Is AI Killing SaaS, or Making It Better?

Speaker A: With AI gaining more traction, there's a lot of changes taking place in the software industry today. I'm joined by experienced investors Ian Levine, Ron Weissman, and Bill McAleer to discuss the impact AI is having on the software business and how SaaS companies are adapting in this evolving era. Welcome to Bootstrapped the Lighter side, the podcast for B2B startup founders wanting to achieve success without giving up ownership or control. This podcast is brought to you by Lighter Capital, the leader in founder friendly financing for SaaS companies. Learn more at lighter capital.com.

Speaker B: In the first quarter this year, the term saaspocalypse entered the text lexicon as investors began questioning whether AI would fundamentally disrupt traditional SaaS. The public markets reflected that fear, with medium public SaaS company valuations falling to just 3.6 times revenue in Q1. And we saw recovery in Q2 rebounding to 5.1 times revenue. But that's still a far cry from the 14.8 times multiple that SaaS published public companies reached at the height of the market in late 2021. So is AI actually killing SaaS or is that the wrong question? I'm thrilled to be here to discuss this topic with three distinguished guests, Bill McAleer, Ian Levine, and Ron Wiseman. And I'll start by asking each of you to provide a brief introduction. And Bill, let's start with you.

Speaker C: Yeah, thanks, Melissa. Uh, I'm a venture capitalist based here in Seattle, uh, with Voyager Capital, and I help found the firm. And we're now investing out of our sixth fund. And we invest in agentic AI and physical AI and agtech. Uh, so that's our focus.

Speaker B: Great, thanks, Bill. And I'll go to Ron next.

Speaker D: I'm Ron Weissman. I'm both a VC and an angel. Recently, uh, stepped down as chairman of the Angel Capital Association. Uh, run AI Investing for the Band of Angels, the oldest angel group in America. And I've been a VC for more than 25 years, um, mostly with Apex Global and now with, uh, Nava, one of Europe's largest VC firms.

Speaker B: Great, thanks. And Ian?

Speaker E: Uh, yes, I am Levine. I, uh, 30 years selling, uh, workflow solutions, uh, for a variety of different companies and also incubated, uh, created, uh, uh, two of my own SaaS businesses along the way. Uh, for the last 20 years I've been an active angel investor. And today I'm one of the two managing directors of Launchpad Venture Group, which is the, uh, largest, uh, seed funder in New England over the last 25 years.

Speaker B: All right, thank you And I'm Melissa Widner, the CEO of Lighter Capital, a provider of non dilutive funding primarily to SaaS companies. All right, so let's get into the question. Um, is AI actually killing SaaS or is that the wrong question? Bill?

Speaker C: Yeah, well, uh, you know, I think what we have right now is uh, quite frankly, SAS software is in what I would call AI purgatory. Uh, and there's this uh, belief that AI startups will be uh, really creamed essentially by all of these AI startup, uh, you know, by uh, you know, all the markets, uh, you know, and I think what we've seen is uh, really a shift when we look at AI from what were traditional software applications, uh, that uh, were normally organizing activities, helping manage products, uh, helping manage business processes. And uh, what we're seeing now with agentic AI is we're moving from what was organizing activities to execution. Uh, so really uh, we're in a position right now where a lot of the uh, software companies, traditional software companies, are kind of in the penalty box. Uh, we're not quite sure why because many of them are deeply embedded in business process. They're using proprietary data and uh, they're really going to be quite frankly somewhat uh, hard to disrupt. Uh, but today, uh, all the valuations, all the investment activity, all the buzz is really around AI.

Speaker B: Uh, well, it's interesting. When you started out talking about what your firm invests in, you didn't mention the word SaaS at all, which is probably true for every VC today, but I would say probably 18 months ago. I've known Voyager for a long time. You would have started with we do B2B SaaS. So has your investment thesis really changed in terms of the types of companies you're looking at or are they just rebranded?

Speaker C: Uh, it's changed to some degree. I think what we're seeing is really the emergence of what's called physical AI, which is uh, AI and software driving robotics. So we're uh, focusing a bit more in that sector. Uh, then when we look at AI companies, we're looking for fundamentally a, uh, business, uh, case that's a strong business case where we've got a particular business process that the software can be embedded in. Ideally there's a moat that can be created by having their own uh, unique model. Uh, and then secondly that there's ideally some proprietary data they're working with. So those criteria could apply to a SaaS company. Um, but what we're seeing is most of the new opportunities are really coming from AI native applications, uh, and so that is kind of where we're spending most of our time. Uh, we do get occasionally pitched a SaaS company, but we have to see that that SaaS company has, uh, really some uh, agentic AI or some AI flavor to it so that it eventually it's going to be competitive in the market.

Speaker B: All right, I will put the same question to you, Ron. Is AI actually killing sas or is that the wrong question?

