Ep 49: AI's Early Returns

Speaker A: This episode's made possible by Planful M. Hello and welcome back to Planning Aces. This is episode 49. On today's episode, we explore how finance leaders are navigating AI's early returns, where experimentation, efficiency and human judgment converge in the planning process. Our three featured CFOs bring distinct vantage points to the table, as you will see on today's show, we have CFO Craig Foster of PAX8 who shares how AI enabled tools are driving measurable efficiency gains. CFO David Opsler of Datadog discusses the challenges of capturing ROI in the midst of rapid AI driven growth. And CFO Ben Gammell of Brex offers a thoughtful reminder that forecasting still depends on the intuition and context only humans can provide. Together, their insights reveal a continuum of perspectives from pragmatic optimism to cautious experimentation on what AI means for fp. And a Zack Sweeney, your co host, joined as always by performance management guru Brett Knowles, our resident thought leader. As we unpack the lessons from this dynamic group of uh, Planning Aces, Brett Knowles joins us after this message. Ready to plan confidently, close faster and report more accurately. Here's what sets Planful apart. Uh, imagine purpose built applications for every department from FP&A to accounting, marketing to HR, all with built in financial intelligence. It's easy to implement Planful in just weeks and with minimal IT involvement. So you can rapidly and seamlessly engage everyone across the business on your key financial progress. Best of all, Planful grows with you. Its unparalleled scalability ensures that no matter how fast your business expands, Planful can keep up. See why over 1500 customers worldwide choose Planful as their flexible, user friendly, end to end financial performance management platform. Go to planful.com cfothoughtleader to see how you can reach peak financial performance with Planful. Hello everyone. We're back with Brett Knowles. And Brett, uh, since our last Planning Aces episode, uh, I think the seasons have shifted once again. How are, how are things looking up there north of the border? Are you enjoying some of the autumn color still?

Speaker B: Yeah. So we have our Thanksgiving about a month before yours because God took our summer away from us about a month before yours. So we're kind of a month ahead of you on the leaves at this point.

Speaker A: Fair enough. Sounds like you're uh, a full month ahead of us on everything. So as we look at today's Planning Aces lineup, Brett, do you have a preference for the order? I should mention I ran the Planning Aces and their comments through chat GPT and it gave me a few possible episode titles. I'd like to share with you. Of course, everyone knows that you, Brett, always get the final word on the title.

Speaker B: Yeah, right.

Speaker A: The top choice was AI's early returns. Second, forecasting the future with AI. Then efficiency meets experimentation. Prudence in the AI era and human judgment and AI tools measuring real impact. Quite a range. Right. Uh, which way are you leaning?

Speaker B: Well, so I would go your third to last one, which was from experience to something.

Speaker A: Efficiency meets experimentation.

Speaker B: So to me, that. Well, so what I'm seeing here is, um, you know, the gambit of AI perceptions. From AI can't do shit all to AI can do everything. Uh, and I think that's where the world is right now. Like, there's a gambit. You gotta decide where you are in that continuum. Some Luddites still think AI can't do anything. It never can. And others, uh, you know, like David, are just throwing everything they possibly can at this machinery to see what comes back. So to me, it's the gambit or. So, uh, it's, you know, the gambit. The, you know, what got you here is not going to get you there. That whole.

Speaker A: Got it. So before we dive too deep, let's sort out the order. We've got one finance leader who offers a healthy dose of skepticism on AI. Do you want to lead with that one, or should we save that for the end?

Speaker B: I would prefer to end with a healthy skepticism as opposed to begin with it. Uh, but that's, you know, my sales style is always to, you know, kick the bar stool out from underneath the guy. So all of a sudden he has to stand up and pay attention. So that's what Ben does.

Speaker A: Perfect. Then we'll close with Ben Gammell. Let's open with Craig Foster. Craig Foster's career spans Audit.com turbulence, investment banking, and six CFO roles. He's our guest in episode 1120 as PAX 8. Craig applies pattern recognition and ROI discipline to marketplace transformation. Here's Craig Foster.

