only accomplished with a smaller team. If you automate, we don’t have time to be copying and pasting 20 slides into PowerPoints anymore.
Glenn Hopper:
Yeah. And really you are, you are preaching to the choir here because I think being in the startup space, and I’m, I’m sure you know, it’s the same at at big companies too. You never have all the resources that you want. But being in a, a space where you’re really strapped for resources and, uh, in my career was in private equity backed companies and the, the demands and the amount of reporting that they wanted and the speed with which they wanted it. If you didn’t lean into automation, I mean, I, I, I just wouldn’t have slept and <laugh>. So, you know, I will spend 50 hours building something if it saves me two hours a week in perpetuity. I mean, it’s, you know, straight line math to figure out, you know, the, the ROI for that. One of the things that you’ve talked about is saving organizations over a hundred thousand hours through automation.
And I’m curious your approach to automation, and I’ll tell you mine and see how this aligns with what you do. So for me, anytime I’ve come into a company as a new CFO, or even as a, a consultant coming in, I wanna look at two things, the human process flow, what people are doing now. And I don’t even mean just in finance. I mean, if we’re, if we’re tracking data, I want to go all the way to the front. If we’re <laugh> looking at the CRM, if we’ve got somebody in the, in the sales funnel as a prospect or lead or, or whatever, let me gather that information there. Let’s pass that information and carry it through the systems, and then look at how the information is being handled at each step along the way till it goes into billing systems and project management systems and all that.
But you start kind of with the process and you see where those bottlenecks and, and speed bumps are, where can, how can we automate this pain point here? And as you do that, the, the amazing thing to me was that automation and data go hand in hand if you want to get the data, the way to get that is to get sort of the, the automation of the, of the process. But it doesn’t matter what the system is. That’s where your requirements come from. And I’m wondering, when you come in to a new company, and well, maybe there’s two ways to look at it. One is the startup where a system hasn’t existed before and two is in a company where there, you know, people might be used to doing things a certain way, very manual, like you said, and, and trying to change that to automation. How do you approach all this?
Mike Dion:
My first biggest thing is to get quick wins. Because especially if you’re coming in to establish processes and people don’t wanna change things, you have to get them on board that this is good for them, that this is improving their life and it’s gonna make their day better. While also kind of addressing that concern in the back of people’s heads, there will always be the concern is, you know, what happens to my job? Which is why I really emphasize this isn’t just about making your day better. You are going to learn the skills that are transferable. If you are the person who can automate things, if you’re the person who can save your leadership money, you are more likely to be the one they’re keeping around because you are at, you’re not only adding value in your job, you’re adding value across the whole finance organization.
And that approach has really broken down a lot of barriers. I had a great example from Verizon where I came into a team and supporting, um, actually the, the CFO’s organization as their business partner. And that is a lot of executives. I think there were, you know, 30 or 40 vice presidents and up, everyone wants their own special reporting. So when I came in, my team was producing 30 individual decks from 30 individual Excel files and was the last one out of the office every single month. Just crazy. I said, this has to stop. So how do we do this? ’cause first of all, I’m gonna have a bunch of senior executives who’ve been here for 20 or 30 years who will want their report. There’s gonna be a resistance change. Then I have the team who’s gonna have to keep doing those reports while we automate.
So how do we keep everybody happy? So for the team, we sat down and literally went through and said, this is what your week will look like. And I guarantee you, I will make you the first team out of the office. If we do this. It’s gonna be two months and it’s gonna suck, and I’m not gonna lie to you. It’s gonna be painful for two months, and then it’s gonna be amazing. Got them on board. Then I went to start at kind of the top of the CFO and said, I want to come to your teams. I’d like to move us to an automated solution. This will be a great product that you are gonna be able to go to the CEO and say, look at this automation that my team is using, and my team is the leader on this. And second, I’m gonna give your team more capabilities.
’cause whereas today they might be getting a three or four page deck, they’re gonna have the capabilities to drill down to the point of, if I click on a variance, I could even see the ledger detail and what’s driving it, right? I’m gonna be able to answer more questions than today and everyone will spend less time on this process. And he said, yep, sign me up. So we developed this process using, uh, clicksense dashboards. We’re able to put this into one single dashboard product that updated entirely automatically. That dashboard product would push out at month end. Once we cleared and validated the data and got accountants confirmation, the books were closed, would push out automatically to the leaders. So if they didn’t want to go physically into the dashboard, they would get the top sheet just in their email automatically. And my team was done by 5:00 PM at month end instead of nine or 10:00 PM Um, everybody was just awestruck at this.