Speaker D: I think it is, um, the right question, but I would reframe it slightly that, um, is AI causing a rethink of software applications, including SaaS? Uh, we've seen some very dramatic movements since January of this year, from everything from exits to valuation to growth rates of SAS traditional SaaS compared to native AI. Uh, my own group is focusing on native AI applications, vertical market applications, deep applications in areas like defense intelligence, aerospace, healthcare and so forth. So we've shifted certainly, uh, where we're looking. But I think that the SAS world is going to be reevaluated, repriced, but also it's going to change because it will, will absorb agentic computing. Uh, the SaaS world is filled with great data, it's filled with great workflow. Uh, and those companies that are deep in their markets, I think have a high survivability. Those SaaS companies that were doing something lightweight in the areas like document management, information review, those kinds of things, uh, I think are going to have a very short life. But there is a wealth of value in today's existing SaaS companies that have very, very deep vertical market knowledge, very deep proprietary data, uh, and very, very deep relationships with customers. Just Melissa, think of how deeply embedded Salesforce is, not just as a piece of software, but as a, uh, as a way of thinking about the whole sales and business development process. Companies are not going to start ripping out Salesforce and changing dramatically how they track their most core assets, their customers. So, uh, I think the answer is it depends. But what's coming out of this I think is very good. And that's a rethink of how we invest, how we price, how we value traditional software.

Speaker B: Thank you. Okay, Ian, same question to you. Is AI killing SaaS or is that the wrong question?

Speaker E: Well, since I'm going third, I want to compliment the first two answers because they were really insightful and talked to many of the things that I think about. But let me start by saying I think the word killing is the wrong word. Um, it's just too strong. It's changing for sure. And then around the idea of what was software as we Thought of it, it was workflows or business process, uh, improvement tools that were created over the last 20, 30 years. And what SaaS. SaaS was a change in the way we think about the pricing of software and turning it into a, you know, a monthly recurring or annual recurring sort of pricing model. And AI is challenging both of those because most of the tools we built for BPO or workflow were built for labor for human beings to actually utilize. And one of the big changes with AI is that it can actually be the labor. So that's a massive change in the market. Uh, but I still believe, um, in terms of the word killing, the things like distribution models, domain expertise, criticality of the workflow flows that you've created, um, are all going to remain. And so software is still going to be around for a very long time. And lastly, picking up on what Ron said, using Salesforce as an example, let's not forget that Salesforce is, is a, you know, it's obviously one of the most successful software companies of all time with its enterprise grade, uh, solution. But then it's got 6,000 apps that operate off of it, um, because Salesforce offers higher security, because they offer distribution, because they offer easier, uh, integrations. And so there's just a lot to think about when you say, is software going to die? Where some of those apps are lighter, they may be in a little bit more trouble because the moat that existed with the code itself is disappearing. Because the one thing AI can do pretty darn well is just the raw coding. So that part of the value equation is definitely eroding with the AI, uh, coming on so strong. Yeah.

Speaker B: And that reflects what we've seen at lighter capital. We'll have that in our portfolio anywhere from 120 to 150 companies, all B2B SaaS, although they all say they're AI. I joke that, um, our portfolio is almost 100% AI, and last year it was maybe 50% AI and it's the same company. So, um, I think there's a bit of rebranding. But even companies that aren't, um, uh, as deeply embedded as like a sales force, we're not seeing churn. They understand the business processes, they understand what the customer's needs are. So it's been surprising that so far we're not seeing AI kill traditional SaaS companies, but we are seeing our traditional SaaS companies become a lot more cost efficient as they spend less money on developers and other things due to AI. All right, so let's get into, um, Bill, I want to talk about, we talked about public market SaaS, company valuations. What are you seeing in the private markets and what kinds of valuations should companies pitching you and other VCs expect to get in terms of as a percentage of revenue?

Speaker C: Yeah, thanks Melissa. Um, you know I, I think that uh, what we're seeing in the private market does somewhat mirror what we're seeing in the public markets. And that is that if you have a pure software company and you're not really able to articulate that you have some sort of AI element to your offering, uh, it's hard to get funding. Uh, number one, uh, something like 70 or 80% of the funding in the last two quarters has gone to AI based companies. So the funding is clearly going to um, something that has an AI story. When you look at the valuation side of this, um, what we're seeing in the public markets is if you have a traditional software company, modest growth, maybe you're getting a 2x kind of multiple. Um, if you have a company that is uh, growing and uh, has an offering, it has some sort of AI extension or some sort of AI story, then it's more like four to five times revenues. And then if you're AI native, which means it's a pure AI application, ideally not built on top of anthropic or OpenAI, but really something that's unique, uh, you're going to get a 10x multiple. And so we're seeing really uh, you know, this whole diversion, uh, or sort of trend I guess towards you know, very high valuations on AI native companies. Uh, we're also seeing pretty robust valuations on physical AI as well. Uh, so where you see software, embedded hardware, where software is really managing some element of hardware and uh, there actually there's a lot of money going into that particular sector and uh, again the valuations are pretty rich, they're in the 10x kind of range. So uh, you know, whether this sustains or not is sort of a bigger macro question on what happens with the overall uh, productivity that people are actually getting out of these applications. Because I think that's sort of unproven at the moment. Uh, a lot of these native applications, native AI applications, particularly in the generative AI space, um, are seeing a lot of churn, um, as it relates to then uh, more the agentic AI. A lot of them haven't been fully implemented, haven't seen two, three years of experience. And so they're really trading on a valuation basis on what is unique and whether there's a moat that could be created. And of course you're getting a much better valuation if there's a combination of a proprietary data, as I mentioned earlier, small language model or distinctive language model, and then also a unique business process.

