Building Your 2027 AI Budget (w/ Justin D'Onofrio, Accordion)

0:00We're in the corporate cards are issued. There's no expense policy, no T &E budget. And some people are teenagers at the mall step of the of the AI journey here. Oh, man, if I could have been at Crossgates Mall with my parents credit card in the 80s, that would have been dangerous. But I'm right in the thick of this. I literally just came from a board meeting 24 hours ago where we had spent a half an hour on this conversation. And we did not have the answer for it, which was the CTO saying, let me tell about the great stuff we're doing and the great stuff we're going to do. And the board asking, well, how much is that going to cost and how do we know it's working?

0:37It's not venture capital. It's private equity. It's the private equity funcast. Pour yourself a drink and have a seat. It is our goal to be useful at Parker Gale and in the funcast rather than smart. And today is probably one of the more useful podcasts we've done this year. Everybody is about to head into budgeting season. And how do you budget for AI? Well, it's a question on everybody's minds. So I've got Justin D 'Onofrio, who's a managing director of the performance management team at Accordion. Accordion is one of our go-to service providers. And Justin and I are going to walk through everything they've learned and the products they've built and the tools they have to help people figure out how do you budget for AI for 2027.

1:24So Justin, welcome to the FunCast. What did I get wrong? And maybe fill in the stuff that I missed. Thank you for having me on. I would start by saying my day-to-day and for the past 25 years has been spent at the intersection of finance and technology, specifically within the FP &A space. So I feel like every conversation I'm having these days is the intersection of AI and FP &A. and the most common one that we're getting question-wise is, how do I budget for AI? Super topical. Every finance team is in the budgeting cycle. And candidly, a lot of folks just don't know how to approach it and don't know where to start.

2:03So hopefully we can demystify some of that, some of the tips and tricks that we've come up with at Accordion. Great. And if people want to go back a few episodes, if they're listening now, your colleague, Kyle Romer, and I spent an hour talking about that intersection between AI and finance and what tools finance teams and CFOs and private equity are using to move things faster, do things smarter, do things more automated with AI and kind of broke down like what's actually working, what's not working and how you guys are working with clients on that. I think already several thousand people have already seen that.

2:41So if you haven't watched that episode or listened to that episode, I'd say go back and do that as well, because you're behind if you haven't listened to that episode, because we basically gave you the playbook for how AI is infused inside finance. So here's the issue, right? You guys just did a report on this, which we're going to link down in the show notes, which was basically called the AI budget nobody knows how to build is due soon, eight steps to get there. So there's a link in the show notes for that. Also go to Accordion as a website for that as well. What's the issue? 91 % of PE-backed CFOs can't size their AI budget, and 80 % of them think the budget's going to go up, right?

3:17So we're basically having this conversation about spending more money in the next 12 months, and we have no idea how to budget for it. 83 % of portfolios have one or more AI pilots going on, right? And only 18%, less than 20 % are actually tied to value. And 65 % of sponsors, so that's me, say that their data infrastructure, data ownership is missing, which is a key issue because bad data, AI will confidently give you a wrong answer when you're feeding it bad data. So that was the overview. How did you guys get to the study? How do you guys do your work to gather all this stuff? How many people did you talk to?

3:56Kind of what'd you learn as you pulled this report together? And then we're going to dive into the tips and tricks and everything you've learned. Yeah. And I think it's important to start with the broad context of what is happening in the market today. So I think everyone listens to the headlines. The one that was going around the past couple months has been Uber spending their full year AI budget in four months. I mean, we're talking about a sophisticated engineering forward organization not being able to size the prize with AI. In addition to that, you'll see the headlines of Meta shifting their big AI investment to be leased rather than used internally.

4:34So I think that the key takeaway is that even the mature folks, the folks that have been using and leveraging AI and making big investments in it, really still don't have that playbook unlocked yet. When you head down to the mid-market, we just saw a survey from a vendor that about 34 % of mid-market FP &A teams are using AI in their process, but don't know the impact or the ROI or how to budget it. So I think it's really important to just bring the broader context here that this is a pervasive issue that a lot of CFOs have on their minds, and we'll talk about the process to help address it. But we're in the corporate cards are issued.

5:17There's no expense policy, no T &E budget, and some people are teenagers at the mall step of the AI journey here. Oh, man. If I could have been at CrossGates Mall with my parents' credit card in the 80s. That would have been dangerous. But I'm right in the thick of this. I literally just came from a board meeting 24 hours ago where we spent a half an hour on this conversation. And we did not have the answer for it, which was the CTO saying, let me tell about the great stuff we're doing and the great stuff we're going to do. And the board asking, well, how much is that going to cost and how do we know it's working?

5:53And we're all trying to figure out how do we measure AI spend, like we've figured out how to measure the ROI on marketing spend and attribution, all these things. And it's really hard because a year ago, we told all our teams, get AI pilled, use it all, play with everything. Katie, bar the doors, right? Just go and experiment. And now everybody is saying to your Uber comment of like, whoa, whoa, whoa, whoa, like we got a little ahead of ourselves. And this is not just token maxing. This is just clawed on the corporate credit card or a perplexity account on the corporate credit card. You've forgotten.

