Subscribe
Share
Share
Embed
Demystifying the conversations we're already having here at RRE and with our portfolio companies. In each episode, your hosts, Will Porteous and Raju Rishi, will dive deeply into topics that are shaping the future, from satellite technology to digital health to venture investing and much more.
Raju: Hello, listeners and viewers. Welcome to another episode of RRE POV. I’m Raju Rishi, and I’m joined by my partner Will Porteous.
Will: Hey, everybody.
Raju: Today we’re going to discuss a topic that has been impacting both public and private markets: SaaSpocalypse. For those of you who’ve been living in a Faraday cage for the past 18 months, let’s define the term, okay? SaaS apocalypse describes market panic and structural shift where AI agents and automated development threaten traditional Software-as-a-Service business models. So, one important stat: $300 billion in SaaS market value disappeared in a single day in January when Anthropic launched Claude Cowork and Claude Code. So, it is impacting both private equity, venture capital, as well as public markets.
And we’re hearing the term all the time, and Will and I are here today to put some sanity into all of this, and some things for you to worry about and some things for you to not worry about and frame it in a way that helps you understand maybe a perspective, for whoever you are, whether you’re an investor, whether you’re a limited partner, whether you are a software company, or whether you’re a user. How should you be thinking about this? So, Will, what do you think? Is it dead and buried, or as Mark Twain once said, have reports of its death been greatly exaggerated?
Will: I think its death has been greatly exaggerated. There are certainly reasons for concern. I think that the castle has been breached, so to speak, that the monolithic power of large-scale Software-as-a-Service businesses, there’s no restoring, kind of, the perceived security of that business model. But those companies, particularly the leading companies, have an enormous amount of assets to continue to capitalize on, whether we’re talking about their installed base of trained users, the proprietary data that they preside over for their customers, the strength of their license contracts, a lot a lot a lot of things, and just the fact that they run a lot of businesses every day, and a lot of businesses depend on that infrastructure, and they depend on the security and governance models around those relationships.
So, I think what is definitely gone is the blind faith in a 10x revenue on these SaaS businesses that we saw as investors across the capital markets. And you only have to look at the growth of private credit in SaaS, the growth of private equity portfolios in SaaS to see how blind that blind faith was, and there’s been a big re-rating of this asset. But I’m not sure, I don’t believe that from a product standpoint, this architecture is gone, so to speak.
Raju: So, thank you, Will. That was fantastic. What we’re going to do today is we’re going to double-click into all of these things. We’re first and foremost going to look at the core threats to traditional SaaS, and then we’re going to follow that up and look at the Mark Twain angle, meh, which is we think this has been overly… you know, simplified the viability is still there, and we’re going to talk about a couple bunch of reasons why traditional software still is fine, and then finally, we’re going to talk about the realities. What’s really going on? What’s the blend between the two?
And actually, that’s not the end. We’re going to actually jump into what software companies should do, ones that exist that have gotten investment from private equity and are thinking about going public, ones that are early in their phase and are thinking about what the right business model and approach is. But let’s start with core threats to traditional SaaS and let’s start with seat-based pricing. The thought is, Will, that one AI agent can do the work of a dozen employees or so, thereby requiring fewer seats. How does that implicate the SaaS industry?
Will: Boy, that really cuts to the heart of the matter, Raju. If we view it very narrowly, it’s a pretty powerful assertion that says, you know, AI agents, fewer people, therefore fewer seats, unless the software companies really modify their approach on pricing. Now, I think the corollary to this is important because we’re having a discussion not just about the SaaSpocalypse, but about the way AI-driven software is going to change the deployment of human labor. And we have a wonderful and important economic principle in Jevons paradox, which basically says that when you improve human efficiency, you don’t just eliminate workers—this is what history teaches us—you tend to redeploy that labor to more productive ends. And so, I would say, if we’re thinking about how to run a SaaS business through this period, you should be looking at ways to reengage the rest of that labor. But you are, I think, going to have fewer seats sold.