And these dashboards then quickly rolled out across kind of the rest of all of the cost base organizations, uh, with a very similar process and similar adoption. Always a couple of holdouts, a few people were, we’re still not big fans of it. Once you get down to one or two, it’s pretty easy to get senior leadership support to say, you know, Hey team, we’re just gonna move forward with the dashboard. That project alone, once it’s spread across everybody, I mean, that was nearly a thousand hours a month of effort at a company of that size by going to these automated dashboards that one person could maintain. And then bonus, it wasn’t just about my team not having to do that work. We were spending 90% of our time helping research variances and make it better for next month, taking things that happened and improving our forecasts, right? We weren’t just sitting around doing nothing. We were working with the business to improve things and make our forecast better. And that’s the real sweet spot because that’s what gets you promoted. A lot of times you’re gonna find these automation projects by themselves. That’s not what’s gonna move the needle in your career. Coming back and adding $2 million in revenue, saving a million dollars of cost, making your boss look good, that’s what’s gonna move the needle for you.
Glenn Hopper:
I know exactly what you’re talking about when you get stuck in this sort of manual workflow, it’s like you’re on an assembly line and work is just coming at you and you can’t step away or step above the assembly line and figure out how to address it. You’re just grabbing the items and doing your work. And I think people fall into these traps of like, well, you know, the first time we were asked to do this report, we, you know, we just put it together as a PowerPoint or whatever. And the, uh, the, the deck just kept growing and growing and the needs, you know, it, it just bloated. And it said, well, we need this. We need this. And instead of stopping, backing up and saying, okay, <laugh>, let’s find a, a better way to make this, they just keep adding onto it and you just, it, it becomes more and more work.
And when you’re in that mode of just data entry and putting that together, you don’t realize that you’re not adding any value. You’re just aggregating information that is not, you know, it, it doesn’t make you a strategic partner, it makes you a worker be the assembly lines get replaced by automation. But the question is, if <laugh> if that’s replaced, then what do we need? And, and like you said, it is those strategic thinkers and ones that are providing value. And I think there’s just so many things that we do every day. If you don’t come at ’em with an automation mindset, then you’re just gonna do ’em the way you always have done. And it’s gonna be with AI and everything else coming along, it’s gonna be harder and harder to justify your position when you are an aggregator of data. And I guess with all that, I mean, I, we can’t go through every aspect of fp and a, but maybe let’s take a couple of examples, um, that people are still doing manual today.
And so, you know, being a a data rails podcast, I think forecasting is one, uh, that people still are, are doing plenty of manual work in. And I’m wondering maybe just, we’ll look at forecasting first and then maybe a couple other areas, but are there some common pitfalls that, uh, you see in manual forecasting and or if some examples of ways that, uh, companies could use automation to have these quicker turns of forecasts. And, um, you know, whether it’s the annual budget or just updating the forecast quarterly or, or, or whatever it is, any automation tips or hacks you’ve seen there,
Mike Dion:
There’s three places where people really get into trouble with the manual forecast. That automation’s great for. The first one is, is just data overwhelm. There’s so much data that they don’t know what to do with it, and they’re spending their time being as I would say, data janitors not actually working on the forecast itself. The the second thing that I see teams just run into constantly is errors when consolidating and that they don’t have a process that pulls things together. That’s where systems like, like data rails are fantastic because the, it just consolidates automatically for you. And if you’re submitting something wrong, it’s just either not gonna go in or it’s gonna fall out where you can see that happen, right? Consolidation is just continuous errors. And then the, the third one I see is that you get so buried in the details, you can literally miss outliers in your forecast because you’re trying to just tweak the things that certain leaders care about.
And there’s an automation solution for all of these. If you think about data overwhelm, I truly believe no one should be manually sourcing the data for their forecasts. That is, if, if you’re still working in the Excel world, that is a power query goldmine. And everybody should be running those processes through Power Query. It saves your steps. You can pull your data in, you can do your mapping. You should not be doing that manually. If you have a system, the system should be connected to your data sources, or you should have standardized upload templates where you are not touching it. All that should be done in the system, whether you need to do it for yourself in Power Query or you need to do it in a system, an important point, don’t wait for people to let you automate, right? Power Query is in your Excel.
You don’t need permission to do this. You’re not talking about getting a system. Don’t wait for someone to tell you. You can just start doing it yourself. From a consolidation standpoint, if you’re working in Excel, again, you can literally have Power Query pull these together. You can even at a most basic level, just have standardized sheets where Excel can add them up and process them, whether you do it, you know, with formulas or with macros, or if you’re doing in a system, which is the best scenario for consolidation, have these processes to remove errors and then build checks in for goodness sake, right? Automation doesn’t mean running an AI tool, it doesn’t mean code. It means having the computer do something you’d otherwise do. Having Excel check your work is automation, right? That’s what automation is, just something you don’t physically have to do. So take advantage of these checks and build that in.
And then on the, the data on the outlier side, I’ve been spending a lot of time testing chat, GPT. And if you upload a data set, say, look for outliers in this data, what might I be missing? And it’s gonna go through, it’s gonna scan it and say, you know, this day the data doesn’t make sense. You might wanna look at this. That is AI’s sweet spot. It’s not that good at giving you a really good usable forecast. It’s phenomenal at telling you where you might be very wrong or where the data is is off from trend and where you wanna look. It is, it is phenomenal at that. Use it, run it through there, run your own forecast. Your value is delivering that great forecast. But AI is great as a second set of eyes to make sure you’re not missing anything. Computers will process data at that scale faster than we ever can. They always have been able to, right?