6:30You haven't logged into it in six months, right? So there's all these things going on, all this spend happening. And now this budget cycle comes, here comes the renewals. And everybody, I think, is going to get a little surprised on the fight over which ones are these renewing, which ones aren't we, where are we spending the AI dollars on which project and which ones are getting starved because they haven't proven it. So this is like real-time stuff. Is that what you guys are getting called on right now? It is. And, you know, it's a new muscle, frankly, for FP &A. The spend for AI is not just a new software line.

7:00It's not a reoccurring fee where, you know, you negotiate the five for 10 % uplift every year. You can approve spend for an initiative, uncap that token spend behind it, and just have a complete budget blowout if you're not managing the process successfully. And again, frankly, this is just a new muscle for FP &A teams that maybe haven't been in the world of cloud spend, where there's consumption-based pricing in the background. So the framework really of starting with an owner is one that we recommend to our clients. Every initiative, whether that's an investment in understanding what AI can do, which we certainly recommend, or it's an investment in improving a workflow, which will add tangible ROI, it's really tagged that initiative, that AI investment to an owner within the business.

7:48Great starting point for everyone. All right. So when we talk and we talk to your team, three questions we have on our mind are like, what does this dollar return? So where's the attribution and how do we measure that? I want to get into that more deeply. Two, what does it cost as usage grows? So these things can get very expensive, especially if you're exposing any of these models out to your customers uncapped. And they could all of a sudden run up the token bill on you. But there's other ways to measure that. And then how has anyone reviewed that number since it was originally approved, right?

8:25So like, hey, a year ago, I told everybody on the team, spend it, play with everything, have fun, let's talk about it. And now we're budgeting for 2027 internally, even here. And I'm saying, hey, guys, let's pick a winner and move on. So walk me through those three questions, kind of what's beneath the surface of each of those. And then, yeah, tell some stories from some client engagements, you know, obviously anonymously. but stories are always a good way for people to internalize what's going on inside your clients. Yeah, and I'll jump back to one of the stats you surfaced at the beginning of the conversation.

8:59Around 80 % of pilots are failing, AI pilots. From my experience, it's not that the pilots are failing. It's just that they don't really belong to anyone. So the ROI isn't defined up front what folks are looking to get out of that investment. To your point, there's no tracking on the spend after the investment is kicked off. And I would just say that the trend has been budgeting for the tool. Hey, we're going to use Claude here. And that's sort of the way the line item is addressed from an FP &A perspective. The change in thinking is really making sure that you're budgeting for the outcome. The tools can change.

9:40There's going to be a new model that saves you money. but making sure that you're thinking about the budget for an initiative AI related in the context of the outcome, not the tool. Yeah, you told me a budgeting question is really more of a question about value in disguise. It's not about the budget. It's about, hey, we're spending what for what. Exactly. And, you know, I would say, too, that the interesting thing is the maturity for budgeting AI doesn't seem to be necessarily linked to company size. So I'll give you three examples of conversations that I've had in the past call three weeks. The first was I was talking to a CFO, media company, and said, what's your AI strategy for budgeting this year?

10:22The answer was, well, we're prioritizing AI spend. I said, that's great. That's something that every organization should be doing. How are you doing that? How are you tracking it? How are you breaking it out so you understand what the spend and the justification is? And the answer was, we're just not doing it yet. So interesting example. I think that's where most companies are at, where they know they want to spend on AI, but they just don't have a mechanism set up to do that. The second example is one you brought up. Services company came to us, pulled up a Claude bill from a recent month. It was many hundreds of thousands of dollars.

11:00And they said, what's going on with this bill? How do we dig into the details of how we got here? They had historically not spent a lot on AI enablement, productivity, so we helped them dive into the underlying causes. But both of these point to the state of affairs today, which is we are investing in AI, but we are not mature enough to manage it. Yeah, and I think we invest in software companies, and you guys have clients across the middle market spectrum. So for us, it's how much AI is in the product and how many tokens are getting burned with our customers using that product, right? So we have to worry about somebody could go run up a huge bill on us using, you know, tokens as part of our product.

11:47And then there's the token usage and other costs to build the product. and then there's just general, you know, clawed perplexity, open AI licenses, subscriptions across all the different departments being used. And then forget, you know, I mean, you can add in all the other AI assisted and enabled software products. As you think about it outside the software world, like where's the exposure there? Is it just like everybody's got a co-work on their desk and only some people use it and some people use it to plan their vacation and some people are actually using it for productive reasons? Like what are the issues kind of outside of token usage for software development?

12:23Yeah, I think we're definitely seeing personal productivity, call it token maxing, still happening. So a lot of companies are early on that we work with, they're probably just deploying Claude for, let's say, the internal F &A organization. And they're in that investment phase of, we just need to understand what this platform can bring to the table from a productivity perspective. And in those cases, again, we always encourage that. But there should be some guardrails around the spend associated with that and whether you expect to get anything out of it. So that's one lever is just personal productivity everybody's investing in.

13:00So, you know, if you think about your budget line items for where you should invest, one is definitely that personal productivity. But again, cap it and have an owner associated with making sure that that spend doesn't get out of control. The second area that we see a lot of investment in is, I think you brought this up at the beginning of the call, AI is as good as the data that it's looking at to make its decisions. So if you don't have a good data environment for AI to make its analysis and outputs more tone, more accurate, then you're going to get garbage in, garbage out. So a big area of investment that we see clients making is sort of shoring up their data infrastructure so that the data AI is using is able to give them valuable insights and improve productivity.