Raju: Yeah. I think actually the seat-based pricing will collapse. I think that’s going to change because I think it’s just an artifact of how companies created a business model that worked for them. I mean, if you remember, Will, we had perpetual software licenses historically, where you charged a lot of money, and it was amortized over a period of five to ten years, depending on the size of the software, and the recurring revenue was in the form of maintenance, and it was usually, like, 20% of the software price. What you got with SaaS and what they kind of touted is you got continuous upgrades and you’ve got a notion where we looked at the number of users as a way to break down the overall costs.
I think that is going to change. I think we’re going to be thinking about factoring token-based pricing in here because there is, in this new world of AI, not an unlimited cap of resources. You are requiring an individual to use tokens, and a token is power and energy and compute and all of that in one, sort of, little bundle. And so, if there’s an open-ended cost loop, there has to be a little bit of an open-ended revenue loop. And so, I think there’s going to be that factoring in there, and I think that there’s going to be a different view, which we’ll get to in a little bit, on how pricing is structured, which I think can be even more innovative than looking at it in terms of a seat.
Okay, so but I do think that is a fair criticism of SaaS companies. Seat-based pricing doesn’t really depict the value that’s created, nor does it depict the costs that are incurred. Okay, now the second thing that the proponents of the death of SaaS are saying is that shallow wrapper companies that are built on top of LLMs are done. What’s your thoughts on that?
Will: Ah, boy, I’m not sure I have a fully-formed view, yet, on that.
Raju: I kind of agree that if you are a software company and the only thing you’re doing is creating a veneer with, you know, maybe a dashboard and some, you know, UI that is particulate to the industry, you’re kind of in trouble, if the LLM is, kind of, doing 99% of the work, and you’re just doing the presentation layer. Because those presentation layers can be, you know, created pretty easily in the new world. I think the reality is, not a lot of companies are getting funded that are shallow-wrapper companies. We certainly wouldn’t fund them, and I think the vast majority of companies have more depth than that and are providing much more value than simply a, you know, veneer on top of an LLM.
And so, I think the folks that are saying, “Hey, the wrapper companies are dead,” yes, true, but I don’t think a lot of quality investors are funding them, and I don’t think a lot of them have hit a market cap that is of substance, right? I mean, I think most of the companies that have substantial software capability and substantial data, you know, they’re not really just a wrapper company. They’re doing much, much more than that. So, all right, the third. Unless you have a thought here?
Will: Well, I just want to interject, kind of, I think… I think the implications of that are important to pause on for a second for the entire presentation layer, particularly as we move to a world of AI agent to AI agent interaction. If you’re replacing the human in the loop, the wrapper per se, or the presentation layer of something is frankly irrelevant in an agent-to-agent world, and that wrapper or presentation layer can be generated dynamically if there’s a human in the loop. So, it’s kind of a service layer if necessary going forward.
Raju: Yeah. And I agree with that. Okay, the third reason people are touting—so one, seat-based pricing collapse, which yeah, sure, but I think there’s going to be a shift—the second is shallow-wrapper companies are done. The people who understand software companies, they’re not betting on, they’re not categorizing a shallow-wrapper company in the category of the substantial SaaS company, so I think that’s overblown. I think that’s just, you know, there, yeah, sure. When AI came out, people came out and said, “I can build a CRM. I can build a—” you know, “—a Salesforce.com replacement,” and the reality is that those were shallow wrappers, and so those won’t exist, but we’re far from that already.
The third is build versus buy. The low-code cost code generation and agentic development tools. Now, companies are going to find it easier to custom-build point solutions internally rather than buying and subscribing to external software. What’s your thoughts on that?