Glenn Hopper:
Yeah, I love that. So whether it’s variance analysis or finding correlations or whatever, I mean, you can run a, you know, you can build a correlation matrix in in Excel or whatever, it’s fine, but it’s why would you do that every month? But if you, it’s a quick thing to, to ask and, and, and because like you said, uh, machine learning is designed for this, but run it, you know, running data through and say, you know, look for correlation between any of these categories or whatever. And yeah, I, I know correlation isn’t always causation, but it, you know, if, if it’s sort of, I guess p hacking your way into finding some, some additional insights, but letting the AI do what it’s good at and then getting insights that you may not see and, and you’re exactly right because the focus is, it’s like, you know, you have the official KPIs that you’ve defined and then maybe there’s somebody on the management team or somebody on the board or whatever that has very specific things that they’re always gonna ask you.
So you end up focusing, you know, laser focusing on this certain area. And sometimes you don’t see the forest for the trees because you’re so focused on that. And I think that that is an area where, you know, outside of automation, just having AI come in and be another set of eyes looking at something and, and catching something maybe that you wouldn’t. Now a lot of times if you, you know, ask AI to analyze data that’s an experienced fp and a person, it’s not gonna come back with any like, you know, earth shattering insights that you wouldn’t have have figured out from looking at data for years. But it is nice to kind of have that, that AI partner there.
Fp and a today is brought to you by Data Rails. The world’s number one fp and a solution data rails is the artificial intelligence powered financial planning and analysis platform built for Excel users. That’s right, you can stay in Excel, but instead of facing hell for every budget month end close or forecast, you can enjoy a paradise of data consolidation, advanced visualization reporting and AI capabilities, plus game changing insights, giving you instant answers and your story created in seconds. Find out why more than a thousand finance teams use data rails to uncover their company’s real story. Don’t replace Excel, embrace Excel, learn more@datarails.com.
What about variance analysis, something that we do every month when we close the books? Are there ways that FP&A teams can leverage automation to kind of simplify the variance reporting and make and, uh, maybe take it to a level where they’re able to add that strategic value there?
Mike Dion:
Yeah, absolutely. So the first thing, just like with forecasting, you should not be manually getting data. You should have a process power query your, your reporting system, whatever it is to get that data automatically. It’s gonna be cleaner, it’s gonna be accurate. You’re gonna know what you’re getting, right? That is the biggest thing. The second thing is you should not be manually building any reports. I still see so many professionals in the trap that their deck, you open to the first page and it’s a wall of numbers. It’s a giant detailed p and l, and they just go down the rows and talk about the variances, right? The value we add is in telling people what they can do about it. Yes, you need a snapshot, you need, here’s revenue expense, et cetera. Of course you need that, but full p and ls belong in the appendix.
You need to be diving into what matters and what has changed in that month. And that’s where having a tool, like a dashboard, whether it’s in a, you know, a power BI or in your own financial systems, having those dashboards where you can drill in, find the variances, and then really focus on that. If you’re putting a deck together, it should be bringing insights. If it’s a wall of numbers, you can get that and you should be getting that from a dashboard. That’s absolutely critical. I’ve been testing with AI tools, um, I’ve used both copilot and and chat GPT for this of just running my variances in the system and saying, based on my prior month comment, which I trained it on, and based on the variances you see here, what should I prioritize researching? I very much will not use AI to actually write my variances.
’cause that is, that’s where we need to come back with value. All AI can tell me is what changed. But what it’s really good at is helping me prioritize. So I don’t go too far in the weeds on any one topic. It’s gonna say, well, last month you talked about this, the variance is identical. It’s likely the same driver. If you look at this, this is a new variance. This is off trend. This is a high priority to research. And it’s very good at seeing those things that have changed. And it’s really good at remembering what I’ve done before, right? Which, which we do this month after month, we can kind of forget. But if I say, here’s my last nine months of commentary and it sees that I’ve said the exact same thing about labor for nine months, I probably don’t need to do much research there, but all of a sudden cost of sales percent has skyrocketed, five percentage points.
I can go look at that. Now, clearly that’s a bad example I would know to go look at that. But if you go, the more you go down into levels of detail, the more AI is able to just pull out, here’s your workflow for you. Go do the work. Don’t use me for Epic. Go do the work and follow that workflow. Um, and that’s a really fantastic way. And most importantly, when you automate, don’t just clock out and go home, spend time with the business, right? What matters? These, these reports, the variance analysis, it’s a tool. It doesn’t change anything. It doesn’t add value in and of itself. What adds value is helping the business change course on things that need to be changed or sustain things that are working. And you need to be spending time with the business to give them that value to bring back those insights. Otherwise, you’re just spending time doing things that don’t move the needle in the slightest. It’s a tool, it is not the end result.
