The impact of AI on SaaS - Lars Pedersen, CEO, Beqom

Speaker A: I think companies, software companies will be faced with three different risks as it pertains to AI. So the first one is upstart, uh, company AI, native company vibe code, something that blows you out of the market. Uh, and I think for most software companies the risk of that is quite low. And then the second risk is you end up in a situation where your customers, uh, where they decide to do it. And I think that will happen. So I'd say that risk is maybe low to medium. There will be some that are going to do that. And then the third one, which I think actually that risk is high, well that's your existing competitors, they move faster than you and that one is real and that's the one we should worry about.

Speaker B: Hello and welcome to the Pep Talks podcast. This episode is all about the impact of AI on SaaS as a business model and, and investment category and more broadly, what all PE back leaders should be doing to stay ahead of the competition. We are joined by Lars Pedersen, a hugely experienced technology and software as a service CEO, uh, on his sixth PE back business, having operated across hardware, payments and software companies throughout his career. Lars shares two PE backed businesses and is currently CEO of Becom, a SaaS platform for performance and compensation management and pay equity, typically for large enterprise customers. Lars discusses the three real threats facing SaaS companies from AI and the five characteristics that separate resilient businesses from vulnerable ones. He also discusses what PE investors are actually looking for and how PE back leaders can think clearly about AI disruption when without being paralyzed by the noise. I hope you enjoy.

Speaker C: Yeah, fantastic to have you back, Lars. 2020 feels like yesterday. I don't know what it feels like to you.

Speaker A: No, it's uh, I think mathematically that makes it six years ago, but it

Speaker C: doesn't seem that way. Yes, doesn't seem that way would have

Speaker B: been one of our first episodes and I think this is number 70 now.

Speaker A: Oh yeah, that's crazy.

Speaker C: But you have been a busy man since uh, well we've seen you lots since then but since we last sat down and did a podcast because uh, uh, you on your second CEO role since the last time we sat down. Just update us on um, where you are now and we're talking to Lars today just before we do that, um, about all things SaaS and specifically private equity SaaS backed businesses. Um, but let's do the intro first. Just tell us where you are and

Speaker A: what my current activity is. Uh, threefold. So I'm sort of quasi plural, you know, in the sense I'm still a CEO. Uh, but I'm also chair of uh, two companies and all three companies are privately owned. And where I'm CEO is a company called BCom, uh, software and uh, software for compensation management, performance management and pay equity, that kind of stuff. And typically for very large company. And then where I'm chair is also software and also private XC owned. Uh, one is called clearcom, the other one is called Max Contact.

Speaker C: And uh, how many, what number CEO role is that for you?

Speaker A: Five, six. Yeah, so as with private equity ownership, it is six. So. Yeah. And actually early in my career it even included a couple of uh, hardware companies, believe it or not. So. And uh, and then I was in a payments company and then since then software and of course payments is also

Speaker C: software and just a bit more background. You have Danish.

Speaker A: Yeah, yeah.

Speaker C: Uh, Danish with a PhD. Aren't you a doctor?

Speaker A: Yeah, yes.

Speaker C: You forget about that now. Is it physics like I was coming back?

Speaker A: Yeah, it's um, so it's specifically fiber optics, but it's like quantum effects in light and that sort of stuff.

Speaker C: So Richard and I are going to

Speaker A: try and keep up actually. But that's many, many years ago. So I did that uh, in Copenhagen and uh, loved doing it. But it came a time where I wanted to get into business and my first, I normally say my first real job but um. And no offense to all of all the people who are in research because that's great. Uh, but my first sort uh, of commercial role was in New York and then since then I've not, I've never worked in Denmark outside of research.

Speaker C: Yeah, but you worked, you'd lived and worked in the States for some time, didn't you? Yeah, exactly. And you built businesses all over the world, but significantly in the US and Europe and now based in the UK. Have been based in the UK, yes.

Speaker A: Yeah. Here now for 16 years. Yeah, exactly.

Speaker C: But your businesses always have a big international element to them, don't they?