Will: If that’s a true statement, it will ultimately be the end of the continuum of this problem which has dominated the software industry from inception, right? You and I have always lived in a world of build-versus-buy tradeoffs in, let’s just say, enterprise software development, whereas, in spite of all of the resources that companies could throw at internal development, they still ultimately brought platforms for performance, for security, for governance, for service, for all of that horizontal feature set that delivered them, or just for standards support, and that gave us a generation of incredible software companies. And I think that a lot of those considerations that made those products commercial grade aren’t immediately going away in a world of supercharged internal development with AI development tools. That’s my gut.
Raju: My gut. My—you’re dead-on accurate [laugh]. Dead-on accurate. Listen, there’s a reason why people buy. There’s a reason why people buy. It’s not your core competency. Not your core competency. Every single time I sit on a board and you sit on a board, and you know, there’s an idea generated by the company and it comes forth, one of the things, the lens I look at—I’m sure you do too—is, is it part of our core, or is this an ancillary function? Yeah, we can generate a little bit of revenue from it, and maybe it’s a way of padding the margin, and unless it’s massive, right?
The reality is, this ancillary tiny little bit of revenue that you’re getting from this other function, if you can buy it, it gives you more time and attention to put toward your core. And I think companies do the same. So, that’s number one. Number one, you know, I think build versus buy is going to be, sure, like, I can save a little bit of money by, you know—maybe, maybe—I can save a little bit of my money by building this capability instead of buying it or renting it, if you will, but I’m going to shift my attention away from my core capabilities. That’s number one.
Number two, I think the jury’s really out as to whether, you know, long-term support of this stuff can easily be done with AI. Who do I point to if the thing starts hallucinating? Who do I point to if, you know, the cost curve goes up or down? Am I going to, you know, start swapping out, you know, the worker models, if you will? Am I going to take Kimi K3 and swap it out for a different LLM that might be cheaper or have better capability, or am I going to let a third-party deal with that?
That, to me, is at the core of build versus buy. Do I want to now manage this infrastructure? Do I trust it? Do I want to manage it? Do I want to continuously update it? Is it really that easy? And oh, by the way, the costs associated with AI are TBD. We talked about this in our last podcast. We don’t know. Token pricing is subsidized, we know that, you know, there’s a bunch of stuff going on where the models that are open-source and/or, you know, distilled are going to be a lot cheaper, and so we don’t know whether the costs are going to come down dramatically yet.
If you’re basing it based on Claude pricing today, it’s one thing to develop using Claude and have a piece of software. There’s another to run everything through Claude or through ChatGPT or whatever it may be for perpetuity because you will be paying token costs that are largely subsidized today, and so buying makes it a little bit less of your problem. And the last is, I don’t think there’s enough subject-matter experts out there yet.
Will: [laugh]. Absolutely. And wherever they are, they’re not all working in your company—not enough of them anyway—to get it right. So, I think we’re firmly in the same camp on this. This is a basis, the buy versus build decision will continue to be too risky, too dynamic, with too many obstacles to success, and frankly, a lot of risk for the people who are responsible for this, who actually are under pressure to deliver performance in other areas.
Raju: I agree. And there’s probably one other minor piece, which is, software companies that build software for a multitude of customers tend to give you best-in-class operating behavior. If you know, best-in-class already, great, build it yourself. But if you’re, kind of, looking for experts in that sector to give you a vote, give you an opinion, to give you, sort of like, drop-downs or, you know, templates, you’re better off going with somebody who’s, you know, played in the space for a long, long time. Okay, so those are the threats. You know, there’s some reality in there, but there’s a little bit of, like, fluff.
Let’s look at the Mark Twain angle, okay? Specialized vertical SaaS platforms that possess proprietary domain data are going to remain resilient because AI agents still require a whole bunch of training, and they need to understand how to deal with secure data storage and how to deal with structured APIs to function, and so, the specialized vertical SaaS plays have an edge that, you know, AI can’t easily replicate. What are your thoughts on that?