Glenn Hopper:
Absolutely. It’s only as good as the human in the loop. And I still think, you know, the, the domain expertise that we bring to it, like the example I always give, and I apologize to listeners who’ve probably heard me say that before on on this show, is if you don’t have the domain expertise of finance, you don’t know the right, you know, you, somebody could hand you financial statements. You don’t know the questions to ask, you don’t know the difference between, uh, you know, cogs and other expenses or EBITDA and net income and operating income and you know, all the things that we look at and you know, really how to analyze it. Uh, you know, the way to look at capital spend versus, you know, OPEX and all that. So it’s, um, you know, you have to have that domain expertise and so you know the right questions to ask.
And rather than just offloading all those questions and everything to the ai, you just get that much more powerful if you know the right questions to ask. And you keep, you know, you use it as an aid to make you more efficient or, or think deeper or be able to spend your time, whether it’s AI or just automation in general. But that’s, you’re getting at the heart of it of how we overcome. And I don’t know how many people still think of finance as a, as a cost center, but you know, we overcome that sort of historical stigma that we’ve had of not providing, you know, true value to the organization on a monetary level. We’re just, oh, it’s a cost center. You’ve gotta close your books. Just like you’ve gotta pay to <laugh>, you know, use software. You mentioned dashboards and also thinking about that wall of, of numbers.
And I’ve been in so many businesses where they haven’t established like sort of the, the clear KPIs. These are the KPIs that matter, and these are the ones that are actually, you know, levers that we can pull, that we can make a difference in a business. And so many times there’s these massive dashboards that are just, you know, chart after chart, after chart, page after page. And there are so many tools out there right now that can give you these real time dashboards that I worry you, you also mentioned information overload. There can be so much data out there and these dashboards, if there’s too much information, it becomes like billboards on the interstate and you don’t know what to look at, what to pay attention to. You just, there might be one or two of ’em that you kind of glance at. But with all this power in our hands now and the ability and, and more and more of, of what we do, you know, kind of moving to this realtime close are certainly, you know, very fast closes for companies that aren’t in that realtime phase. But what do you think are the components of a, of a great realtime dashboard and are there specific tools or technologies that, that you like for building them?
Mike Dion:
Yeah, the biggest thing which goes counter to the most of the dashboards I’ve seen is for a successful dashboard, you should put as little data into it as possible. Absolutely as little as possible. The less data you put in, the more successful your dashboard will be, which is counter to most people who wanna put an entire, um, snowflake or SAP database into their dashboard. It’s slow, it’s overwhelming. It doesn’t answer the question. Small focused data sets focused on actionable items. What are actionable items? The way I think about KPIs from a dashboard perspective is almost like what are our forecast drivers? What are the things that we can change or influence that will drive the business forward? I like to build my dashboards with almost like a, like a top sheet, right? Where you’ll go in and these are the key KPIs. If I’m responsible for managing an hourly cost base, my KPIs are gonna be, you know, what is my overtime percentage?
What is my um, hours? I’m not gonna put labor rate on there because I can’t control it, but I can control the amount of overtime run, I can control my double shifts, things like that. What are my actionable items then on the back? So you can click in that and say, okay, I have this labor variance, I wanna, you know, I wanna research further. You can go back and you can get a sheet of here’s all the labor information. Then you can, then you can see if you wanna research it and follow that thread. But what is the truly actionable item? I as an operator can raise or lower my overtime spending and that will change my labor rate. That will change my labor cost. That is the KPI, not the labor rate. So building with this few kind of key KPIs for what that leader needs, and then having the ability to go back to other boards in the back of the dashboard if you want additional information or detail that you are looking for. That way I, as the finance professional, not overwhelming you in the close, the data’s smaller. A dashboard is lightning fast and it gets used because if it’s something that I can change, it gets used.
Glenn Hopper:
I love that you said that, ’cause I, one of the things I’ve been planning around with and what, what I think for generative ai, a lot of ways that we’re gonna see it used are software providers, your, you know, NetSuite and, and data rails and uh, all, all these tools out there are going to implement generative AI in as a way to interact with the data. So people aren’t gonna be building their own large language models and they’re not gonna have to everybody become an AI engineer, but it’s just, it’s gonna start being included in software. And one of the tools that I’ve built, you mentioned low code and no code tools. I’ve been playing around with repli lately, which is just prompt based building apps, but incorporating AI analysis into, you know, financial statements. So the idea is you close a month, you get, you get the financial statements, you upload ’em in, it does a first pass on ’em, does kind of the charts and graphs, kind of what we would do when we get financials, you know, when when they’re closed and we do our first pass on ’em, we do that kind of variance analysis and look at how we’re doing, you know, compared to budget historically, all that, but then provide sort of that high level of, uh, footnotes on the financial statements.
