What is ROI and Why Does it Matter in AI Software Investments?

Live from Schenectady, New York, it's the Metrics Brothers, Growth and CAC.

And I'm Growth, better known as Ray Reich, founder and CEO of Benchmarket.

And I'm CAC, better known as Dave Kellogg, EIR at Balderson Capital, independent consultant, and the author of Kellogg.

And together we are the Metrics Brothers.

We are indeed, Ray. And I almost said SaaS talk on the intro. I had to fight myself to not say it. Uh, but we are indeed the Metrics Brothers and we go together like investments and rates of return.

Now I can get behind that one. No food metaphors. And I do love investments, especially those with returns. So, you know, some people actually refer to me as an ROI guy.

Well, you know, people call you a lot of things, Ray, my metrics brother, some of which are not appropriate for this podcast if we want to preserve our family's safe rating. But you know, Ray, I lived in France for five years, and you know what I say about ROI?

Je ne sais pas.

Close. Try it again, Ray. One more time.

Je ne sais. Okay, tell me, Dave.

We rehearsed this, guys. Okay. What do you say? I say ROI is king. Uh, and if you happen to speak French, it's a great dad joke because the French word for king is roi. So anyway, ROI is king, right? You get it?

Okay, Dave, you've out-dad-joked me for once. But with that one, CAC, you know, I want to recover before starting this episode. So I think let's hear a word from our sponsor before discussing today's topic, and that is, what is ROI and why does it matter In AI software. The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for the remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave @Kellogg and Ray on LinkedIn @RayReich. Thank you so much for listening to the Metrics Brothers Podcast.

Okay, Ray. So look, while ROI is indeed king in more ways than one, it can be kind of confusing. So, could you please walk us through basically the basics? Like, what is ROI?

Well, return on investment is a way to measure the financial benefits of investing in almost anything, but for this episode, in an AI tool or an AI agent. And you typically are going to compare the gains, like the revenue cost savings or the revenue increase, productivity improvements, and you compare that to the cost of investing in that particular tool or system. So simply stated, ROI equals the net benefits in dollars divided by total investments in dollars.

Or even simplerly, if that's a word, stated, um, what you get versus what you gave. I always think of it as a give-to-get metric. And if I give $1 million and I get back $300,000, then I have $300,000 divided by $1 million, I have a 30% ROI. Correct, Ray?

Exactly.

Okay, so Ray, if I step into a room and I'm attending a meeting where the subject is ROI, In addition to the, to the word ROI, what other words am I likely to hear?

Well, the reason I really wanted to do this episode, Dave, was I recently did some research with Scale Venture Partners about using AI software in the go-to-market organizations. And the number one measurement of ROI was increased productivity. And we're talking 70 to 85% of marketing executives, sales executives, And customer success executives were measuring as my— name the role— SDR can do more highly researched, personalized outbound communications. Or I can do more campaigns with better creatives. So this increased productivity and improved quality of what I'm doing, I hear so often. The benchmark research says it, but it doesn't carry it all the way through to— And because of that increased productivity, I'm getting more revenue for less cost.

Okay, so let's— I'm going to put a pause there, Ray, because I want to come back to that. But, but right now, if I step into a room to talk about ROI, I'm going to help answer the question, my own answer to the question. One, I'm going to hear about ROI, so return on investment. So we're going to talk about gains and costs, but I'm also going to hear other words. I might hear words about what's collectively known as discounted cash flow, DCF analysis, where you're going to hear words like net present value, NPV, or words like IRR, internal rate of return. Those words can come up. I think many of you in the audience are familiar with them, so we won't rehash them here. And if you're not, just go to Investopedia and either look up DCF analysis or look up IRR or NPV, and you'll get a basic primer. I would say for our audience today, although I'd never really thought of it this way, that ROI analysis, just putting gains divided by investment, is basically a degenerate form of DCF. Putting that aside, so there's 2 things you could hear about. One, just ROI itself. 2, discounted cash flow analysis and related terms. But 3, I know the one that comes up a lot that you're passionate about, Ray, is payback periods. So could you tell us a little bit about what a payback period is? And I know you have some war stories, maybe you wanna share one.

Yeah, and this is 'cause I would always do a pretty detailed business value assessment when we were selling enterprise software. And I had this CFO who said, okay, I'm gonna invest $125,000 this software, we need to make sure we get the payback in 12 months or less. But one of the debates I had with them, it's like, yeah, but if you're investing $125,000 right now and you're going to be using the software for 3 to 5 years, which is the typical life of our customers, the benefits really accrue year 2, year 3, year 4. So you're missing the actual entire return on investment, which is going to be a compounded return over time.