13:49That is super hard, Justin. It is. It is very hard, absolutely. Really hard. I mean, we've got companies who have built MCP servers. And so for internal work, so think of like an intranet on steroids, right? Any employee can ask any question about any topic in the organization. and that data governance piece of it is just really hard to get right because everybody's got a file server with version one version one you know dot one version one f version no this is the final version and the model doesn't know which one to pick right so that gets really confusing we also have another company that has built basically an access layer to all the all the models that they use for all their employees.

14:39So you basically go through a portal that's a company portal, and then you ask it. There's a chatbot there. It will decide which model to use based on the cost and kind of the complexity of the question. And if it already has answered that question, it's been cached somewhere in its internal model and pulls it. So it's zero cost. That's pretty sophisticated stuff. It is. But mostly largely for, you know, just internal employees having it all on their desk using it because, you know, it's a few hundred employee company, giving everybody a perplexity or a clawed license or an API, yeah, you run into these massive bills, you know, hundreds of thousands of dollars that you weren't expecting, which, you know, could be a, you know, basis points on gross margin that are a big surprise for CFOs, which nobody likes.

15:30And, you know, I think we're going right back to one of the comments that we both talked about earlier, which is the Claude usage. That's really the old world or the last year version of budgeting the tool. The real way to think about it now, particularly from a CFO, FP &A lens, is we're budgeting for an outcome. And that outcome might be we expect productivity of the accounting team to improve by 20 % with this use of Claude. That way it makes the investment in case that much easier to talk through and rationalize and track versus we just have a line item for clawed spend. Again, there's going to be some of that based on the maturity of the organization, but most companies now are switching from the tool budgeting to the outcomes budgeting.

16:16Again, to go back to my board meeting yesterday, that was a question asked towards the end. Hey, are we expecting some efficiency gains in the 27 budget by a department, not just out of the engineering team, CFO's response was, yes, we don't know what yet, but each department will be asked to basically deliver some efficiency or productivity gains. My guess is that'll be very hard to measure whether that was AI related or just general better business practices, top grading, et cetera, et cetera. And there'll be different justifications for that. So even the sophisticated companies who are thinking about it, who've been playing with these tools for a couple of years not playing catch up right now, are kind of winging it, Justin.

17:02And I'd say, it's okay to wing it. You got to start somewhere. Well said. Perfect is the enemy of good, I think, with where we're at with AI spend. So you kind of hit the nail on the head, which is companies at least are aware now that they need to start tracking this. Because again, to go back to the opening conversation, the spend is happening, whether they're tracking it or not. The important thing as a good CFO, good FP &A team is to make sure you're setting yourself up to have the business justification and the financial tracking to actually measure what that spend was. So, you know, our sort of framework starts with that first step in the investment process for the FP &A team, call it the capital budgeting process, call it your business case development.

17:46There's a lot in our accordion study that I think is really useful. But the two key points about that portion of the process that I would love to just dive into is one, starting with that ROI, ROIC, MPV view of the investment that you're going to make. So that's a pretty standard finance exercise. A lot of companies do it for CapEx planning. The change here is one, assigning those AI initiatives an owner. Just again, the governance here will be able to help with giving accountability to someone in the organization to track and manage. And then second is the EBITDA lever that you think will come out of that investment.

18:29So is this a play on increasing revenue? Is it a personal productivity play where you expect the workforce to be more efficient? Is it a reduction in cost on another software line? So if there's two things that we recommend every FP &A team do, It's get your ROIC and your capital planning process AI-ized, for lack of a better term, with tagging to initiative owners and EBITDA levers. Yeah, which then, you know, you think of when the budgeting season starts. So everybody's listening to this in the fall of 2026 or, you know, just the fall beginning soon here. You're all in this mode. You're used to asking marketing, hey, you have a million-dollar budget this year.

19:16Tie every dollar to some outcome. Well, we're going to spend this much on conferences, and we expect to get this many leads. This is the top of the funnel. It converts like this. Boom, boom, boom. You know, 10 years ago, 20 years ago, that was hard. It was a little bit of a black art. Today, system standard. It feels like we're in the black arts part of the AI spend. but there are initiatives you can discreetly tie to ROIC in, I think, finance and in certain departments. Are there some low-hanging fruit? Or if you were giving somebody training wheels, where would you start? Which one that somebody listening can be like, oh yeah, I could tie that and tag that to an initiative.

19:55And I could measure it at the end and report against it every quarter. Yeah. And I think, so I always like to go back to, and again, we recommend this to clients, Think about the context of a workflow that can be a small workflow or a really large workflow for the organization and the benefit that you'd capture with that. So one that we typically get engaged for from Accordion is call it optimizing the F &A organization to make sure that they're spending more time on value add work and less time on button pushing and manual work. Pretty typical one, I'm sure all the folks listening can relate to the board deck that's due every month and the painful process that may be happening behind the scenes to get that put together.

20:39A lot of clients will come to us and say, hey, we need to do this every month. Obviously, it's very important to the business being PE backed. It takes us 40 hours a month to put together, to do the analysis, to dive into the trends. And if we can save 30 of those 40 hours, we can now put a clear ROIC case together on why we should invest in using AI to speed up that productivity and that output. So I wish I could say there's clear areas of investment for every organization. There are certainly common trends that I see working mainly with the F &A org. But thinking in the context of a workflow, some that are near and dear to my heart are accounting and finance.