And I, I think that that’s how ai, a lot of us are gonna see that we’re gonna start seeing that incorporated more and more into ERPs and CRMs and, and GL tools and all that. It’s, that’s how most of us are gonna interact with it. And I think that goes along with the very simple high level dashboard where if you wanna drill in, you know, you open dashboard and you can drill a couple levels deep sometimes to get that additional layer of information. But when you hit that dead end, historically that has meant, um, now I’ve gotta go to the BI team or to fp and a and say, I need a report to see this. But I see in the very near future as people figure out how to incorporate generative AI into these systems, instead of having to wait for a person to pull that report, you just ask your dashboard the question and get that level of detail there. So then it becomes even more important to minimize that top level to really make you focus on what’s important. But you know, you can dig in when you need to.
Mike Dion:
Absolutely. We still need to be the guides. That’s be, that’s becoming the role is not just strategy, but the navigator and navigating the business through the finance process and through data and through the value that we can bring to the table.
Glenn Hopper:
I guess going back to the generative AI you have written and you, you share it with me, the ultimate guide for AI prompting for finance and you know, I, I hate the idea in general of prompt libraries because it makes people, I mean, I think it’s, it’s a good introduction, but really people need to know how to interact with generative AI and not memorize, you know, plug and play libraries, even though, I mean, if there, but that said <laugh>, if there are things that you do every month and you’ve got, you know, you know, the perfect prompt, but I just saw a study yesterday that, you know, the exact same prompt is, it’s generative ai, so it’s generating new content. You’re not gonna get the same results every time you do it. So, you know, another reason that prompt libraries aren’t my favorite. But that said, with even without a prompt library, we need to know a basic approach. And there’s a million of ’em out there, but a basic approach for how to interact with these and the fact that you’ve made one targeted for finance. Can you tell me a little bit about the guide?
Mike Dion:
Absolutely. So just one anecdote on on what you were saying about generative AI and the prompts not being continuous. I was actually reading a study recently that, uh, AI models provide worse responses in July and December, if you prompt ’em, because human content is weaker in July and December. So they’ve actually done studies that it, it’s not as effective. And AI models, which in theory should be able to read all the text you give them will focus more on the beginning of text and the end of text. Just like people, like they’re building people like tendencies over time, which also influences the prompt you’re getting just like you prompt a person that’s different. The generative AI is constantly learning and it’s picking up some of our bad habits, which I thought was just, um, fascinating and also a little, little scary at times. So when I wrote the AI guide for prompting, what I really wanted to focus on was, as you said, teaching people the methodology, teaching people how AI tools work, how they think, how they’re trained, and how you can structure your own.
I do include in the back, I include some example prompts, but very clearly as a way of kind of getting inspiration for what you can do versus, you know, copying and pasting, right? It’s, it’s, here’s some things you can do, you may not have thought of. So in that, I have a, a spark framework which teaches you how to construct a prompt anytime. And I think the two most important things about that process are the first and the last step, which are also the things that most people forget. The first one is to set the scene. AI tools perform very well if you tell them how to behave. So if you’re starting a finance prompt and you tell the AI tool act as a financial analyst, I found you’ll continuously get dramatically better prompts by telling it what you expect of it. And then the last thing, again, AI models are trained to talk to us like a person would talk.
So I end every prompt by saying, ask me any questions you need to provide the best responses. So it gives me the extra boost that if the AI doesn’t fully understand something, instead of guessing, I’m telling it to ask me. And it will, sometimes it will just run it and it understood me. Sometimes it’ll come back and say, could you con clarify X, Y, or Z? Those two steps are so commonly missed. And the quality of my prompts, no matter how rough I am in the middle with the actual task and the output, I want back that piece of telling it how to behave, it puts it in the mindset of a financial analyst. And it will, it will not make guesses of what I want. It will ask me what I want. That is really the goal. And this way, if you learn this framework and if you build that into every prompt you write, you can work at any system.
’cause I don’t focus on any specific system in this guide. It works for all of them. Claw, jet, GPT, copilot, Gemini. I want it to work with whatever you’re in front of. To your point, if people start putting these tools inside of software, it’s still using these same really three or four large language models. These prompts will work for you and you’ll know the secret to success for everyone. And then it’s easy to train other people. If you wanna teach people how to do this, you can easily train them because you have this kind of cohesive framework of how to ask AI for what you need.
Glenn Hopper:
Have you experimented with any of the new reasoning models? Like whether it’s uh, you know, oh one or oh three from open AI or I, I guess Claude 3.7 has some sort of inherent ability like it, it will determine how much it’s gonna reason. So rather than the historical way that they’ve done where they just immediately spit out a response, these reasoning models will stop and kind of think about it. The reason I I’ve brought that up is your prompt is, is sort of forcing that reasoning step even on a standard model where you’re saying ask me questions if you, if you need more information. And I think that the reasoning models are, are ideally doing that on their own. I was wondering if you’ve, if you’ve tried out any of those and seen any impact from your prompts or any, any difference between the two? What
Mike Dion:
I’ve been using reasoning models for isn’t those kind of one step tasks. ’cause typically you wanna, you wanna give ai, if you’re using traditional models, one step, I’ve been using the reasoning models to test kind of the agent, the AI agent functionality where it can do multiple steps. And what I’ve been working with a lot there is giving it a prompt. So like I’ll say like I, I still follow the telecom industry closely. I’ll say act like I am Verizon and I, I need to understand how at and t and T-Mobile are performing in quarters relative to me. So I’ll say, first of all, you know, here’s the 10 Q for Verizon research study, understand this and ask me any questions you have to make sure you’re on the right track. Then go out, get the information from at and t, get the information from T-Mobile, study it the same way you study Verizon and then complete a, a SWOT analysis.