So just to be clear, a payback period is how long it takes to get your money back. So I think of it as how long your money is on the table. So if I have to invest $1 million, right, and the purpose of all this stuff in general, people, is to compare projects to figure out which one to invest in. So if we have 2 projects that both cost $1 million and one gives me my million dollars back in 3 months and one gives me my million dollars back in 13 months. Which one's better?

And sometimes you're comparing apples to oranges. You might be talking about, hey, this investment is for my— to make my sales team more productive, and this investment is to make my customer service team more productive, right? Sometimes—

Notice how Ray avoided the trap question. But Ray is too clever to take my trap because that was actually a trap question, Ray, and you avoided it. A lot of people would say the 3-month one's better because you get your money back faster. But, but the actual answer is it depends on the how much money I get back in total, right? If the one that gave me back $1,001,000 in 3 months, so yay, it's got a 3-month payback period, but I have a total value of $1 compared to one that gives me nothing for 13 months and then gives me $5 million, right? Uh, it's the second project. The 13-month project is much better. So, and before we get off this, Ray, I— I dislike payback periods. I like the notion of how long my money's on the table, but people think they're a return metric and they're not. They're a risk metric. They're not measuring how much money you get back. They're not actually measuring ROI. They're measuring risk as measured by how long your money's on the table with the idea being that the longer it's on the table, the riskier it is. So that was payback periods. Are there any other metrics, Ray, that are likely to come up before we move on to the next topic? Are there other words you might hear that conversation. I don't think so, but I'm just checking.

No, I think you mentioned the IRR, internal rate of return, because that's something else people use versus net present value and discounted cash flow.

Yeah. And actually, I know in the prep, one of the things you talked about was comparing an IRR to a hurdle rate. That's something that might come up where a company may have some minimum rate that you have to exceed. Otherwise, the project won't be considered. Typically, the hurdle rate's one between 8% and 12%. But that's another word you might hear. Again, you can get that by Googling it. So, Ray, I want to switch gears now. You talked about what you called a financial ROI, or you talked about what I call soft benefits versus hard benefits, or maybe even soft costs versus hard costs. Could you help us understand what do you mean, or what's the difference between kind of— I'll do the benefit side first— between soft benefits and hard benefits?

So a soft benefit, and I consider this a soft benefit. Let's just take you're investing in AI software for your sales development team, and they are able to do more high-quality personalized research and outreach. And at the end of the day, they may only be doing the same amount of outbound activities, but the conversation or connect rate goes up by 10%, and the conversion rate from a conversation to a scheduled sales call.

Yeah.

goes up by 10%. I'm just using 10% as an easy number. Those are benefits, but they're not often, not always translated into, and what that means is I can generate this much more pipeline and this much more new ARR for the same or less expense, which would be a true financial return.

Correct. And Ray likes to say financial return. I personally think it's redundant, or certainly financial ROI to me is redundant, but, but it's okay. He likes to emphasize that the ROI is measured in, you know, dollars, or actually measured in percent. Um, net present value is measured in dollars, IRR is measured in percent, just to be a units freak. But, um, Ray's point is a good one that, in my opinion, a lot of people fumble the ball on the 5-yard line because they'll just talk about, hey, the SDRs Like a soft benefit. The SDRs are happier. They enjoy their job better. That's a soft benefit to me. They can make more calls. Okay, that's trending towards a hard benefit, but it's still soft. Well, how many more calls? Oh, 10%. Or the conversion rate in your case. The conversion rate goes up 10%. Well, again, that's getting harder. So, the formula I like to apply is what in marketing we call so what. This is the way you get from features to benefits in marketing. You just keep saying, so what? So what? So, so what? The meeting rate is 10% higher. So therefore, the cost per opportunity goes down. Ooh, that's getting financial. Therefore, I could generate the same opportunities with less cost. That's definitely financial, right? Because I'm one step away from converting that into dollar savings. So to me, that's the art. And look, you're not always gonna— I think one of the great myths, Ray, is that any soft benefit can be converted to a hard one. And I don't buy that. But I think many, many can.

No, I agree. But, and I think I would, I just would want to say this because it's why I love, you know, talking to CFOs so much. I love it when I could walk in and say, hey, for this $500,000 investment over the next 3 years, I'm going to be able to generate $4 million incremental ARR and save $1 million in people cost to do it. So I'm going to generate $5 million of net cash benefits for that $500,000 investment I'm asking for. That's a very different discussion with the CFO, Dave.