21:21That great example of the management reporting pack, you can quantify the dollars saved through the productivity increases that you'd get, saving that manual time. Yeah. Hey, let's get the finance team more strategic. Let's take the monkey work out of this. How do we close the books faster? How do we do collections in a more automated way? how do we get reporting up to our private equity overlords more systematized? Because the monthly is mostly just a data pack, right? The quarterly gets a little more strategic, but that data pack that goes out. So you could say, hey, we're going to spend this on this tool, right?

21:59We use a tool internally called Maximor across some of our portfolio companies. It's a startup. I'm sure NetSuite will come out with a competitive product here in the fall as they launch out a bunch of new products around automating things. But basically, yeah, it just makes journal entries. And they've sped up, one of our companies has sped up the close significantly on a large business with a very small team. And given that you're dealing mostly with middle market, lower medical clients, it's not like we're going to take 10 people out of the finance and accounting department because we're automating things.

22:31In fact, you're probably more like, well, we just don't have to hire people as much as we get bigger. Is that right? 100%. So it's a great point. I think the trend we're seeing for the companies that are thinking about AI spend is the key metric that most folks are looking at is revenue per head and how that is increasing or flattening over time. I think most folks have a hypothesis that with AI, you can increase revenue per head over time through efficiency. I have not seen a lot of clients look to do cost takeouts yet with AI. So the important thing is measuring the productivity enhancements, and you can do that with dollars behind the scenes on the workforce impact.

23:15But most folks are looking how they can, every PE-backed company has generally a smaller operations team, smaller F &A org. They want to make those folks more productive, free them up for the more strategic work. It's the same song and dance that we heard with automation in general. when those tools came out, RPA and things like that. But the goal is to make sure that the workforce is as productive as possible with obviously a limited set of heads that you can to do those things like that routine analysis on a monthly basis. Yeah, but it's like let's make finance fun again. When private equity shows up, the data requests go through the roof.

23:57They don't end after the deal closes. They probably only intensify. finance teams who are listening to this who work for private equity know what I'm talking about. If you answer things really fast, then you just ask for more things. If you can put a 40-page data packet together at the end of the month, like maybe it should be 60 pages, like let's measure and look at everything. So that is overwhelming for these small teams. So certainly in the finance department, you can say, hey, we can use these tools and can get better. Now, right now, it's a, I think if we use these tools, we can get better.

24:30So I think a lot of people are leaning on you to say, no, no, we know, we've seen these things. If you spend this much money, we can deliver this much efficiency for it. How does a CFO have that conversation when they're asking for something for themselves, when most CFOs usually aren't in position to say, hey, give me more money, right? That's for sales, marketing, and engineering, not for F &A. Like, hey, I need 50 grand for this tool so I can save 150 grand. I think the first thing is obviously starting within your own organization as a CFO with that workflow is important because you'll go through the trials and tribulations of we thought this was a good investment case and it didn't happen in the way didn't unfold in the way that we had expected.

25:13So picking a workflow within your own org as a CFO is important so that if the marketing team, the operations team, the IT team comes to you with an investment case, you have that direct experience from your own team to weigh in and sort of gut check if that's the right approach or thinking. So you just got to do it. I know that sounds pretty cliche, but making sure that you have a strong investment case, tracking it. So not every investment is going to yield the ROI that you plan, but at least if you're tracking it and you get ahead of any spend overages or lack of value capture, for example, then you'll be able to adjust and reorient that spend to go somewhere that's more valuable.

25:57I think the, I'll just say like the experiments are happening, right? What we recommend with the experiments is give them a leash and give them a clock, right? So like in 90 days, kill it if it's not working out, fund it if you're seeing good returns, and scale it up if you're seeing great returns. So experimenting is not bad. There are going to be misses. There are going to be lack of value capture. But the measurement, the muscle that FP &A is good at, which is tracking, forecasting on an ongoing basis, is really very important, particularly in AI, when that value is not as attainable, maybe in some cases when adoption is low and things like that.

26:40And boy, do you have a lot more credibility as the finance team when you've used these tools yourself, if you've put yourself on the clock. And Kyle and I talked about this, and I've talked to your team about this a lot over the last 12 months, is this is almost as much a cultural, maybe more so a cultural issue or challenge in the organizations than it is a technology challenge of learning new tools and rebuilding workflows. It's like it takes one or two people in each department to say, we're going to do this. Let's spend some extra time on the weekends kind of playing around with this, see what works, find the edge cases so we don't get dismissed right out of the gates because we didn't think through some things that are unique to our business and then kind of just driving it through.

27:27So when you do it, then you can go to sales, marketing, and engineering and say, hey, guys, we're doing this. And we have the smallest budget and the smallest team on the company, which is usually middle market F &A teams. So I want to remind people the three questions from this report, which was what's the return? So what is the ROI from this spend? What's the cost as usage grows? That's an issue that often people miss. They're like, oh, this is a$25 ,000 light item. It's like, whoops, oh, now it's$125 ,000 line item. And then three, have we looked at this since we approved it? Because it changes all the time.

28:03And you may say like, hey, we approved this thing. Nobody's using it. So put a clock on it. I want to bang through the eight steps because I think a lot of finance and private people listening to this are like, yeah, this sounds like a great conversation. Where do I start? How does it even work? But so the report does a really good job of laying out the steps. Number one, know where your sector sits before sizing anything. Does that mean like your industry? like, oh, we're healthcare versus industrials versus tech. Is that what you mean by know what sector you're in? It is, yeah. So you tend to see trends within industries of where the value capture is for AI.