And then you’ll see it kind of reiterate that SWOT analysis, um, that kind of a multi-step process is really good for the reasoning models. With my prompt structure for single tasks, I don’t see the need to go to the slower models ’cause they’re, they’re pretty slow. If you go to oh one, it’s, it’s a lot slower, but it’s the multi-step tasks. I can kind of run it like I’ll start the AI agent running over on this screen. I’ll pop over my other screen, get some work done for things like that. I love the reasoning and it’s really good at, at keeping its logic through different steps and following the original ask, while it runs multiple things I just haven’t found for a single process, like, hey, here’s this file, read it, summarize it. I haven’t really found it’s been necessary to use that, um, kind of model for that.
Glenn Hopper:
Yeah, makes sense. Well you’ve obviously been experimenting a lot with AI and I think a a lot of our listeners are too, and I know I I probably sometimes err on over talking about it, but I think it’s, so I, I’m a fan boy I guess of, of, uh, the, what, what’s possible with generative ai. But I think all of us now kind of see the writing on the wall with this. But there are issues still, uh, e even the, the brand new frontier models are, there’s still issues with hallucination. And while you might be able to get away with hallucination in a field where there are gray areas, you know, in marketing, if your marketing copy is slightly off, that’s not the end of the world. But, you know, the trial balance has to balance. There are no gray areas in, in numbers, so we need to be specific.
So, you know, I think maybe adoption is a little bit slower, but people are starting to lean more and more into it. And I’m wondering as you’ve, and, and like you said, you’re not offloading your job, you’re using as a tool to help in your job, but how do you see, especially as the models, as the technology grows, the models get better and, and with your focus on automation prior to ai, maybe with that lens, how do you see AI kind of transforming fp and a workflows in the next few years? And with that, uh, I guess maybe a two part question. What advice do you give to people working in fp and a now who want to get started and, and kind of thinking about that, whether it’s what they need to be focused on learning how they should implement it today? I dunno, that’s a lot. Maybe we, we need to pause between, but what do you think about what’s, what’s coming in the future?
Mike Dion:
So I think one thing I’m spending a lot of time thinking about and and concerns me for the future of the profession is I think we’re gonna see a lot less entry level roles because it’s the entry level kind of that analyst role where we’re doing a lot of the automation. I don’t necessarily see the lead analyst, uh, individual contributor managers. I don’t see kind of those like expert level, individual contributors really shifting a lot. ’cause there’s this technology needs a lot of support. But I do worry about entry level roles and I think that we have to be cautious about the pipeline. And I think that same is true for accounting as this technology rolls out. That’s something I’m really thinking about. Something I encourage, I’m encouraging a lot of people in college now to study data science because that will give them another way into finance if they are not able to find an analyst role.
So that’s something I’m really encouraging. Everyone will need to have some level of understanding of data science, right? If you’re working with AI tools, you are so much more effective. Whether even, you know, even if you’re not fully designing it, even if you’re not, you know, coding things, if you understand how the tools work, you’ll get dramatically better responses back from it. So I’m encouraging everybody to spend some time, even if you just are, you know, watching a couple YouTube videos or listen to a podcast, start understanding data how it works, not just the financial or GL data, right?
You also need to be very focused and you have to get outta this mentality, like I was saying earlier, that the reports and the forecast themselves are what matters. What matters as I would say it’s, I call it the real deal. Three things, increasing revenue, decreasing expenses, and making your leaders look good. Those are the three things that move your career, not the reports, not the forecast tools to get to that. So as the workflow goes, you have to find ways to add value and you have to find ways to add value that people above you notice otherwise you are gonna be one of the ones being automated outta the workforce.
Glenn Hopper:
Yeah, absolutely. And I love what you said about data science because to me it goes hand in hand with what I said about your domain expertise in finance. If you understand how to read financial statements and you understand the, the role of, of fp and a, you know, the right questions to ask versus just handing someone off the street financial statements and asking them to analyze them. The barrier to being able to do data science in the past used to be where you have to know Python and you have to really understand all this at a, at a developer and engineer level. But with generative ai, you can now interact with your data just with your natural language. So your natural language becomes the new programming language. But if you don’t know the basics of machine learning of are we doing classification or prediction or clustering, you don’t know the right kinds of questions to ask.