Speaker C: We think that we can do 20% more with 20% less just using tools that are available. So there's this, you know, if we do things correctly, there'll be a massive pickup on efficiency and output from the company in terms of the things that we want to do. And, you know, we're going through our planning cycle right now, and it's really fascinating to see how people have manifested how AI enablement m inside the company can do things, from support to things that we can do in accounting, to automation of projects that were very, very manual, et cetera, externally. You know, we have this incredible distribution for all of these vendors and partners. Uh, they're building agentic components, we're building an agent, we're a marketplace. We need to incorporate those different, I don't know if you want to call them bots or AI components or some of this or modules and enable our downstream clients for efficiency as well. I think a lot of this, like big industry is actually moving into the SMB very, very quickly and we need to be in an AI enablement platform for them.

Speaker A: So Craig's clearly seen tangible efficiency gains ahead. Ah, 20% more with 20% less. Brett, what stood out from Craig's comments?

Speaker B: Um, I thought it was interesting because you know, Craig specializes in what all of us dabble in, right? That whole analytics and uh, trying to recognize patterns in workflows, financial flows, business flows. Um, and you know what really caught my attention was his quote of, you know, you're going to get 20% more with 20% less. What's interesting about that is when you and I started talking about AI a couple of years ago, maybe three years ago, Jack, people were shying away from anything that looked like AI was going to take away there was going to reduce costs, reduce headcount, that kind of stuff. Um, where Craig's right out there with it, right? I want to get 20% more, 20% less. That's a 40% gap, uh, incremental performance improvement that he's looking from this. And that was just a glib off the cuff comment. This is not serious analytics. This is pragmatic what's happening in the analytic world.

Speaker A: You must reflect on this quite a bit. I know. How do you see AI enablement for these small to mid sized businesses changing, you know, the competitive landscape? I mean what's, what's going to happen here?

Speaker B: Uh, well, you're right. Um, I do think about it and it's incredible, like every gambit of it. So for example, onboarding a new employee. We were talking to an uh, organization last week and they said, well it takes about two years to onboard a new employee. Uh, and then we said, well you know what if you had an AI agent and use that agent and every bit uh, of information that your field team gathered went into that agent. How long would it take you to onboard someone? And they said well, less than six weeks. So you're going from 24, uh, sorry, uh, uh, 104 weeks to a couple of weeks to onboard people. Like that's you know, huge. What does that mean? It means obviously people become productive and revenue generating positive revenue Generating early. What it also means is, um, uh, you can almost be less selective in who you hire on the technical side because AI is going to fill in that gap. Now I'm going to hire more for social skills and the ability to, to problem solve and help clients. Um, it means faster, ah, introduction, new products. Because now in uh, the current world where I have humans, uh, doing the helpline and support, I've got to train them on this new product. Whereas in this new world that training happens exponentially faster. So a huge amount in terms of the human capital, in terms of uh, the, the strength of small businesses is always your ability to enter and exit markets faster than the big guys. Well now you can even do it faster with the same tool set that the big guys are using. So that really allows us to operate in an incredibly agile format, whether it's creating products or supporting products. It just, it's a whole new world that levels the playing field between a small corporation and a large corporation.

Speaker A: That's really interesting. I, uh, heard something, uh, that echoes what you're describing, um, at a CFO dinner I recently attended. And I'm not sure if I can say, uh, publicly what the gathering was, but one point really stuck with me and that was that the jobs most at risk from AI as we move into the next year and the year after that, the year after that, they may not be on the front lines, but in the middle tier. That's where the layers of management and often the bureaucracy resides, uh, within organizations with new AI tools, younger professionals could move up faster, taking on what they once, you know, once was required years of managerial experience, doesn't anymore. Does that idea ring true with you?