Will: I think that is true, but the advantage hinges on proprietary data more than anything, more than expertise, more than install base. And I think we need to, in general, evolve our thinking about the software stack because that, in an AI-driven world, while that proprietary data—and we’ve talked a lot about data and data governance in the past—that AI is going to produce a more liquid, if you will, software product experience than ever before. And so, if you’re a vendor of one of these platforms—we’re going to talk about the pressure points for them in a minute—you have to recognize that your product has to now be constantly evolving. And the world has changed in terms of the idea of releases and shipping. You still need all the quality and security elements that we always insist on and good customers insist on, but it is a much more dynamic asset that you are now trying to put in front of these customer environments.
Raju: Yeah, and you coined it well. I’ll rephrase it just a little bit, and I’ll say there are elements to the verticalized SaaS platforms that are going to remain highly stable, highly stable, and that is the fact that they have a whole set of proprietary data, proprietary workflows, they have a litany of capability that is built around know-how that is very, very nuanced that I think is going to remain. It’s going to be hard to break that open with AI right away. I think user interface, I think tunability, I think basically saying I don’t want to ever see this piece of software, like, this line item anymore because these five things are important, but the sixth is not important, I think you’re going to be able to dynamically change that, and so the verticalized SaaS companies run the risk of trying to maintain their UI/UX and maintain the, you know, sort of, inner workings of the software as a single package, I think those might get actually split apart.
Will: Mm-hm.
Raju: And so, yeah, Mark Twain saying those guys are safe, but I don’t think the totality of what they provide is going to be safe.
Will: I agree with what you just said, and I think it creates a profound challenge, if you’re responsible for one of those legacy platforms, which is, my product if no one is really seeing or [laugh] touching it as my product anymore, right? How do I define the boundary conditions around my software platform, as it were, which is really now an agent delivering an application experience to other agents, sometimes to humans, with a dynamic presentation layer inserted in between, how do I define the four walls of that? And it’s not like there aren’t precedents for this, right? Very few people actually touch and feel and interact with a database, and yet, you know, there have been great database companies throughout the history of high tech, but the problem is going to move to the fore and how we describe the company’s capabilities, how we differentiate expertise, true enterprise-grade expertise in a functional area from what would be the equivalent of something internally built is going to become harder and harder and harder.
Raju: I kind of think of the notion of services, like a service call—
Will: Yeah.
Raju: —and I think the reason I’m bringing this up, this service call notion of it’s a dip that happens, right? So, you might have this veneer that’s AI, and you might have this software like, you know, dental practice management tool or construction estimating software or, you know, something, that is going to be a dip that you’re going to leverage. You’re going to call upon it as a function call, and it’s going to dip, and it’s going to provide you with an answer, and the veneer might be something that is much better GUI and much more attuned to what the customer needs, which means the pricing model is going to change. So, that’s why I’m alluding to this because I do believe the pricing model is going to have a massive ramification. One of the biggest ramifications of SaaS is not that the traditional ones that you know about are going to be dead; I think that their pricing models are going to change. Okay, another reason why they will survive: regulatory, regulatory, regulatory.
Will: I couldn’t agree more. You took the words right out of my mouth. The moment that we’re in right now has some parallels to the app ecosystem explosion that we all lived through before, and if we go back to your example of a moment ago, how do you differentiate applications in this space? If you’re looking for your contractor estimator AI tool, well, you pick the one that is certified by the government regulator, or that you know meets the compliance standards for auditability that your regulators, your auditors, et cetera, are going to need from an oversight standpoint. It’s that accountability dimension that you know you’ve got to have.
Raju: A hundred percent. I’m going to tell you two things that are going to happen. I’m a predictor. I’m going to be like—what is it? Carnac?—
Will: [laugh].
Raju: —but he gave the answer, right? And—
Will: The Oracle at Boca Raton.