Speaker E: Well, I have an opinion on that one for sure. First of all, we have a relatively small portfolio launchpad, 50 companies. But we had two exits last year. One was what I would call software and it was a 6x uh exit for us. And, and then, and then we had uh, ah, another one that was a 13x for us and that was at um, roughly 18 times revenue at the moment. So you know, and that was more AI ish. But I think it's really dangerous what we do where we just use multiples like 3, 5 and 14 because you have to unpack that. And if you look at software, um, uh, M and A over the last five, seven years and even longer, it's really, really segmented by vertical and it's really segmented by growth rate and by vertical. Even like three years ago, MarTech was only selling it two or three most of the time unless it had hyper growth going on where cybersecurity or maybe something having to do with the cloud was double digits, 10 or 15 exits. And so there really has been this uh, uh, sort of further bifurcation that's been going on behind these multiples that we constantly use that I think is really important to not lose sight of because we oversimplify where software is worth 10x and let's go. Um, so that's a really big point, uh, that I think we uh, need to focus on. Hassan it.

Speaker D: I would agree with Ian that things are, must be segmented vertically but there are some um, real torpedoes in the water right now, uh, that will affect everyone. First, there have been no SAS IPOs in the past six months, uh, and there's very, very few on the docket, uh, perhaps canva, uh, but there are very few that we're looking at. Second uh, is we have nearly 500 uh, SAS unicorns that were pre AI, that is they will be getting older valuations and those companies, uh, according to some folks are waiting to die. They will not get new financing, uh, and they may not get exit opportunities given how slow the exit markets are. Um, third, uh, I would say that emergence, uh, has a very interesting white paper out there which shows that native SaaS companies are growing at 4x, uh, the speed of even bolt on AI on top of SaaS, uh, and these do suggest that a long term cyclical change is happening.

Speaker B: Sorry Ron, did You say you said

Speaker C: native, I think you meant native AI.

Speaker B: Native AI, not native AI.

Speaker D: Native AI is growing forex, uh, faster, uh, than the non native, the non native like the bolt ons, to existing SAs, uh, and this has been attributed to the fact, as Bill said, that we've moved uh, from uh, helping employees get their work done to doing the work for them, uh, and that's happening um, pretty rapidly with agentic computing. I just want to make one last comment. Uh, Bill talked about physical AI and I think the broad secular uh, movement towards robotics, towards defense, towards aerospace, towards hard tech, towards data center technologies, even things like waste management has been building steam for about a year, year and a half, uh, and it's intersecting at the same time that we're seeing traditional software only SaaS, uh, take a significant hit. So we're in the middle of I think a secular change, um, for lots and lots of reasons and that we as investors and uh, as entrepreneurs should be willing to rethink what we're building and for whom around these trends. I think that uh, for me the next major trend really is about physical AI, but not limited to data centers. We're talking about longevity, we're talking about lots of work in health tech. And as I said even humble waste management is seeing a major upswing in that new shift, not just because of robotics.

Speaker C: Yeah. And I would say that to add to that, uh, we're looking quite a bit at AgTech agricultural technology and the application of software uh, AI and uh, physical AI to that space. And uh, there's a number of factors that are driving demand uh there now even despite the fact that the sector is a tough sector because of what's happening with the economy and uh, the oil and fertilizer and so forth. But what we're seeing is a real emergence of physical AI in the agtech space. One of our portfolio companies, carbon Robotics that does laser weeding behind a tractor to pull behind, uh, has seen extraordinary growth. And you know, I think it's an indication of uh, the buyer is looking for, you know, a strong value proposition and the application of robotics into certain sectors like agricultural tech or defense, uh, you know, is definitely a wave that we're going to see.

Speaker D: The European Fund I advise Naver, um, based in Turin, which invests worldwide. Many um, of their investments are in aerospace, uh, in uh, aerospace related technologies, agtech related technologies. And so I'm seeing exactly the same thing. And this is again not new for 2026. This has been going on now for three or more years.

Speaker C: Yeah, Uh, a couple years.

Speaker E: So my question comment back would be do you think some of this is because of new data capture, because of AI, in other words, I've seen some ag tech stuff that uh, things that we couldn't capture in the field before are much easier to capture with um, AI. And uh, uh, just all the changes in what can be captured with photography of different sorts and that's creating hard tech opportunities that are backed by AI, that in the end are software running.