28:38And then within those industries, you also want to look at the function that you're investing that spend in. So finance or accounting is going to have a different return than maybe customer service, for example. So I would say this is the guiding principle of just be aware of what's out there and what other value capture companies in your same space are experiencing to be informed. It's the do your due diligence step of the process. Yeah. Well, I tell you, here's the easy button. Just call a court in and say, okay, you guys have a bunch of healthcare clients on a healthcare person. Where's the fat?

29:14Where's the red meat I can go after right away? So number two is tag every dollar. That makes sense. FB &A teams, finance and accounting teams are really good at tagging the dollar. But I would say don't tag it to engineering. Don't tag it to something large. Tag it to this project in engineering. This is a DevOps project or this specific project. This is paying down tech debt. This is building new products. This is automating testing and getting rid of QA, bot headcount because of it. So is that right? You got to tag it down to a very specific item, not just a general R &D item. Yeah, exactly.

29:57I think the three key tags are who's the assigned owner within the org, give them ownership, make sure they feel empowered to track and measure that over time in conjunction with the finance team, tagging to that, call it value capture, EBITDA lever, is this a revenue increase? Is this a productivity savings? Is it a cost takeout? And then lastly, to your point, tagging it to some sort of program, what is the overarching spend that we approved already through an ROIC process that would make sense to tie this to? Yeah. Number three is I think the hardest one of all, and it starts sooner than the budgeting process, but doesn't mean you shouldn't start if you don't have this, is fund the data foundation as its own investment.

30:43We own a company called Prophecy. There's a plug for Prophecy. It's just a master data management business. Using it at a client now. Yeah, thank you very much. It's a master data management product that cleans up all your internal data. So here's a stat that will blow your mind. There's way more private corporate data sitting on servers in the world than there is data on the public internet. And the LLMs were trained on the public internet. But the challenge is, you know, your FP &A person who just left six months ago has stuff sitting on a file server and email, you know, accounts that you don't know.

31:16You don't even know it's there. So it's just very hard to get to, like, that golden record, that, like, you know, that really clean data foundation. But you have to start at some point, and you have to chip away, chip away. So what are you seeing on that, you know, maybe with this client you're working with Prophecy on? The interesting thing is one of the biggest upticks that we've seen in business at Accordion is we have an offering called the data platform, which is effectively, you know, the old school data warehouse, obviously using all the latest tools and technology and AI enablement of that.

31:45But a huge increase in clients asking to put in that foundational data platform for all the reasons you said, right? AI making decisions off of data are only going to be as good as the data that you have, how clean it is, how up to date it is. So there's just generally a big investment there. I wouldn't say that's specific to AI. It's just been exacerbated by the need to deploy AI and where do you point that decision making process to? Generally speaking, it's a good data platform. Prophecy, master data management, that's one key component of it, right? If you want AI to make a decision on which department has the most opportunity for efficiency gains, you need to understand what your department structure looks like from an FP &A finance investment case.

32:35And then your data, if you're in the FP &A team and you want to understand why vendor spend is up, You want to make sure that your vendor naming conventions are clean and organized and aligned. So this is just the age-old adage of making sure that your data is clean so that you can make good decisions, AI now helping you make those decisions. The one thing that I'll add here is the historical sort of data cleanup exercise has largely been an exercise in actual data, so-called numbers, dollars. The thing that AI is introducing into the mix here is cleaning up your contextual data. So what are your policies and procedures?

33:19Where are your strategy documents? For AI to make a quantitative decision, it's helpful for it to have qualitative information to go along with that. So we're seeing a lot of clients sure up sort of their knowledge repository along with their legacy data platform. Yeah, an amazing way to get the new strategy and new rules and regulations out across the organization. When you have an MCP server or some sort of internal chatbot, to clean that stuff up, it gets disseminated pretty quickly as people are asking questions. So you don't have to be in every office every day repeating the things over and over again.

33:56You can disseminate it pretty quickly. Absolutely. So then it's, you know, you say number four, fund the tech, the workflow redesign, the change management altogether. That sounds like a lot to bite off. What do you mean by that? So it's like focus on everything was one of my old boss's sayings. Yeah, I think that the key takeaway there is, you know, this is not specific to AI, but the implementation, if you will, of the technology is not the only portion of the change. It's are people going to adopt it? Are you thinking about the shift in the skills that the employees on the team need in order to deliver that technology, that project?

34:37In this case, the technology being AI. So, you know, it's more of an art than a science for sure. But the key note here is don't just think about it as a funding of a tool. Think about it as it funding a broader change for the organization, because that's really when you're going to capture the value. The tool can be great, but if nobody's using it, you're going to get zero value out of it. We decided a year ago that the single easiest and most effective thing we could do across the portfolio and internally is just teach people how to use the basic tools that were available to them. That was it.

35:12Absolutely. Play around with it. Get good at it. Understand where Ethan Mollick would call that jagged frontier of AI is. Like, it's really good for this, but not good for that. So you're not kind of extrapolating like, oh, it helped plan my vacation. Now it can like, you know, rebuild this whole workflow. Or not saying, oh, it couldn't play in my vacation, so it can't do anything, right? Obviously, the models have gotten so much better and smarter over the year that I think giving people the basics and then, you know, that cultural change we talked about. This is like tools can do lots of cool stuff.