So like while you have this powerful tool that could help you with customer segmentation and, you know, churn predictions for your forecast and all that, if you don’t know the right questions to ask or what it’s good at and what it’s not, then it’s a very powerful tool in the, in the hands of <laugh> someone who doesn’t know how to use it, it, you know, all that power is sort of negated. So you’re, you’re spot on there. And I did talk to someone the other day who got an MBA and analytics degree at the same time, and I, I’ve heard about those degrees for years, but I’m, I’m kind of surprised that even today I’m not seeing as many of ’em out there. ’cause I think the two go hand in hand. And I think like people kind of on their own are gonna have to take, take this on to learn data science.
And it doesn’t mean, you know, we, we went to school to be finance professionals. I, I’m not saying we all have to become machine learning engineers, but like my last book, AI Mastery for Finance Professionals, it’s very little about generative ai. It’s mostly about just, this is how machine learning works. These are the applications of it. Because to your point, if you wanna survive as things are being automated, like those entry levels are gone, if you want to provide the value, you have to understand the technology, you have to be able to use it. And then that’s how you’re adding value to the organization is by having these powerful tools that you can use appropriately to make you better at your job.
Mike Dion:
It’s so true. You, you never want AI to guess, right? Like I can get chat GPT even to spit out a forecast, but it gives me a terrible forecast unless I tell it how to forecast. If I say use time series analysis with seasonality, it gives me a great forecast. It does very well with that. If I just say gimme a forecast, who on earth knows what I’m gonna get back? Um, even with machine learning, using ai, you can quickly overfit your data models. And if you don’t know how to avoid overfitting, you’re going down a path and thinking you have a usable product when you don’t. And that’s why that’s just so critical. It’s still needs, direction and guidance. You don’t want it to pick for you. ’cause I may pick wrong.
Glenn Hopper:
Yeah, absolutely. Like I could talk AI all day and I, especially when I have someone like you on the show, I, I feel like we could geek out on this for <laugh>, uh, for a couple hours. But, um, I wanna get back to one more thing too because I, you know, everybody’s talking about AI and people are conflating now AI and automation, but you’ve been doing automation long before AI was, was part of it. And I think that while we, you know, AI is the shiny new thing that we’re all kind of chasing right now, we can’t overlook just systems, process, software, sort of the, the automation that’s been around for years. And I think getting back to that, if you’re talking to other fp and a professionals like, and they’re thinking, yeah, we got a bunch of annual processes, I don’t even know where to start. Are there like two to three like low hanging fruit kind of processes that you would recommend? And, and maybe this is too general a question, I don’t know, but are there, you know, some specific processes that you’d say, Hey, take a look at this and see maybe there’s something you’re doing today that you could automate?
Mike Dion:
Yeah, so this will sound a little, little counterintuitive from a finance guy, but I think if you’ve done no automation, the first thing I encourage people to look at for automation is their email inbox. And there’s two very specific things here. There’s kind of the, the bills and invoices and then there’s approval workflows. You can take two approaches. One is a great start to automation is literally learning how to just use rules and filters in your Outlook or Gmail inbox. That’s automation and it, it’s so natural, it’s so easy to do. It will save you so much time, it gives you confidence. The second is you can quickly set up things with approval workflows. You can do it with, you know, SharePoint, you can go do it with Google Drive where approvals will just automatically route in and out and you can use that instead of tracing emails through 20 people who need to approve something, it has statuses.
And that’s a really easy, approachable way before you get into places where you have data, where you have to consider, you know, data cleaning and all those things. It gives you a really fast, quick win and immediately saves you time even though it’s not kind of a core finance process. But quick wins are so important. From there, the next thing I would really focus on automating is data collection and consolidation. We all have Excel in our computers. We all have Power Query and Power Pivot and they’re free and they’re sitting there ready to use. If you are a fp and a professional, you have a use for Power Query and Power Pivot. I, I guarantee it. Unless you are an executive who doesn’t touch Excel, you have a use for Power Query and Power Pivot and that is the first thing you can learn. And I have literally developed programs where you can teach people this in each software in 20 minutes. You can get the basics, you can get up and running and it will save you hours. That’s number two is just look at your data, look at the reports you’re building and automate the backend using Power Query and Power Pivot.