Raju: [laugh]. The Oracle at Boca Raton. I think that’s good. That’s good. So, this is two things going to happen. There’s going to be an LLM that’s let loose on a bank, and it’s going to have both buy-side and sell-side information. And buy-side is going to request an answer to something, and it’s not going to divulge that it knows sell-side information, but it’s going to provide what it thinks is the right answer. And there’s going to be a shitstorm. That’s prediction number one.
Prediction number two is healthcare, clinicians, hospitals, doctors, margins are thin. We all know this. There’s a handful of hospitals where the margins are tremendous, right? Like, but they’re the behemoths of the world, right? They’re like the Mayo Clinic and the Cleveland Clinic and Memorial Sloan Clinic, and there’s a whole bunch of others that are, like, you know, basically skin of the teeth kind of stuff. And they want to make more money, and AI gives them the ability to make more money, and they’re going to say, “Unlock my data records, obfuscate them, so you don’t know it’s Raju Rishi or Will Porteous, but let that data be leveraged with AI, and let’s see if we can come up with some good outcomes.” The hospital makes a bunch of money, and we get new drugs or new treatments or new whatever.
And it is very, very possible to unobfuscate the data, and it will happen, and there will be leakage of records, and there will be a massive case. And I’m telling you, the regulatory as—look, AI just jailbreaked itself out of a firewalled instance where it kind of went out and decided to, like, terrorize Hugging Face [laugh] you know, and it’s hacking its way away, right? Like, no guardrails are going to prevent some of the—well, we need to have guardrails. We are in the first pitch of the first inning of AI, and until you can structure things that are regulatorily sound, I think you got to look at your traditional SaaS companies, software companies, and say they figured it out; they’re good, and until AI can kind of figure out—until you can be safe and sure that it’s not going to happen, I think people are going to slow roll this a little bit.
Will: Well, I think you’re completely right, and at the risk of repeating myself, I will just stress that I think data governance practices are going to be the—a certain amount of a customer firewall or a firewall in the customer relationship, so to speak, in terms of the longevity of the paying—what are you paying for?—in the present world, particularly as we grapple with the dynamic elements of AI becoming this distributed layer. We kind of each individually want to be able to cross that threshold with our personal data of our own choosing, and those companies that really pay attention to data governance practices, I think, will earn a lot of loyalty and have the opportunity to introduce concepts to their user base that they may actually want to embrace on the AI side, but let it be of their choosing.
Raju: Exactly. And the last reason why Twain might be right is adaptation of the incumbents, right? Like, do you think these guys are going to sit still? Do you think Salesforce is going to sit there and say, “Pft?” Do you think Zendesk is going to be like, “Ahh, let AI come? We’re good enough.” You know? I mean, you don’t think they’re building capability into it?
And I’ll kind of go a little deeper into this one because I kind of feel like they will adapt. And I don’t know if they can adapt forever, but definitely in the short term, you don’t need to destroy the market cap of all SaaS companies overnight, thinking that, like, it’s done, but yeah, maybe five years from now or some period of time from now, there’s going to be innovation that, kind of, supersedes it, but for now they’re sitting on a mound of customer data, a mound of it. That is, they have the relationship, they have the data sets, they have the dashboarding, and the question is: is the octopus that’s underneath the covers so tangled that they can never get to AI, or does AI kind of say, “I don’t care about the octopus. I’m going to deal with the octopus, and I’m the one piece of technology that can, you know, get underneath the covers of it.” So, whether an LLM provides the software to detangle, or you know, the companies themselves use small language models and some processing powers from the bigger LLMs to do it themselves, they have a potential of doing it. So, any thoughts on that piece of the equation, Will?
Will: Well, I think whether it happens quickly or slowly, I think we’re both feeling instinctively that that erosion happens, and that it happens through some combination of the, particularly the small language models, context engines. You and I have both spent a lot of time thinking about the power of context engines and knowledge graphs to basically normalize that data layer and make it consumable for AI. That’s actually where there’s enormous amount of entrenched advantage for a lot of legacy software companies today. So, I think there’s a lot of eating away at different parts of the franchise in a typical enterprise installation. I think that their customers want them to win, [laugh] ironically. Like, their customers have a lot invested in these relationships and in the future of those product platforms, that, you know, certainly doing it themselves doesn’t necessarily promise them. And so, where is the future of those product platforms if they’re depending on a lot of more generic AI tools? I don’t know.