Speaker C: Well to that point you're seeing emergence of a lot of drone technology that uh, is uh, looking at the fields, is uh, monitoring, we're out road tracks, uh, you know, so a whole bunch of uh, applications in the drone space which again sort of goes into this physical AI category.

Speaker D: Underlying much of this is the growth in new sensor technology.

Speaker C: Yeah.

Speaker D: Uh, because we're applying sensors to be measuring, you know, everything from the, the output of seeds, uh, to agricultural inspections. Uh, and of course the next gen of robotics is going to be very, very much dependent on the quality and the ubiquity of sensors.

Speaker B: Ron, you recently posted an article on the Angel Capital Association's website titled the SAS Apocalypse Implications for Angel Investors. Um, can you give us the abbreviated version for those who like to learn by listening rather than reading?

Speaker D: Sure. So the um, uh, I, uh, made four points. The first was Wall Street's negative reaction to anthropic Cowork, uh, particularly its extensions for legal, finance, HR and sales, showing how agentic technology, uh, could not only provide better data but actually do the work themselves. It was Jefferies that coined the term SAS apocalypse, noting that the per seat revenue model of SAS would be threatened as agents replace humans. So by the end of Q1, market multiples, as we've said, and virtually every SaaS category fell by 50%. We lost $2 trillion in the public markets. We've seen very few exits and OIPOs and so forth. Um, the second point I made is that much of the SaaS apocalypse is probably overhyped, uh, in reaction to things like the war against data centers or the fear of the jobs apocalypse. I think things have been blown out of proportion because there are many categories of SaaS that cannot and will not be replaced. Think of regulated industries like healthcare or telecom or aerospace, where every change in software needs multiple levels of approvals. And uh, your software validation, um, model may be binders thick. These aren't going anywhere fast. So again, my second point was I think we're overhyping this. My third point is that uh, whether we're seeing SaaS Apocalypse now or SaaS Apocalypse Never. The events of Q1 are causing us a big rethink of what does SAS mean, uh, for angel investors? Um, I have suggested that they do several things. First of all, look at which SAS segments are really vulnerable, uh to AI. Uh, things like record keeping systems, uh, data manipulation systems, uh, lightweight analytics and focus uh, on less vulnerable segments, things that relate to security, absolute systems of record, uh, systems where there have to be humans in the loop and so forth. Many of these are uh, m much less vulnerable uh, uh than lightweight SaaS. So if you're going to invest in SAS, uh, look at vulnerabilities and look at deep SaaS applications with deep moats, deep data, deep workflow. Uh, and finally I said that angels need to uh, re evaluate how they do their investing. We're seeing SAS move at the speed of light, uh, where SaaS companies are moving very, very quickly. They don't need uh, as, as much capital as they used to. There are companies that have gone from 0 to 100 million in seven months with a staff of 6, 7 or 8. And so angels need to be aware that capital requirements have changed. But most importantly, in my experience, uh, what AI companies are looking for is angel and VC groups that have a large existing AI, uh, body of investment to partner with, to learn from, to sell to. So you need to really up your game and become part of the AI community not by dabbling, but by going deep. And finally, because the speed of AI is so fast, we'd be much faster and much quicker about how we do our decision making around deals. CEOs are not going to wait for us. So we need to become smarter and deeper about what's really AI to come up with decisions faster to meet the time frames of these very, very fast moving AI companies.

Speaker C: Well and I would reinforce that. I think the velocity of what we're seeing in this trend, and I've been through three major tech um, trends here, uh, the velocity here is unparalleled. And I think the advice that Melissa, uh, we would also want to give to your companies is really be aware that you've got to really ramp up the velocity. The velocity of uh, development, uh, which is now being really uh, enabled by uh, Claude and other, other applications, uh, AI applications and also velocity of go to market, uh, both of those are being impacted quite a bit by AI.

Speaker E: Uh, um, just to add on things we're looking for at Launchpad is even deeper domain expertise. It always matters. It matters even more now. We also think distribution has become super critical and unique. Distribution is becoming a new moat. And so if you can hit it in a unique way and then the last thing is just where is your data. In other words if you're just entering AI by handing it over to the big models, that's a no, no in our eyes that just isn't a value creator. The way owning your data, having proprietary data, learning from your proprietary data and having that closed loop system that builds value over time. And so when we look at software we're looking for those things. And back to the physical AI comment earlier where we often used to say to ourselves hard tech's hard, be careful. As an angel investor we're actually starting to just flip that one uh, hundred eighty degrees and say yeah, a little piece of hardware to this solution to could be a really good thing because it goes back to in this age of AI, the code is no longer remote, that it was just not as deep.

Speaker D: Just a comment on that. I completely agree with you Ian. Uh, also uh, that software is more protectable, more patentable if it has a strong link to hardware. Ever since the ALICE decision uh, rendered most software patents uh, uh, not very defensible. One of the strongest defenses is tightly integrated hardware and software. So wrapping your solution uh, with hardware is uh, for many, many reasons a much stronger moat uh, than software companies have used to have.