Speaker B: Yeah, I would, I would expand, I would say absolutely. And I would expand on that. I would expand on it to say, um, first off, a lot of build management's role is to coach, measure, protect and train to help develop that new set of people. Well now the agent is going to be able to get you there faster with less errors. Uh, in this particular case, uh, that I was talking about last week, part of the issue is keeping abreast of technology is a Cisco switch or someone else's. But also the regulations, what are you allowed to do and the regulations in your jurisdiction? So, you know, these days, if you're supporting someone from India, how do you know what the regulatory requirements are in the UK it's tough to do if you're a humanoid, really easy to do if you're an agent. So many of the things that we used to do with Middle management now um, can be taken over by an agent. What middle management also has to do is monitor their employees behavior and as early as possible, um, nudge them in the right direction. Again AI is going to have way more signals of what that employee is doing, their success rates and those sorts of things and therefore that activity of monitoring is much less. The simple example is in sales calls in the good old days the sales manager would go to you for, with you to you know, 30 to 50% of your calls to get a feel for how you're doing in those calls. Of course the dilemma is as soon as your manager's in the room you behave differently. You know, with recent technology we've all been using agents in an AI zoom call to analyze the sales call and see how many times closing words were sent, buying words were sent, all that kind of stuff. Now my sales manager doesn't have to spend 40% of their time in joint sales calls. They can do that, be just as effective with 5% of their time. More importantly the rep can get feedback at the end of every single call. As the agent says, great call Jack, but there are three closing signals that were sent that you didn't respond to that you know, these times and this is what was said, this is how you should have responded to it. Uh, and also you know, you created five objections at these points by saying you know, A, B, C and D. All the stuff that you used to do as a middle manager, sales, um, management person, now AI is going to do a better job of guiding.

Speaker A: Before we started recording we noted that Craig Foster and our planning ace David Obstler shared a similar mindset while our third guest would take a uh, more skeptical stance. So let's move to David. Interestingly, both he and Craig have had six CFO tours of duty. David's uh, journey spans IPOs, M&A and Datadog's rapid rise of course. Here's David Opsler, CFO of Datadog.

Speaker D: Yeah, it's something that we talk about a lot now. We talk about the invest with the investors a lot and it's a multi pronged answer. First of all there is a demand um cycle going on for AI tools. We call them AI native companies, um, where um, there you know, are a lot of products that are being used by either consumers or enterprises. And um, and we have been able to monetize that we are the leading platform for cloud and AI, uh, native companies. So we've gone through a significant demand cycle. We have, in our last earnings call we announced that this sector is growing very rapidly, around 11% of revenue. That eight of the top most valued AI companies in the world, AI native companies, are using Datadog. And that um, we have, I think we said we have um, over a dozen, over a million of ARR 80, over 100,000. So we've been able to be the observability platform for these companies. But that's just the tip of the iceberg then um, within our platform we are making investments in order to cover the AI workloads that our clients are using. Um, we uh, have integrations with a variety of information sources and uh, 4,500 of our over 31,000 customers are using that data. So we're covering then in terms of the platform, um, we've always been a company that uses information to be predictive and to allow our clients to remediate. And we've been investing significantly in large language models. We have some models that we've created, we use other models and we are um, essentially enabling our platform to be able to be used uh, with um, generative AI that we believe will speed up um, the information flow, the analytics, the remediation. At our DASH user conference we announced a number of sets of functionality that um, um, we call bits. And one of the most powerful ones, where clients are really excited is bits, um, used in service management. And what this does, it enhances the analytics and understanding what's wrong and then helps the path in figuring out how to correct it very quickly. And then as you said internally there's a lot of ways we're using AI, some examples of that, or we're using the various coding uh, agents to help accelerate our software development, um, cursor, et cetera. And we're also using it for sales intelligence. We're using it for some of the more automated functions within finance. And so we are trying to see if we can um, get greater productivity and speed in the production of our software, in our sales motions and in our back office admin. So there's many ways that AI is affecting um, datadog both in terms of our customers as well as internal. At Datadog we are in a situation where because of our product, the first thing I mentioned, monetizing through the investment cycle in modern AI software companies. It's pretty easy to see the return there. So that's where you know we have 11% of our revenues are there, um, in terms of our platform itself, itself. We're still early days. Somebody like me in my CFO seat is looking for that evidence a lot. Most of the Time. It takes two to three years to get return on your R and D investments. So I think, um, we're still, um, in that area. We're still in, um, the days of developing products, putting them in GA and determining what the return is. But we're pretty confident, given our platform and how it's used, that there will be a strong return. And internally I think that's a very good point. We're in experimentation mode. I think we're basically trying to lower the barriers to adoption. Um, we're doing it in a prudent way. We're doing it so it's not changing our cost structure yet, but we want to see where we're getting return. And we'll start to develop metrics, for instance, encoding. Are we increasing our productivity in software development? And I, in my CFO seat are waiting for those types of metrics to be able to calibrate the level of investment and what we do going forward.