Raju: The word SaaSpocalypse to me implies, like… overnight, you know? I agree that there is going to be degradation and a reshuffling of value that’s created, and so my thought is that that time period is longer than people think and the market cap erosion that happens overnight is unwarranted, but yeah, maybe your future earnings you discount a bit, but you don’t have to tear them down. So, what are the realities? What do we think is going to happen here? And I will give a couple of thoughts, and you should give a couple of thoughts on what you think is going to happen in the near term, right?
I think one, one thing that is going to happen, for sure—I have a few—but I think one, CIOs are not going to buy multi-year contracts. I think that is going to happen soon. I think the folks that are locked into five-year, six-year, three-year, two-year, even, agreements for software companies have got to be prepared for one-year cycles because the world’s moving. It’s changing, and if I was a CIO—and I was, [laugh] you know, back in old school days, my first job AT&T and Lucent—you can see a scenario where, like, you could get in trouble for getting locked in, right? What do you mean you signed a five year contract?
The implication of single-year contracts is pretty significant, Will, because it means that churn will go up. And there is a metric in SaaS called lifetime value, which will change, and so you’re going to have to rethink as a software company, what metrics do I count on for the return on investment of selling? So, SaaS companies, it’s not a perpetual license model where I sell you and I build the cost of selling into it; I’m basically giving away my cost of sales so that I can amortize it over the course of the lifetime value of an expected customer, which is three or four years. And, you know, I give up the commission structure in year one, but year two and year three is much more profitable in SaaS companies. Typically. That’s the way of the pricing model. I think we’re going to get one-year contracts, and I think we’re going to see some churn.
Will: Yep.
Raju: So, that’s a thought. Anything from your side? If you have one, great.
Will: I started out up front, framing this from the investment standpoint, and I want to come back to that for a second because it follows right on what you just described. So, from an investor standpoint, we’ve gone through this fundamental rerating of the SaaS business model. You and I have talked a lot about products and customers today, and you know, the punchline is that the monolithic software product is essentially going away. It is being transformed over time by all the forces that we’ve talked about, and with it, the long-term market leadership position of a bunch of notable companies is under threat, and the associated multiple of the business model is going away.
And I think this is, sort of, profound for us as investors because we’re seeing software become this much more dynamic capability. And, you know, when I roll that thinking forward, I think about what are we going to be willing to pay a premium for in the future? And it comes back to something you said, Raju, which is that deep product expertise, the expertise about what the product should do-is still going to be really valuable. The best-in-class market leadership expertise in a category, people are still going to be willing to pay for, but the box that they’re paying for has changed. It’s not this fixed thing that it once was. It’s far more dynamic, and with it, we kind of have to see a new appreciation of value… emerge.
Raju: Absolutely, I love that thought. I love that. I think business models are going to change for software companies. Soon.
Will: Yeah.
Raju: Very soon. Subscription pricing, if that’s what you call SaaS, you know, then you can maybe say, “Well, okay, there’s death of the subscription model, but the software companies are still alive.” I think the two aspects of this where I see it headed—actually three. Three—one, token-based pricing. It’s going to be consumption-based pricing versus subscription-based pricing. So, how much did I use your software? I’m willing to pay for it—
Will: Yep.
Raju: —versus, I pay you a subscription that’s an all-you-can-eat. And the reason I think that’s going to change is because the cost structure associated with AI mandates it. Requires it.
Will: I want our listeners to really hear what you just said because I think this is the most profound and important change.
Raju: Well, wait, wait, because I don’t know if it is the most profound. I think the next one might be even more profound because I’m, what, the Oracle of Boca Raton?