Speaker C: Uh, one other piece of advice I guess to give to uh, the listeners here, uh, if you're in the physical AI ah, company, um, or you're starting one or you're pitching one, uh, one of the other things that you really need to be able to articulate and understand is the interplay of the hardware and the software and the fact that in order to scale a physical AI company it's a lot more complicated than a pure software company because in this case you've got manufacturing of the, of the unit or of the hardware, uh, you've got to make sure that your cash flow is aligned right between demand and supply. Uh, and so it is a more complicated model. And what we're finding is we do see some folks that have really good technical chops that talk about the solution. But then when we ask them okay, how are you going to go implement it? And also what's your distribution network going to look like? Are you going direct, are you going through a dealer? Uh, they're kind of lightweight on those answers. And so the uh, other advice is beef up your team, get people in that understand the subtleties uh, of a hardware company that even though it may be largely embedded with software.

Speaker D: This is true especially for CEOs. Um, if you're a CEO who's never run a hardware business before, get help fast because you're not going to have all that much credibility if you say, well I built these two software companies, but hardware is radically different as uh, Bill just said.

Speaker C: And domain expertise is really important too. So if you're doing one of these opportunities and you've never done agtech before, you have to understand how agtech works to make sure that, that when you're inserting this into the business process that uh, you understand how it's going to impact the business process.

Speaker D: We've talked about uh, about the potential collapse of some SaaS companies. I'd also want to suggest that there's a potential collapse of many AI companies, companies that have lightweight domain knowledge who

Speaker B: don't understand the customer problem.

Speaker C: Yeah, right, Yeah. I mean the other thing that uh, people are ignoring in the Saskopolis ah, uh, thing is uh, when you look at uh, some core SaaS companies, you know, that are successful, uh, there's a couple of factors that need to be considered. One is that there's high switching costs if you're embedded in a process. Uh, secondly, and sort of it's once in, it's slowly out. Right. Um, AI software as is currently being marketed and so forth is not always up to the security requirements of the enterprise. And so that's another issue that I think is important. And then uh, where I do think there's going to be disruption is in the pricing. Um, so while you may be embedded and you may have high, you know, you have all these elements of high switching costs, good data, et cetera, I think the pricing model is probably going to get disrupted. And that is where, if you look at uh, valuation impacts, I think that's kind of where the street and us as investors are looking to say, okay, what do we think that pricing model could be? Is it going to be a volume based pricing model as opposed to a seat based pricing model? Is it going to be multi year, single year, single use? So I think there's some things still unsettled in terms of the four or five variations of pricing models that are out there for AI today.

Speaker E: I do think one of the things that's going to change though Bill, it plays off what you just said is one of the switching cost problems was implementation. A enterprise grade salesforce.com implementation can take a year. Uh, and with AI there is a belief that that might be knocked down to weeks. Um, now still Needs to be proven. But that is one of the beliefs that out there. And so if implementation gets really easy, then rip and replace becomes much easier as well. It's got a ways to play out. But I mean uh, Pegasus, which is another CRM company out of the Boston area, highly successful, been around for 20, 30 years. You know, they've built AI agents to implement their, you know, their implementation is AI driven, you know, and it's sort of, I do think that is the future.

Speaker B: I want to dig into that a little bit. Ian. So if implementation is easier than rip and replace is easier. Right. Okay, so what, so what does that mean?

Speaker E: Well, it just means when, when, if you're, if you're. I used to run I, I own salesforce.com at a Fortune 500 company 20 years ago. And for me to make the decision to switch to Microsoft or any other system back then you had to look at a lot of factors. You know, how many systems is it touching? And, but I just invested a year.

Speaker B: No, no, I understand that. But now somebody approaches you and they go, we have all these features, you know, we just started. It's native AI. We have all these features and we can implement this in two or three weeks and give you everything that you, you need that you're using with your whatever Salesforce or whatever system you're using. So, so let's, let's just play that out. What actually happens? Yeah, well, if Salesforce go okay, they can just take it.

Speaker E: Well, Salesforce isn't going to go okay. But let's get to the cost question at hand.

Speaker C: Salesforce will buy them.

Speaker E: Yeah. Is my license with whoever and is it a ten million dollar a year or fifty million dollars a year enterprise class license? And can I do that for 1/10 of the cost? If those are the types of magnitude that you're talking about, people are going to do it. Some of it comes back to uh, the whole customer success thing. Are uh, companies truly tracking the outcome value they're getting from the software? There's always a hypothesis when you go in that you know that you're going to create this incredible value. And let's just say most good software does create a lot of value. Saves time, makes life easier, maybe lowers expense, but it rarely matches up to the, you know, uh, this is going to save me 10x productivity because if it did, every time somebody, the multiplication on that is impossible because big Enterprises are buying 50 pieces of software and everyone's supposed to deliver 10x and it just not what's happening in the real market. So back to your question. I actually think that cost savings will drive it. People are going to start looking at their utilization of a software in a different way. They may be paying for 5 or 10,000 licenses and they see that only 500 are really using this application on a regular basis. And that's going to be a big change. And so you, you start getting, uh, you get into all sorts of things. There's going to have to be adjustments by the big incumbents around pricing models, around pricing in general. They may have to give up a little bit of margin and they have the room to give up margin to hold a customer. Um, but I do think you are going to see where you haven't seen it in your portfolio yet, Melissa, over time we are going to see rip and replace happening because of AI. It's just not happening overnight.