35:46But if people aren't bought into it and you haven't explained the why, that's going to be a challenge. So number five is, of these eight from the report, is put a name on every seat in the room. What does that mean? Yeah, that's going back to the comments we made a few times, making sure there's an owner. That's one of the easiest things you can do. Assign someone the ownership, the governance responsibility, the backing of the return on investment case, the adoption. Call it a project manager. Call it a program manager. That can be someone in the C-suite if it's a really strategic investment.

36:23That could be someone on the ground if it's more of a tactical sort of productivity investment. But who is the, hate the term, but throat to choke in this example? Who's going to be in charge of tracking it, realizing the value? Easy one that no one does today at volume. But from a process perspective, making that sort of a guardrail, a gating step is very important. And we say around here, if you have the information, you have the authority, right? So push that down into the organization. So you would say throat to choke. I would say give that person the authority to call timeout. Hey, this isn't working.

37:01I'm hitting roadblocks. We're spending this money. Give them an incentive to come back and say, let's not spend this money. We're not ready yet, right? my CEO is AI pilled and thinks we can do all these cool things. He gave me budget and, you know, authority and I want to please him or her. So I don't want to tell him it's not working. And so then I find out six months in at the board meeting when some private equity associate asked a question at the board meeting and you get really embarrassed that you haven't been tracking this or you haven't, you know, called it because you weren't measuring it, which gets to the next one.

37:33And number six, ask for the acceptance rate, right? That's right. Yeah, this one, you know, a little bit more for mature organizations, but needs to be done nonetheless. So you put AI in place to help you make decisions within a process. What does good look like in that? If AI gives you, let's take a simple example, a deep dive into your vendor spend, and it's giving you a good output 50 % of the time that you accept and don't edit, you know an explanation a uh a value capture on where to increase spend with a vendor or rationalize spend with a vendor what's a good acceptance rate what would the team find valuable and that's going to be case specific and it's a little bit more of a mature muscle to have but making sure that you understand an ai output is not going to be a hundred percent bulletproof what's the acceptance rate that really drives value and where do you have diminishing returns with that.

38:32Yeah. So like you said, don't let perfect be the enemy of good enough. Like, you know, keep moving forward. The models will get better. The tools will get better if you're waiting for them to be perfect. They're not going to be. And one thing I've talked about with people is we'll never get to perfection because as the models get better, as the tools get better, we're going to ask them to do more. The context window is going to get bigger. So we're always going to give it more and more complicated things. Just think of the things you asked ChatGPT to do a year ago versus what you asked it to do today.

39:03You're always going to have this sliding back to 50 % good, and then you work your way back up to 90, and then the context window opens up even bigger, and you can ask it for more complexity. So just play with it. Get in there. Use it. But then also, we're at a place where we're beyond tinkering. We need to implement. Last year was the tinker year. This year is the implement year. Next year is going to be the measure year, I think. if you're with me there on that, Justin. I would call it the reckoning year, because I think it's already starting to happen where we kicked off the call with some of the big headlines, but those same organizations that were token maxing blew out their budgets in four months.

39:43They're really starting to put guardrails in place now, so the spend per person is limited to X thousand dollars a month or$100 a month, and I think you're going to see more and more of that. Again, companies are in different stages of maturity here, but I completely agree that the transition is starting to happen for sure. Yeah. We're going to get into the next section here in a minute of how the CFO isn't put in the position of being Dr. No for all the fun. That's right. But we'll get into a second. So quantify the workforce question. Is this about like headcount and stuff? Is this what we're talking about?

40:15Yeah. This is kind of one of the most important ones, I think, that we haven't really touched on much yet. But the way we typically recommend handling it from a structure FP &A lens, CFO lens, is you take your census of your workforce today and you align those folks to call it a cohort. So across the business units that we have, the departments, the job titles, is this person an FP &A person? Is this person a customer service person? Is this person an engineer? So you get to kind of common set of generally aligned on functions within the organization and then plan out effectively the headcount growth slash the impact that AI is going to have in that cohort over time.

41:02And really the goal here is to make sure that your workforce plan, so your growth in headcount or lack thereof, has strategy behind it in the context of AI. Because again, some functions within the organization are going to be severely impacted by AI. You think of stuff like customer service reps who a lot of the level one support can be pushed off into an AI chatbot, AI agent. and then you have back office functions that may be less impacted. And again, the industry lens is important here, but finance exercise, they're doing their workforce plan, they're planning their headcount growth. There needs to be a, here's our assumption, our strategic goals on how headcount's going to grow in the context of AI.

41:48Gotcha. And then the last one here is test every item against a diligence question that hasn't been asked yet. You got to explain that one to me. What are we getting out of here? Yeah, this one I think is really just the private equity lens of what we kind of ended the conversation with a couple comments ago. And you're going to start getting this transition to why did you spend this? What was the investment case? What return are we getting here? So make sure you have that answer up front. It's going to come quickly if it hasn't already, particularly in the PE space. So, all of this ending point here is about put the process from the prior steps in place so that you can have a clear, articulate version of why you invested and if it worked out.

42:36And if it didn't work out, that you cut it quickly because you were able to get insight into the return that did not happen or wasn't realized. Here's the challenge every private equity management team is having right now, and it falls on the CEO and CFO mostly, is that they're being asked to do more AI. More AI, more AI. I heard this company did this. One of my other companies is doing that. Different industry, different scale, different sales cycle, different whatever. But it's an easy comment in a board meeting. Some portfolio companies are ahead of whatever company you're talking to, whether it's yours or somebody else's.