Will: You’re the Oracle of Boca Raton. Take us there [laugh].
Raju: I’m taking you there. The second is output-based pricing, which is, I am a help desk, right, and I charge you a SaaS subscription based on the number of seats. What if I charged you based on the number of resolved tickets or tickets created? Because now we’re going to a value system. Consumption-based pricing assumes we’re getting value. Assumes we’re getting value.
I don’t know if you always do, and frankly, you’re going to have that because you could ask a stupid question—like, not you, Will. You never ask stupid questions—but somebody else can ask a stupid question and leverage a bunch of resources, and the value isn’t being created, but the cost is being incurred. So, you have to have a little bit of that built in there, but I think something really elegant and interesting is output-based pricing, which is what am I asking this thing to do? Well, I’m asking it to produce a CIM—a company [unintelligible 00:35:17] Farsight, where they produce CIMs, like, instantaneously, effectively, for M&A banks—what if I charged you based on that output creation? Because you know what it costs you to manufacture and develop that, and if I’m getting a lot of value out of that, then I’m going to be selling output, not tools. Selling output, not tools.
Which is going to really unlock something special, which I mentioned modestly in one of our previous podcasts. And I know you and I have talked about this, but when you look at any technology inflection point, the way it rolls out, the types of companies that are created, typically, the first set of companies that get an inflection point are services-oriented companies. The second is tools, and then applications, and then platform companies. And we kind of said, “Oh, let’s just coin the platform companies winners today,” and make them kings and queens, and… bogus, you know, it’s all falling apart, you know, what we said two years ago was being realized. But when you move to output-based pricing, does it matter if a human does the work or AI does the work or an AI-facilitated human does the work because the cost of the output is the output is the output.
And I think we are going to see the birth of some very, very, very interesting tech-enabled services companies because the business model needs to shift. They could not exist in that way when you were competing with a software company that was just selling you a tool or an application, but since that’s going to go away, the seat-based pricing goes away, and you have this combination of, you know, tokens for cost accounting and output, all of a sudden, services companies can exist in a really interesting way. As long as they’re trusted, you know, and you’ve been certified and all that kind of stuff. So, I think business models are going to shift in heroic ways, in absolutely heroic ways.
Will: So, this is a profoundly powerful idea and what it—I think it gets forced on the market by customers themselves because they will refuse to absorb the risk of token-based pricing going in, the input cost side of things in such a dynamic way. And so, the implication of output-based pricing is that the vendor must, in fact, bear responsibility for that and drive to its lowest-cost architecture to produce the output and the results in order to remain profitable, and that can presage a very powerful era [unintelligible 00:38:16].
Raju: I love that, Will. And double-click into our last—our podcast two podcasts ago, which says, which models do I use? Do I use a frontier models for everything? What? No. Why? Why are you going to use a frontier model to do a simple calculation? Use a cheap model. Use something that is open-sourced. Use something that is distilled. You don’t need to use a frontier model for all of the services.
And so, that’s all going to go into the cost calculation. Token-based pricing, yes, but a token is not a token is not a token. That’s the truth. So, the complexity of this is really, really interesting. So, I think if they account for token-based pricing, I think we’re going to be thinking about outputs, and I think we’re going to be thinking about maybe services companies, tech-enabled services companies that might look like software companies providing that capability, and we’re going to have a brand new world. We’re going to have a brand new world, and I think that is going to come quicker than people think.
So, all right, last point, and then we’ll wrap this up because it’s been a fun, fun one. I love this one. You and I live in this space, like, we live in this space. So, what does a software company do? I’m a software company. What the heck am I supposed to do? I’ve been charging SaaS subscription model, [groans] I got all this stuff happening and hitting my thing, and I’m only getting one-year contracts now. And I expect churn to go up; what am I supposed to do?