Speaker B: I just want to dig into that rip and replace because who is better if it's all because of cost? You're saying we're going to get the same features. You're a big company, you're going to get the same features. Ripping out Salesforce and going with this new company, it's not going to be that painful. The implementation isn't going to be that painful. I mean Salesforce, I would argue that Salesforce is not going to lose their, that customer. So if it's going to be based on price, they can drop their price. And nobody's better positioned than Salesforce to understand the customer's needs.

Speaker E: Right, I agree. This goes back to Ron's comment of what type of company. So I agree with you on Salesforce because of the complexity involved there that they're not going to be experiencing a lot of this. But there are tons of software companies out there that have much simpler workflow solutions that are going, this is going to happen to them because remote's just not that deep. And what they're doing was they solved a really important problem. Uh, I invested in a company that solved reference management for high tech companies. It was a great product offering but you know, it was, it was a small market. Call it less than $100 million total TAM and there's a couple companies in it. I just don't see how a genic AI is not just going to, they're all going to get ripped and replaced over time. It's just that simple.

Speaker B: With the domain expertise are going to be the ones that are best placed who already have the customer relationship, who already have the domain expertise are going to be the ones that are best placed to service. That customer and now their costs will go down dramatically. So they can afford to service the customer at you know, half a quarter of the cost they were previously charging them. I mean I just, this, this big churn, I just we're not, one, we're not seeing it happen and two, if it is going to happen, it's the incumbent that is best placed to keep the customer or sell to that customer, sell new products to the customer.

Speaker C: Well, I think the subtlety here though that maybe we, we need to emphasize is the fact that a lot of these applications are organizing or processing or tracking, right? And when you get into agentic AI and now uh, it moves into execution, that's where then the value proposition is a lot stronger. So I E, you know, if you have a Salesforce application, it's great, it's tracking you know, your leads and uh, you know, what's happening with your leads. But then if you can actually eliminate the outbound by saying, okay, now I'm going to use an AI agent, a voice agent to contact a customer to do some action, whether it's loyalty, whether it's follow up, whether it's whatever it might be. Now suddenly you're eliminating some people potentially, but also you're enhancing where the application was. So I think it's going to be critical for these traditional SaaS companies to layer in that execution, uh, part of their solution for it to be sustainable. Because that's where I think the disruption is going to happen is in the fact that these agents are actually go out and execute uh, actions on behalf of the uh, user.

Speaker D: There's one more aspect here and that's risk. It's not just about cost, it's balancing cost savings and risk. Um, many CIOs got fired in 2000 when the bubble broke because they were buying technology from immature companies who ultimately died. And so the scrutiny of companies in terms of their viability, their partnerships is an important part of the equation. And let us not forget that every day we process around $3 trillion worth of transactions using COBOL solutions. You know, existing solutions may be pretty sticky and for major mission critical applications there's going to be a lot of validation, uh, a lot of vendor sourcing, analysis and so forth. Uh, so yeah, for lightweight applications I think we're going to see some rapid transition, but for mission critical stuff I think it's going to take a lot longer. Uh, we're seeing mission critical projects being canceled, uh, with AI at a pretty great rate and CEOs are complaining, I'm not getting value, I'M seeing risk, I'm turning this off until this stuff is more mature. So let's temper our enthusiasm for cost savings by understanding that real companies evaluate risk as well as cost savings.

Speaker C: They're also finding that, uh, they are going to have to pay more for the tokens too. So the initial ROI that they thought they were going to get is going to be a little bit different because, uh, the token cost is going up at least in the near term here.

Speaker B: Let's get to a few quick questions here. When are we in a bubble? Are we in an AI bubble?

Speaker D: Uh, yes, I believe we are in an AI bubble where AI companies are selling and borrowing from each other to fund this enormous build out. Uh, it is very hard to imagine revenues in the future for OpenAI, Anthropic and the others that will justify in the near term the massive capital expense. So I think we will see a retrenchment and I definitely believe we're in a bubble near term, long term AI will solve humanity's problems. Near term we have to pay for it and finance companies to be able to stay alive. And I think we're already seeing some cracks, uh, in the AI ecosystem around fundability. Just look at what's happening to companies like Oracle right now, uh, that really can't afford the build out that they're trying to fund. So I think the short term is messy, long term is great, but we currently are in a bubble, in my opinion. In a, in a financial bubble.