Yeah, so I hear you, Ray. And I think, look, to cut to the chase on this one, because I like the soft benefits. I like the happier employees because I do think happier employees leads to less turnover, which leads to less recruiting cost. But boy, that's not an argument I want to make in front of a CFO. So, I'd rather have that be icing than cake. I want the cake to be the hard benefit ROI analysis. Cost per Opti will drop. You can hold us accountable for that because we're going to sign up for the same OptiNumber with 2 fewer SDRs if you give us the $100,000 package for qualified or whatever. Right? That's hard. So, to me, it's the icing on the cake. And the cake is the hard ROI analysis, what Ray calls the financial ROI analysis. But I'm like, don't forget the icing. Hey, these calls will be better. There'll be fewer mistakes made. Just because these things are harder to measure doesn't mean they don't exist. So, I'm a big believer of kind of a left-right format. And maybe with the CFO, you only show them the left side, you know. But with the CRO, you show them both sides. And maybe the CEO. What do you think, Ray?

Well, I live with the CFO and I make sure I show her both sides. The side that she goes, Ray, we have 2 options in our budget. We increase revenue or we decrease our expenses. I go, what about my happiness with the new pool? She's like, yeah, that's nice. But how do we, how do we fund it, Ray?

So let's talk, Ray, about personal productivity, because you started the episode by saying you did some research, I think with SCaLE, on AI projects. And one of the things you found is that they're often justified based on personal productivity savings. What would you do instead? Or how do you feel about that? And what would you do instead?

Maybe I take this a different direction, because we've talked a lot about that personal productivity or some of the SOP benefits and how to translate that into the hardcore, you know, increase revenue, decrease costs. But the other thing that came out in that study was about 40% of the go-to-market executives said they were going to purchase AI application software to introduce AI into their go-to-market processes. 30%, so almost a third, said they were going to actually build an AI tool or agent on top of a large language model. And Dave, this is where I think an old problem becomes new again, because there's a lot more cost to building a piece of AI software than buying a packaged app. And maybe we can dive into that?

Sure, Ray. You jumped the gun. Let me do one point first, and then we'll come back to the build versus buy, which is what I thought you were going to say was people are successfully getting projects approved using personal productivity as the rationale. And that works today, but I think we're both worried that it may not work tomorrow. Any thoughts on that angle before— I promise you we'll get back to build versus buy.

Oh yeah, like 80 to 85%, I forget the exact number, said they have invested into an AI or AI tools in the go-to-market. And it's because they have to. They don't want to get left behind. They're not even being forced to do hardcore ROI. They're like, yeah, we need to use a clay-type product to have better data enrichment and doing more personalized research. But that we have to do it just to kind of try to experiment and stay ahead, that's only going to get you a budget a year or two from now. It just won't.

Yeah, and that's the main point that Ray made in the warm-up when we were talking, which is, you know, you get away with kind of weak ROI analysis today because there's such a tailwind around AI. Everybody wants to approve AI projects. But between this MIT study, which I don't think is very good, that's telling people no, no, 90% of AI projects fail, And other kind of AI backlash sort of messaging. I think Ray and I both think it's going to get harder to get projects approved in the future than it is today. And that means if you stop at just making a personal productivity-based argument, you may not win your approval for your project. So you need to get better at doing the full ROI, which in my mind means multi-year. Basically, it's got to be typically a 3 to 5 year ROI analysis that looks at all costs. that translates soft benefits into hard benefits as we demonstrated, and then complement that with the soft benefits in a list. Now, let's go back, Ray, to what you want to talk about. So, everything old is new again. Everybody wants to— not everybody, but I think you said a third of the people want to build AI apps as opposed to buy them. This feels like déjà vu all over again, Ray. Talk to me about build versus buy.

It does. And it's really, you know, build and run because, you know, and a lot of times I get this from RevOps leaders, right? They've convinced their CRO, hey, I can go out and get Claude or get ChatGPT, and I can build a research and outreach engine for my SDRs. And they're thinking about, hey, I can build this in less than a month, and it's going to take me maybe 100 hours. So that doesn't seem very much, right? But then you've got the true operational cost, which is going to be the inference and token costs for that model. And often you don't know exactly how much it is going to cost until Until you have some history and experience, because the prompting that you use sometimes can be different between user 1 and user 2. So you're not gonna always know the operating costs, unlike seat-based operations. You're gonna have some— if you're using your data, you're gonna have some model training costs that needs to be factored in. Now, one of the big things under that total cost of ownership kind of orientation is what happens when ChatGPT changes from 3.5 GPT to 5 or to 6, you might have to go back in and modify your own programming on top of the model to take advantage of all the new capabilities of ChatGPT 6.0. And then I already had one of my clients see this. Oh, wait a minute, my RevOps person who built this a year ago left. I need to make some enhancements and there's no documentation. of what we built. So, this is the classic. It's not just building it up front. It's running it, maintaining it, enhancing it over time. All needs to go into your total cost of ownership and your ROI model.