43:16And then they're going to come back in budget season and say, all right, well, here we go. Like, here's all the spend we want to do. And then the response is like, whoa, whoa, whoa, wait a second. What are we doing all this for? And I was like, where's the efficiency coming from? And it's kind of like, we're, you know, we got to start with something, make a guess, and then test against it. So the challenge is we're heading into really squishy budgeting season for all this stuff with unknown results out of the efficiency gains or the productivity gains or the growth, you know, what's going to unlock the growth or other things.

43:53And you want to make these bets, but you haven't measured them yet because they haven't happened yet, so you can't measure them. So you're going to have to build confidence around the tooling that you can measure this stuff. So the eight things we just walked you through hopefully give the CFO, show the confidence to push back on your private equity owner in budget season to say, like, listen, you want me to make some bets? These are the bets we've decided to make. Here's what we think the ROI is. And hold our feet to the fire on that monthly pack of that we're measuring the right things. Did I get that wrong?

44:24Or if you're sitting down with somebody, how are you having this conversation? Given that, we're all just making it up right now, I think, other than maybe on the engineering side. That's a little more settled. But everywhere else, we're kind of making it up? Do you feel that way? Yeah. I mean, probably the not so secret secret of the FP &A organization is that really good when you have some historical data to trend off of or to use as a reference point in forecasting the future. Not great when you don't have that historical context. And a lot of companies are in the place where the historical context of AI spend and the return just isn't there.

45:03So yeah, there is no magic bullet. You're never going to be 100 % accurate. You're actually probably going to be more inaccurate than accurate. But if you put a process in place to one measure, so start off with the right tagging, the right business case, if you will, then kind of build the new muscle, right? So spending on tokens is much different than spending on software or spending on a standard increase in an OPEX line. And then track, track, track, right? Did we start to realize the value we expected? If not, as a good financial overseer of that project, you know, kill it, fund it, scale it like we talked about earlier, making sure that you're forecasting and understanding the results.

45:49That's really the best and most widespread recommendation we can make is use the framework, track it. And then next year, hopefully, we're having a much different conversation where everyone Everyone has gone through and understood where the areas of their business were most impacted, and they have a much clearer historical data set, historical reference point to go off of. And you may want to have for 27, certainly for this, a first half and a second half budget, right, where you're like, hey, let's decide this for the six months, and if it's working and making progress, we'll greenlight it again for the next six months rather than a 12-month budget.

46:26That may be too much time to let this stuff run, and you may want to be really explicit about that. So I want to spend the next 10 minutes, 5 or 10 minutes, talking about what an engagement looks like for you on this, right? The phone's starting to ring. Second quarter, first half's over. We're heading into the final three months of the year. Budget season, everybody expects to get a draft budget before Thanksgiving so we can all beat up our management teams between Thanksgiving and Christmas. to finalize it by the end of the year. We got to get to the banks by the end of January and approve budget, all that stuff.

47:01So how are you engaging with a client? What does it look like? What do they need to be prepared to talk to you about? Is it a workshop? Is it like, hey, pull a whole team in? What does it look like to go through this with you guys if somebody picks up the phone, which I'm sure they will, and call you and say, hey, I need your help here? Yeah. So the question we get asked most often, and then I'll jump into how we kind of address it is, I need a plan for my AI spend. It's that sort of high level and generic. And I think you gave an example of that to kick off the conversation. So there's a couple of ways we as Accordion help there.

47:36First is from an FP &A lens, we'll put all the templates, the tracking, the financial rigor in place and give you a structure for that ROIC capture, that mechanics of understanding model modeling in the FP &A org, what tokens look like. So we'll give you sort of the FP &A toolkit, for lack of a better term, to help improve your tracking and management of AI spend. That's one typical engagement. All right. So I'm not staring at just a blank Excel workbook saying, I got to build this from scratch. You're going to give me something to start and we'll tune it to our organization. Exactly. And we'll do that, you know, if you're in Excel doing your planning, we'll do it in Excel.

48:17If you have a more upmarket platform like an adapter, and Anaplan, so on and so forth. We'll build it in those. But the important thing is meeting you where you are with the structure, not necessarily the end result or what technology you use for that. So that's one typical engagement. Another is, you know, we just don't know where to spend, right? So that goes to the conversation of like, where's the biggest opportunity as I'm building up my budget? And Accordion does most of our work in the F &A org, Office of the CFO. typically that starts with a workshop where we do benchmarking of your internal process how much time does it take what's the roi we do your standard sort of matrix of investment to deploy an roi an ai use case and the roi for deploying that and we give you a roadmap of what opportunities we see in your organization to spend so really two different offerings one is how you do the budgeting and the other is where you're actually spending that money in the budget Yeah, because everybody's getting asked right now in their Q2 board meeting, all right, as we think about 27, how do we grow faster and spend less money?

49:27Absolutely. And that's, A, I can help you there for sure, but it's making sure that you have the structure in place to make that case. Great. Okay. If you get hired to be CFO of Parker Gale Portfolio Company, You start in September 1st, and now you need to get ready for a budget season. Can you think of the steps that you're going to go through to get the organization around this? Because here's what I worry about is the CFO is going to be put in a position to be Dr. No. And the AI projects that are going to get funded are the ones that sound the coolest, that have these big ROI, rough and tough guesstimates around.