Speaker E: I think we're in a bubble. But inside the bubble there's going to be some really big winners. But it's a bubble because even some of these largest heavily funded models are going to become commodities and they're not priced that way today. And at the other extreme of the market, these models allow you, through vibe coding other things to launch a company faster than ever before. You can be, you know, an in business software company with a product and that doesn't necessarily mean you're going to be successful. These, you know, we call, you know, the terminology AI wrapper companies. It's just sort of like they're easy to launch now, now way easier than it was to create a software company 20 or 30 years ago. But, um, you know how long they're going to survive, there's going to be a lot of flame out in them basically because that's all they are, AI rappers and they don't, they don't have enough value.

Speaker D: Yeah, absolutely, absolutely agree.

Speaker C: It also feels to me that we're in a bubble. Uh, you know, A lot of the money that's gone into the investment in AI is, is in the infrastructure layer. Um, you know, if you look at uh, I think the estimates from one of the analysts, ah, it's a $1.4 trillion AI market. But you know, of that 450 billion is going into applications, um, and the rest is going into infrastructure. So we're at the infrastructure build out layer and um, there's going to be lots of opportunities for applications on security tooling and all the typical things that will follow the infrastructure layer. But I um, think initial large language models, ah, particularly OpenAI are very capital inefficient. Uh, the question really is can the public markets continue to absorb the level of equity and debt that's being created not only from the large language models but also from the hyperscalers. So there's huge amounts of capital going in to this build out and it's a build out that's built on largely somewhat inefficient models. So if the models get more efficient, uh, we're going to, it's going to be interesting to see where the money goes. And I think all we need is something like OpenAI to uh, hiccup or blow up, whatever and suddenly the whole thing uh, will change.

Speaker D: Two related points, Bill. Uh, one is the AI buildout you mentioned is strangling the rest of the early stage investment.

Speaker C: That's right. It's absorbing a lot of capital.

Speaker D: Well, 86% of all of venture capital in the US in Q1 and Q2 went to AI. 86%. That's up from 65% last year. Um, the second point is, is that we've already seen what Deep Seek did last year, uh, which was a torpedo in the water, uh, against the traditional AI build out. Now it did not last, but it's coming back. And third is there was a tremendous amount of research going on. It had a cost reduce AI. So we're looking at green AI to reduce these capital expenses. We're looking at uh, leaner matrices, we're looking at sparse data ways to really change the game. So we don't need these mega rounds that we used to, are they here now? But they're going to emerge pretty fast. So I think that there's a great fragility, uh, in the assumption that we're always going to be spending billions of dollars on AI build out. That's going to change hopefully in the next two years.

Speaker E: Melissa M. If I could, the word bubble. I was thinking about it as both Ron and Bill were talking and buried in that word is valuation. So are things grossly overvalued? That's part of a bubble. I guess people could challenge that definition. But that is how uh, I think about it. And it is an interesting time because Some of these AI companies are growing 10 or 100 times faster than any other companies we've seen. Now that's not the m majority of them, but they are out there. And we have talked about hyper growth rates are a crazy valuation driver, more so than earnings or whatever. And um, you and I had a conversation in the past about ultimately valuation is all about durable cash flows over time. Do you truly believe they're going to continue? And we, we all got really hyped up on the rule of 40 over the last decade. It's a great rule. We all kind of measured against it, thought it about a lot as we did every um, you know, uh, investment that we made. And then, you know, one of the things that happened that, that's a little less known is Bessemer did a, a, uh, study on this about four or five years ago and they created a new model called the Bessemer 3x which says, yeah, the rule of 40 is right, but growth within the rule of 40 should be valued three times as much. And that's how you arrive at the valuation of a company. And I, you know, I, I'm not sure the Bessemer 3X model isn't still the right model. It's, it's, it's, you know, uh, it's, it still makes sense to me. Growth covers all sins and, and is where the bigger exits are found.

Speaker D: Well, growth covers all sins is obviously right, but look how much.

Speaker B: As long as the economics are correct, I have to, you know, as long as there's positive unit economics, look at

Speaker D: how much of AI spend is being AI companies buying from other AI companies at some point when the merry go round stops, who is going to be funding that kind of growth in the future? This is not sustainable growth if it's a handful of companies largely buying from each other.

Speaker C: Yeah, that's a little bit of what we saw. And I mean that is what we saw in the early stages. The Internet.

Speaker B: Right.

Speaker C: Startups, uh, buying from startups.

Speaker B: Right. Or, or a company that never had a sustainable business model but they were growing so they could keep raising at a higher valuation. Right, um, we want to get to a couple other things quickly, but let's talk about the exit market.

Speaker E: Well, I'll talk one second. On the exit market, one of the more fascinating things, if you look at the data over the last several, uh, years is AI companies are being bought by Strategics. And historically, over the last couple of years, private equity has been the buyer of software. Now that slowed down a lot in the first half of this year, but it's still happening. Um, and so it's just, it's just interesting that strategic buyers want the AI companies and the private equity folks who have too much dry powder, they have a problem. They have so much money from their LPs they need to put to work. They're looking for ways. They are still willing to just look at cash flows and try to figure out durability and put money to work. Not at the pace they were. But I just find that one of the interesting dynamics within exits, uh, these days.