Will: I don’t know. First of all, defend the castle, sort of, with your customers, with your reassurances about the security of their data and information, and about the fact that you are essentially changing the engine that delivers the result to them over time. You have to construct a vision of the future that embraces AI while giving your customers a lot of continuity. And frankly, they need to be able to look at you and believe that you have the balance sheet and the financial wherewithal to carry through that period. And a lot of great software companies have piled up a lot of cash and can underwrite a pretty dramatic transformation of their business. Customers need you to take them there and take them with you. If I were advising a portfolio of SaaS companies, which I’m happy that I’m not, I would be focused on that.
Raju: Well, after this podcast, we might have a bunch of consulting work. So [laugh]—
Will: [laugh].
Raju: All right, I agree with you. I agree. It’s a tough road. I’m going to tell you three things you need to do. All right, and then come call Will and I, and we’re going to—it’s just a million dollars a day; it’s not much. We don’t charge much, but we’ll give you good advice.
So, I would literally look at cannibalizing your seat pricing. I really would. I don’t think it makes sense in the new world. I don’t think—if you’re not going to cannibalize it overnight, at least have a game plan for it, right? When a contract renewal comes in place, which happens all the time, offer somebody something new and different. Think about output-driven, think about reflecting token costs into this, and think about maybe layering in some services for what they do. A good platform company has a bunch of professional services people anyway, so they can be put to work.
Second, I would say, secure your data and build that data moat in ways that is deeper than even you’re thinking about right now. The data that you’re looking at is not yours, right? I mean, if you’re a software company, it’s your customers’ data, but what is your data is the amalgamation of it and the, sort of, insights that come across multiple customer segments. And I would just do that, and I would figure out your segmentation so you understand at a much deeper level the sophistication of a particular organization may be the axis that’s more important, not the size. And most people just naturally go to size and industry. And there’s different ways of saying, “Hey, that customer actually looks like this customer, totally different size, totally different industry.”
And if the data sets align, that can give you a lot more visibility into how to present the data, how to sift the data, how to normalize the data, based upon the type of customer that matters. You’ve been sitting on this stuff for decades and not doing anything with it. Now, AI can give you the tool to do that with the permission of your customers, right? They want you to do it and you have the ability to do it, but that’s a securing of your data at a level that an AI engine or an upstart coming into this play can never do. And the third is I would build some sovereign agents. I would build your own agents. I would offer them for free for a little while, and you know, as opposed to allowing the outside world to, sort of, bring their own agent to the table. I would do that. I would start thinking about those three things tonight, and you know, I think you have a chance. I think you have a real chance, but you got to really change your business.
Will: To sharpen the focus for a minute on those sovereign agents and what they should do, I mean, that is the future interface point for your product, as it were, across your customer ecosystem, partner ecosystem, with other companies, sovereign agents, et cetera. I mean, that’s the powerful frontier of your product that you need to be investing in.
Raju: I still love Clippy. I love Clippy. I mean, Clippy didn’t do anything. It wasn’t good enough, but it was cute. I would love a Clippy on everything [laugh].
Will: You’re going to have so many eager, helpful Clippies trying to do so much for you.
Raju: I love Clippy. So cute.
Will: But that raises an interesting point because back to your emphasis on compliance and certification and regulation, as the walls come down, as software becomes more and more liquid and dynamic, our ability to recognize a trustworthy agent or product capability is going to become more and more paramount, and how we instrument that side of things is—there’s a crying need for this in the new ecosystem.
Raju: Absolutely. All right, I’m going to let you wind us down and take our users onto their evening activities, or afternoon activities, or morning activities.
Will: Well, Raju, thank you for leading a terrific discussion of the SaaSpocalypse. We’ve watched this unfold together during the first eight months of 2026. It has profound implications for investors, for customers, for software companies, most of all. At RRE, we’re fortunate to be able to live in this conversation, and we’re pleased to be able to share it with you, once again, in this episode. So, thank you for joining us, and we look forward to having you with us again soon.