50:11We can grow this much faster. If we put AI in the product, we can sell more of it. We can sell it for a higher price. Or if we put AI into the customer service department, we can get rid of half the heads because this other portfolio company did this or something. Because oftentimes, and again, lower middle market, middle market, where you don't have these big, big teams, it can often be like loudest voice, biggest personality, momentum, shiny object, other things. And you have the CFO sitting back there having to be the referee of this. And unlike last year, having to tell people, no, we're not funding this project.

50:46We're funding this other project of your peer over here. He's getting everything he asked for. And you're not going to get anything until maybe June. Or maybe you get a little bit tested out and come back to me and see how it went. So how, yeah, you're the CFO of a company. Walk through how you go through that and get an organization around the idea that this is coming. I think the first thing to do is, with the management team, senior folks, really determine the strategic buckets of spend that you have as an organization. So if you're a SaaS company, you probably need to invest a lot in product changes in order to increase your market share or keep your costs flat.

51:26So think about the big buckets of spend that the organization can value. So again, things like revenue increasing programs and productivity enhancements. So start there, right? We agree that we want to put spend behind these broad initiatives. And then give structure to the folks in the organization to put those individual ROIC and PV cases together. Have them do the justification and really, again, just be the referee. Because if you're doing all the work for them, you're going to be accountable for the outcomes versus giving them the structure to think properly about how the spend actually unfolds and how you can capture value from that spend.

52:07That's just the best thing I think a CFO can do at this point is teach a man to fish, so to speak. Yeah, and I do think one early communication like, hey, guys, this is new for me. We're all going to figure this out together. We have initiatives. We have covenants. We've got goals for the organization that you all heard in the board meeting. We're not going to be able to say yes to everything. We've got to pick and choose where we're at. So please come prepared with some sort of ROI. And if you can make the decision easy for me rather than, hey, I need money for AI next year. So I need this much.

52:45I'm going to spend it this way. I'm tracking it this way. Here's how I know if it's working or not. Here are the check-in points. So I'm talking to you, sales and marketing. You know, make it easy for the CFO to say like, yeah, we're going to give you some to experiment here because they know they're going to track it. You never want to have that always no CFO lens hat on, right? Particularly with where we are with AI. So, you know, we said it before, but in that process, it's also fine to have, call it a slush fund, call it a internal enablement fund. Set that aside. Know those are going to be probably low ROI, but good for the organization to kind of understand the muscle that they're going to need going forward.

53:27But for the big bets, make sure you have that process in place that we talked about. Yeah, and make it easy for your private equity sponsor, too. The first page of the deck should be the five keys to the investment return. Whatever those five things are and how you're measuring them. This was our VCP. Here's our tracking against it, right? That should be the first page of every deck. The second page maybe next year is here are the AI initiatives we kicked off, and here's how we're tracking against them. If those show up every quarter and it's the second page of the deck, then everybody's going to have a lot more confidence that we're not just spending into something.

54:00And if something's not working, I was like, hey, we tried it. It didn't work. We thought we were going to get this payoff. It didn't work. So let's cut it off now and move on. I think you just build a ton more confidence. What did I miss, Justin? And what either in the report or stories and interactions you're having with clients on a day-to-day basis that people need to know in the goal of being useful, not just smart here? I think just in general, to summarize the conversation, we started off with the corporate cards are out there, spend isn't tracked. There's no guardrails around what the T &E budget is.

54:34The step in the right direction for everyone is really starting to put those guardrails in place. the corporate cards can still have spend associated with them, but know why you're spending, knowing where you're asking folks to spend is the most important part of where we are today. And next year, I think we'll be having a much different conversation. The final closing comment I would add is on that note, if you're not doing a quarterly reforecast or a monthly reforecast specific to those AI spend buckets, you need to be because the cost can balloon. Changes can happen quickly in the value capture.

55:10So just get the FP &A muscle going of more frequent reviews of that spend in the form of a re-forecast or a monthly forecast. You're going to get a lot of credit for being able to double down on a project that's really working. You're going to get just as much, if not more, credit for cutting off a project that just isn't working for lots of reasons because the tech doesn't work. It was too ambitious to start. It was poorly scoped or culturally the organization wasn't ready to adopt it. So just move on. And I think the FP &A team for tracking it and the CFO for being the hammer on this stuff, you're going to get a lot of credit.

55:50So let's do this. For more on this topic, we're literally engaged with Accordion at the moment on this in one of our portfolio companies. So you can read the white paper, you know, the AI budget, nobody knows how to build is due soon, where they go in more in depth on those eight steps. And you can also get in touch with the team at Accordion at AI at Accordion.com. What does that AI stand for? Is that somebody's initials? I've heard that before. So AI, AI at accordion.com. And I highly recommend go back and watch or listen to the Kyle Romer episode where we talk about kind of how AI is transforming the finance and accounting team.

56:30Don't forget to like and subscribe. The only way people find this and the crazy people have listened to tens of thousands of hours of this content over the years is when you like it and subscribe it. And we'll see you next week. Justin, thanks for joining me. I think we're going to have you back a year from now, if not sooner, to check in on how the budgeting process went and what we learned from it all. So bye for now. I appreciate you having me.

57:16I will see.