Speaker C: Well, one of the, uh, on the PE front here, uh, you know, there's about $500 billion of dry powder in the PE market today. Um, and there's 4,600 companies that PE has backed, um, without exits. So they're all sitting there, software companies. They had 79 exits to date in 2026. So if you do the math, that's 28 years of backlog to, you know, exit at the current exit rate. Now, hopefully the exit rate will improve. But to your point, you know, PE investment is down 56% in the first half of the year. Uh, and the tech deal value overall is down as well. So what that has led to is a shift, as you mentioned, to strategic buyers. Now a much higher proportion of the deals being done. It's basically in 26 to date. It's 62% of the exits have been strategic buyers versus 28%, uh, pe, uh, versus, I'm sorry, 28% in the prior year. So Strategic has stepped up and they're buying some of these AI, ah, companies and maybe a few software companies. But the, uh, PE guys are definitely out of the market.

Speaker D: This has significant implications for venture capital funds and for their ability to raise cash in the future. Um, given the fact that this enormous backlog of unicorns is probably not going to exit very, very well. Uh, and yet many VC portfolios are filled with companies that look great in 2021 and don't look so great today.

Speaker C: Well, and usually what we'd like to see is if we're running a process, oftentimes in the past, what would happen is the PE firm would make the first bid that would get the process going, and then ideally we get the strategics to kick in who usually paid higher premiums. So that dynamic has Changed now that uh, the PE guys are sort of sitting on the sidelines. Unless it's sort of a purely agentic AI kind of company. They're just not nibbling at any of the software companies. So Melissa, for some of your companies where maybe they've taken a loan and they're thinking about, you know, when do they exit. Uh, one of the things that they have to really think through is the fact that the PE buyers today are not buying anything that's traditional.

Speaker B: They're buying more unique assets and traditionally lighter capital. Most of the exits for lighter capital portfolio companies have been to P.E. firms. Yeah, um, yeah, but that said, we are still seeing exits sort of in the 4 to 6 multiple range. And these are companies that grow. Are growing, but not hyperscaling. Um, okay, we are getting to the end unfortunately, because I think there's so much more we can discuss. But I want to ask each of you what is one prediction you could make today for what the market will look like? We can say five years or ten years where you, you'll either look brilliant or foolish for making it five years from now.

Speaker C: Okay, well, I'll jump in. Uh, I think OpenAI is going to have a hiccup of some sort. Um, I'm not sure if it'll blow up entirely, but I question whether it'll make it public. And if it doesn't make it public, it's a lot of implications for funding the company.

Speaker B: I love it. That's a bold prediction. You're either going to look brilliant or foolish. Bill.

Speaker C: I give it a try.

Speaker B: All right.

Speaker E: Yeah, I'll take a stab at it. So, um, uh, I believe that this agentic AI is real and its first massive impact back to the concept of uh, the salespocalypse around labor is going to be the sales and marketing departments. They're going to get gutted really fast and really hard. And I do believe that most people are saying that it's going to be the software programmers today and I'm going to jump over uh, uh, uh, the shark here and go to sales and marketing.

Speaker B: Huh? How?

Speaker D: Okay, um, a prediction. All right. I think that in the next three years we're going to see uh, that agentic is going to win. But uh, we're going to move from a hype driven market back to fundamentals. We're going to see a massive shakeout uh, amongst all of the me too wrapper companies. Uh, and companies will suddenly become valued again based on fundamentals, not on how great they're going to be. Ten years from now, uh, and we're going to move back to a more normal investing environment, a more normal development environment. Uh, the winners will have emerged by then, uh, and we'll see market consolidation, uh, slowdowns in company formation, and a focus on, uh, the few winners in most market segments, as opposed to, we'll fund anything with the word agentic, uh, in its name. So we're going to, I think in three years we'll have a more stable, more rational market. Uh, and I'm, I am praying that that happens.

Speaker B: Great. All right, well, we will wrap this up now, but thank you, Bill, Ron, and Ian, for a great discussion today. I think we should do this again in six months or a year from now. See how. See how much things have changed and how much our views have changed. And that's where we'll leave it. I want to say a big thanks to my guests, Ron, Bill and Ian. And yes, it's pronounced Ian, not Ian. Sorry for my mistakes earlier in pronouncing your name. Thank you for bringing your incredible insights. I know the audience appreciates it and I hope we can get you back in the future.

Speaker A: Thanks for tuning in to Bootstrap the Lighter side. Don't forget to hit that subscribe button on your favorite podcast platform to catch future episodes. If you found this helpful, why not share it with other founders who might benefit? And if you want to learn more about founder friendly financing options, just head over to Lighter Capital Dot com.