Glenn Hopper:
And those lessons, are those on F nine Finance? Those
Mike Dion:
Are on our YouTube channel. Yeah, they’re on F nine finance. YouTube. Last one I would say is monthly reporting decks where you don’t have a lot of variation where it’s kind of what, you know, the standard ones, the KPIs that you’re doing month after month. Put it in a dashboard. If your company has an Office 365 subscription, there’s an 80% chance that you have access to Power BI for free. Pop it in there. Again, my biggest thing is don’t wait for permission. Go see what is there. I know everyone of everyone listening who has Excel has Power Query and Power Pivot, and almost everybody who’s on Office 365 for enterprise will have some type of power BI access, use it, it’s free, get started, get these small quick wins and then it just compounds from there. You get so excited about what you’ve done that it just, it’s just a snowball effect from there. And bonus, you become the go-to person on the team and it’s really good to be
Glenn Hopper:
The automation person. Absolutely. Being the automation person is so important because right now, executives, senior management and companies, all they’re hearing is, you know, the, the boards, the investors, whoever it is that’s, that’s driving the company or you know, even from the CEO, they’re all hearing about AI and automation and you know, we’ve gotta get more efficient. And if you’re the person who can actually make that happen, whether it’s AI or not, I mean, you know, everybody’s kind of AI washing everything that they do now and saying, oh, that’s AI when it really, it’s just a rule based, something that’s been around forever. But if you’re the automation person, then you know, that is how you, uh, show that you’re valuable to the organization and make ’em more efficient. So timing is, is perfect on that. And I guess on that note, before we kind of get into the personal and fun questions, uh, part of the show is with everybody talking about AI and automation right now, is there a common misconception about automation or AI in, in finance in particular that, uh, that you, you’d like to debunk?
Mike Dion:
There’s still a belief among almost everyone I talk to that AI and automation requires Python are advanced coding skills, you know, $10,000 a month software, that that is still the, the understanding from most people. I’m starting to see, you know, more companies are starting to get copilot. It’s starting to get a little bit more approachable. No code is becoming bigger, but that’s still just the general perception from the corporate fp and a population that it requires these advanced coding skills. ’cause you know, five to 10 years ago, we, all we were talking was Python. That’s what I really wanna break again, if you set up email filters to not have to touch emails, you’re automating. If you use data validation to make sure you’re, you’re not putting crap into your forecast. That’s automation, conditional formatting automation. I teach like all these tools. These are automation and you need to think about it that way.
You are automating if you use these tools, ’cause the computer’s doing work for you that you’re not highlighting every cell yourself. If you come with that mindset that there’s all these tools out there and it’s just about, you know, what am I spending time on that I could spend less time on? It becomes so much more approachable and people have an easier time getting started versus I have to go take a class to learn Python. I have to get permission to buy the software to work with Python. I need it permission to even use Python on my computer in the first place. It becomes just this overwhelming mountain versus hey, I’m gonna open up Outlook and set a couple rules for these emails I don’t need.
Glenn Hopper:
Perfect, perfect. And quick wins right there too, so. Exactly. Alright, let’s, let’s bring it home with our last two questions that we ask everybody. So first one, what’s something that not many people know about you?
Mike Dion:
Uh, I was actually an improv comedian in college, which is not normal for finance people, <laugh>. Um, but, uh, I grew up loving, whose line is it anyways? Um, so I spent, I spent four years on stage, um, making people laugh before, you know, going into finance and making people cry <laugh>.
Glenn Hopper:
Right. Perfect. Perfect. Maybe the takeaway from improv is like the number one rule is never end in a no, don’t stop the flow. So you just keep the, uh, you know, everything is positive and, and stay moving. Maybe that’s something that you can apply in your finance as well, <laugh>. Oh, absolutely. It’s never be the yeah. Best way to work with clients. Now everybody’s favorite question. What is your favorite Excel function and why?
Mike Dion:
I have kind of a, a more obscure one. Mine is equals RRI, um, which does, uh, basically CAGR calculations really fast.
Glenn Hopper:
Oh,
Mike Dion:
I know. For years we all just did our CAGR manually and then someone pointed this out to me and it’s just a, it’s a copy and paste CAGR that you can even build it to be dynamic. So it counts the number of periods by itself and will move the present value and future value if you use Offset. I love that formula and I almost no one knows it. Like every time I pointed out to somebody, uh, I’m always the first one telling them it exists. So that’s my favorite. That’s
Glenn Hopper:
So, yeah, that’s, I love that you said that because I just did a CAGR calculation yesterday and I was so annoyed, like having to go through <laugh> and type all that in, like, just in on, ’cause I was doing it in multiple cells and as, as someone who was an executive for a while, I’m not, I’m no longer cutting edge on, on Excel stuff, so it was a, a very clunky thing for me. So now I’m gonna go take that same spreadsheet and go apply that <laugh>. So super excited to try that out, man. This has been great. I guess just last thing before you go, uh, and we’ll put, we’ll put all this in the show notes and everything too, but how, how can our audience connect and learn more a about you and F nine finance and, and, and just kind of follow what you’re doing.
Mike Dion:
Sure. The two best ways I put out a, like a ton of free content for people on automation. Um, we’ve got our newsletter, the Finance Automation Insider. You’ll find that on F nine Finance. That’s just weekly tips, just free stuff on here’s how to automate processes both for work and also how to automate your career for success. Um, and then I do a ton of automation content, live videos, tutorials, walkthroughs on our YouTube channel, which is also AD F nine finance. Um, just trying to, to get the word of automation out there and keep giving people free tools to set their career up for success. Awesome.
Glenn Hopper:
Well Mike, thank you so much for coming on the show.
Mike Dion:
It’s been an absolute pleasure. Thanks for having me.