Speaker A: So AI adoption, still searching for the roi. That's how I framed it in my notes. Brent, what were your key takeaways from David's comments?

Speaker B: Well, you know, I always love taking a contrarian position. That is the least important message of David's conversation. Uh, the AI thing you had to torture out of him and he was like, is like drinking from a water hose. Like, there are four different ways that. Or eight, I forget how many. There's, you know, a multitudinous of different ways that AI is playing out and he just rattle them off the top of his head. Um, which was completely impressive. And none of those, I mean, they all sound impressive to me. They sounded like they had a very powerful roi. Later on, when he asked him the question about roi, he said, well, we're still at the experimentation stage, but certainly, um, whether there's an ROI or not, that was not a constraint to his organized thinking in any way, shape or form. That was like an irrelevant data point that probably would never have come up if you hadn't brought it up. In other words, you wouldn't have said here are the four things. Or I forget. Four, eight, whatever it was, these are the four things. Ah, but you know what, Jack? I really don't know what the um, ROI is. I'm not sure we can go forward with them. Like that was not part of the equation.

Speaker A: That's an important distinction, especially around managing cost without fixating on near term ROI. It feels like CFOs are evolving their playbook here, doesn't it?

Speaker B: Well, have you heard of this little organization called, uh, Google. Do you remember when it did its IPO? What was its revenue line?

Speaker A: 0.

Speaker B: $0. The standard metrics we use, those RI metrics would have had it valued at zero. But people recognize the value, intrinsic value in having eyeballs. And you got eyeballs, you can advertise. And a great IPO and tremendously undervalued. Um, be that as it may, there's a whole bunch of numbers he rattled off that if we had those numbers in any business. So the first one, right under a man cycle, 11% of his revenue is coming through AI at this point. That's a significant. How many people would want an extra 11% of revenue? Yeah, anyone would. And then he says, um, you know, uh, eight out of ten are driving value through AI. Again, whether you, whether you waste your time calculating ROI or not. You know, if eight out of ten of your customers are getting value from it, you can, you can monetize it, right? 8 out of 10 are using AI is an important component of their business. So, yeah, again, you know, roughly 10% of your business is dependent on it. Like, there's just so many things that, you know, make it clear. And he even said some off the cuff stuff like, you know, his third point was platform, right? They've made some of their own LLMs. I don't know if you've ever looked at the cost of building an LLM. It is astronomical. Uh, so, yes, he's leveraging off other people's LLMs, like most of us are, but he's building his own as well. Well, you got to know he's got some numbers up his sleeve that would substantiate that level of investment. So whether he's disclosing that to you or not, there's just a whole bunch of stuff that came out in his conversation that made it clear to me, uh, that he, you know, had a strong case for. His fourth one is about using coding agents, right? Sales intelligence agents, finance agents. Like, he's got them in every single department. Don't tell me he didn't already figure out what value it was giving. So I'm kind of taking the opposite point. Like, yeah, he says he's experimentation. We've, we've talked about this before. Like, people carefully craft, Jack, what they share with you. Uh, and, uh, you know, if you look beyond the words, I, uh, think he's got a clearer handle on the ROI and the revenue being driven through AI at Datadog.