Yeah, look, there's a reason most apps today are bought because we lived all through this. Back when I started the software business, people would build their own HR app. They'd literally build their own PeopleSoft or Workday employee table, hire, fire, promote. And then people kind of realized that, gosh, that problem looks very much the same across different industries and companies. And there's no reason we should go build that because, you know, while you can whip it out in a day, it's quite expensive to integrate, update, maintain, add new features, stay compliant, yada, yada, yada, do authorization. Everything gets hard. So, we kind of lived through this once before. Ray talked about that. Like, my joke is we've been misunderestimating, to quote George Bush, we've been misunderestimating build costs for a really long time.

Yeah.

There are the low-code vendors out there who do a good job, in my mind, filling in the gaps on apps with basically custom-built but quickly and easily built apps. They're kind of an exception here. But for like core enterprise apps, I think we've learned a long time ago that they're way more expensive to own than they are to build. It's kind of like, oh, there's some— oh, it's like a puppy. They call it the free puppy, right? Hey, the puppy can be free, but the total cost of ownership is high. Vet visits, I had my dog at a neurologist, a dog neurologist, right? So you can find these things are much more expensive to own than to build. And I think Ray pointed out that in the development area, you've got model training costs. What's different about AI? Model training costs and development. In operating costs, you've got the inferencing and other usage costs. And in maintenance costs, you have keeping up with the evolution of the foundation models. Those are just some of the extra costs. you need to factor in when thinking about AI applications.

And then one other thing, just kind of wrapping up, is, you know, we talked about payback period is nice, but looking at a 3 to 5 year ROI model is where I can optimize what the true return is going to be over time. But one of the nice things I think a lot of companies don't do but would be really valuable is to actually do a review. Hey, this is how we justified the investment. Now, I've done it after 1 year, 2 years, 3 years. What's been my real return? Because if you actually do that, you're going to be better than 95% of the people and companies out there, Dave.

Yeah, I know what you're saying, Ray. I have in my 35 years, I literally have never worked at a company, including those I ran, where we looked back at past basically capital budgeting type questions and said, gosh, 3 years ago, we looked at a model, we invested in this thing. Did we, you know, how did we do? How did reality work out relative to the model we built? I think it's a good practice. I think very few people do it. It sounds like, Ray, you have. Have you?

Have you ever worked at a company that did this? Well, good habits die hard. And that's my own saying because it should be bad habits. But, you know, as we've talked about, you know, up there in Schenectady, New York, I was trained by GE. And GE had a lot of huge capital-intensive investments. We're talking about hundreds of millions of dollars. And they forced us to have the discipline of whenever we did an ROI justification model upfront, we would actually get called on the table after 1 year, 2 years, and 3 years to do a review and show what the actual investment outcome was. But that's old school, I guess, Dave.

Yeah, it sounds like a great practice, to be honest. Too few people do it today. So, Ray, I see we're coming up here on 22 minutes. Are there any final things you wanted to add before we wrap up?

No, I just think, you know, with AI software, what's old is new again. If you're going to think about developing it, make sure you understand the total cost of ownership. And please, as much as possible during these good times of, hey, go do some AI stuff, try to understand not only the soft benefits like productivity improvement and take that all the way to income statement impacting. How does it increase revenue and/or decrease cost?

And by the way, grab the money while it's available. Get those proposals in while you've got the big tailwind. So thanks everybody for listening. We will see you next week.

Thanks, Dave.

The Metrics Brothers: AI, SaaS, and More is a production of The Metrics Brothers Growth and CAC and a member of the Benchmarket Podcast Network. By accessing this podcast, you acknowledge that The Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided Or the humor content of the jokes told.

Ahem, Ray.

Information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes only. This podcast should not be considered professional, or for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of, or God help you, reliance upon the information presented here. Ray Grothreich is based in New York City and available on X @rayreich. Dave Kellogg is based in Silicon Valley and available on X @kelblog. Schenectady, which is French for unspellable, is not our actual production location. Thanks for listening.