Speaker A: How important is building metrics early when you don't yet have a clear ROI story?

Speaker D: Right.

Speaker B: So I mean, he's got to be clever. Like, they're publicly listed. He's, you know, SEC is going to be looking at him. There's only so much disclosure that he can give you. Uh, but, uh, as one pragmatic, uh, financial person to another, uh, you know, if it walks like a duck. The interesting point is, and I've badged you about the ROI thing because I believe firmly, we, as the financial team, unfortunately have built this trap of quarterly earnings and roi. It's our own doing. And in this world, um, the roi, he said it's two years. I don't know if it is or not, but it's not as simple as it is in buying a new printing machine or new, you know, truck for deliveries. Uh, but there's got to be enough, uh, ancillary data to indicate whether it's pushing in the right direction or not. So I think ROI is a wrong question to think about, especially in this, this emerging technology where, you know, what was expensive a year ago is now free. Like, you can't do an ROI when God knows what the cost base is going to be when I'm trying to recover that investment a year from now. The part that I thought he said that was most precious was be careful not to drive up costs. So all this stuff he's doing with AI is not incremental cost. I should be able to say, if I got to invest 100,000 in, uh, programming agents, then I need to throttle back program expenses somehow by 100,000, whether that's less bugs, faster turnaround time, less programmers. Don't drive up cost. So, uh, I'm saying stop focusing on roi, but don't take your hands off cost. This is still a free cash flow world.

Speaker A: That's an important distinction, especially around managing cost without fixating on near term ROI. It feels like CFOs are, uh, evolving their playbook here, doesn't it?

Speaker B: Well, again, it's a trap that we made for ourselves. Like, God didn't say that we have to divide the year up into four sections and report earnings at the end of every four. Um, and you know, your buddy Trump is suggesting maybe we don't need to do that. Interesting perspective. But, um, the point is, there's no magic to the quarterly number. We just made that up. There's no magic to an annual financial plan. We just made that up. And right now we're suffering from those arbitrary rules because they don't fit what the new world looks like. So ROI fits in that category. In my mind, uh, it's an old fashioned view of business that will limit organizations ability to take advantage of this new world.

Speaker A: Which brings us to our final planning ace Ben Gammell of brex. Ben's career has spanned global banking operating roles and leadership at brex. Here's Ben Gammell.

Speaker E: I think candidly for from my vantage point I'm um, yet to come across an AI tool or product that is able to forecast with a higher degree of accuracy than candidly even our finance team is able to forecast with more rudimentary approaches. Where I'm seeing AI be incredibly helpful on is obviously the speed of which you could understand your historical results, the depth of which you can drill into those results and understand either correlation or causation that may exist over large swaths of data which obviously Brexit is a large amount of by virtue of us being our own payments infrastructure platform. That being said, when I think about being able to extrapolate pass 90 days, I think that's where I see the AI sort of degradation of uh, fidelity as it relates to forecasting accuracy because there are just dynamics that I don't think the models are able to sort of understand and ingest. As it relates to what is your actual business strategy. How are you thinking about the ramping of sales reps or the deployment of marketing dollars or the you know, especially as you move into new markets and new sort of uh, growth verticals such as maybe this EU or UK expansion or so even these partnerships that only existed for the last 12 months. An AI model just really would struggle because there's just like not enough historical data in which to rely upon. And when companies are growing as fast as Brex are adding new products, adding new markets, there's just like a less of a like you know, 10 plus year look back that those like AI models can use to then sort of prophesize the future. Whereas I think with just humans in the loop, I think you have a bit more of an intuition around some of these things that will enable you to get a more accurate forecast.

Speaker A: Ben raises an interesting point about the role of human intuition in forecasting. Historically it's been inseparable from the process. But is that dynamic beginning to shift in some way?

Speaker B: Well sure. And it turns out that the wheel and productivity are inseparable for many centuries. Not uh, the case anymore. So again Ben, there's uh, nothing that you can cite with Ben. What he said has been correct historically. The question is, is history a good predictor of the future? Don't think so. I think there are Weaknesses in his point. Now, it's biased about his market, his customers, how he's using AI. I don't think it, it fits his global truisms other than pay attention, keep humans in the loop, and don't get carried away with AI solving a bunch of things. It's just another tool in your, in your bag of tricks that you can take advantage of. Now, whether it can do the analytics or not, I would argue it can. But, um, you know, I'm not in his business, so I'm not well equipped to argue whether it can do what his organization needs.

Speaker A: Across all three conversations, there's this shared realization that AI isn't just about replacing tasks. It's reshaping how finance teams think and work. Would you agree?

Speaker B: Well, I think, uh, a couple of common threads. One is, um, we've got the whole gambit here. We've got from, you know, if I exaggerate, Ben being a Luddite, stay away from AI. Only humans can do this stuff to David. Let's throw AI at everything possible in the business and see what great things happen. So I think that's interesting, and that's the creative gap that we need in our lives so that we don't, uh, get too caught up in our own trenches. I think secondly that, uh, we're seeing people being much braver now about saying AI is going to take cost out of the system or improve productivity and there will be consequences. Whereas a year ago, we're being very careful not to say those things. Um, thus we have, I don't know, people protesting outside the plant or investors, uh, getting anxious, uh, so it's no longer a taboo. And I think we're seeing more and more of that. And I think finally we're seeing it being insidious. It's going into every crack and gap in organizations and finding a place where, uh, it's helping us do stuff better, faster, cheaper.

Speaker A: And just to be clear for our listeners, we're certainly not calling Ben Gammell a Luddite. He's the CFO and president of brex, a, uh, unquestionable technology leader. Brett.

Speaker B: No, I would say yes to what you just said. Like, you know, obviously I'm trying to poke, um, at things to make our brains work on what we've heard. So obviously I've taken out of context.

Speaker A: Exactly. And it's worth noting CFOs like these have to be careful in how they share progress. They're speaking to investors, employees, and partners, each with different expectations. So, uh, full transparency isn't always possible.

Speaker D: Right.

Speaker B: Think of that as a pie chart.

Speaker D: Right.

Speaker B: If you draw a pie chart of all the stuff they're doing, what portion of that can they actually share? Given regulatory constraints and investors and everything else, But I have no idea. Make up that number in your head, then say of what they can share. Um, how much do they want to share in a podcast? Like, they've got to have a motivation to be here of some sort or the other. Don't know what it is, but something is motivating them. And therefore, that also puts a bias and a slice on that pie chart of what they're going to share. Uh, and then they got to think, uh, of what actually comes to mind during the conversation. I know sometimes you brief them with questions ahead of time, but the conversation drifts and flows. So again, we're getting that, that pie chart sliced down to a pretty thin slice at the end for Brett to take that thin, thin slice and try to extrapolate back to the a hundred percent. Man, if I could do that, I wouldn't be wasting my time on podcasts. I'd be like, investing like crazy. So, you know, for sure I'm off track. What I'm trying to do is provoke your listeners to, uh, you know, listen and look at opportunities in these great conversations that you've been able to bring to the table.

Speaker A: You're right. And it's remarkable that these conversations are even happening. Two decades ago, few finance leaders would have felt comfortable publicly exploring topics like this. Now, podcasts provide that forum, and CFOs are stepping forward to share complex ideas with multiple audiences at once. And it's a big shift in how finance leadership communicates.

Speaker B: Yeah, I think it, uh, you know, I think it takes a village like, none of these are doing it on their own. And to a certain extent, they can move forward rapidly or slowly based on what that rest of the village, investors, customers, competitors, employees will allow. And so, uh, I think this breadth of conversations that we keep on seeing reflects the diversity of that stakeholder environment.

Speaker A: Well said as always, Brett. I think that's a, ah, fitting close for this volume of planning aces. And as the season turns, here's wishing everyone, especially our friends in Canada, a happy Thanksgiving for Brett Knowles. I'm Jack Sweeney. Thanks for joining us on Planning Academy.