Speaker A: Yeah. So common to all of them is uh, I would say global or at least international and technology.

Speaker C: Yeah.

Speaker A: And private age owned of course. Uh, but within that it's been pretty much every time I've changed jobs, I've also changed the industry.

Speaker C: Yeah.

Speaker A: Uh, sites, you know, which is, you know, it's a uh, bit unusual but uh, if you run a company for three years, you get a point of diminishing return in terms of the value of what you learn. And you can learn a lot in three years about an uh, industry. Uh, um, and then it's a lot of pattern Recognition. A lot of it is the same. There's a lot more that's the same than this difference between industries, challenges and so on. A lot of the software companies, even just as a category of company, um, if you talk to another software CEO, the challenges we have are very similar.

Speaker C: Um, so what stimulated this um, is two things. Um, one really, since I don't know when it was released, this Global Intelligence crisis, um, Think tank article, uh, which is sort of doomsday piece really, which sort of went viral. Citrini researched it and it sort of went viral in February. Uh, I think I got a look at a bit towards end of February, beginning of March. Loved it because it's sort of fascinating scenario of uh, oh my God, is this what the world's gonna come come to? And is this what AI is really going to mean to us all? Uh, and then we caught up. You very kindly came to one of our CEO sessions as our guest speaker. And uh, in there we were talking about lots of things outside of this topic. We were talking about M and A and international growth and practicalities rather than thinking about how AI is going to eat our breakfast tomorrow. But you at the end sort of said, well actually I've got a pretty strong view in terms of, very strong view in terms of where this is going to go for the SAS industry. So yeah, perfect. That article and seeing Lars voice together, I thought, right, that's, that's a great, that's a great opportunity to do a podcast.

Speaker A: Yeah.

Speaker C: So, um, you know, that's, I guess that uh, the stock market um, crash really for listed SaaS businesses that came has come this year and to some extent some of them crash post. This article, uh, has led to investors, CEOs, any SaaS leadership team to be thinking very carefully about their business model. Where are you on this? Where are you in the extreme of what does SaaS look like in tomorrow's world?

Speaker A: I think there's so many different ways of thinking about it. Um, and when I talk to management, M teams or other CEOs where I'm at, um, I talk about three different things that can happen. So one is the singularity and uh, global abundance. And none of us are going to work and we are all going to sit on the beach and there's going to be an abundance of everything. So that's Elon Musk's way of talking about it. Um, there's a slight variant to that which is it's still the singularity, but we all going to ah, uh, not have a job and Nobody's going to have any money and consumption is going to dive and it's going to be horrible for the economy. Same, uh, trigger, which is full automation. Yeah. And then, then of course there's a third variant which is, uh, the robots are going to kill us all. And, uh, which is also coming from Moscow and which of course is building these robots. Well, it's probably going to have a slight negative impact on, uh, the economy as well.

Speaker C: Yeah.

Speaker A: If we're all going to die. But, uh, and then there is a fourth one which is it's going to lead to productivity gains. And uh, actually there was an article in the Financial Times that showed on a logarithmic scale, the growth in productivity year over year in the US and on a log scale ever since, like the beginning of time. It's pretty much a straight line. And then of course, Covid sort of down and up, you know, a bit of a squiggle, uh, and then, and also during wars and so on, uh, of course a bit of up and down. And um, that fourth view is that that's going to continue, but with a slope that's more steep. Um, which of course that's the only way to think about it. You know, when you run a business, you plan for the future. You cannot plan for we all going to die. You cannot plan for, you know, we can't feed our children beyond our control, but we can't plan for we all going to spend our time on the beach. I tend to ignore all those, uh, in my thinking because it's not productive to even think about that. So that's the first step. Now, looking at the stock market. So there are two parts to it. You have application software and then you have infrastructure software. And if you look at valuations since COVID times, what's really gone up has been the infrastructure stuff, the application stuff gone a little bit up and down. Uh, and what has primarily been hit right now is the application software, not the infrastructure. So if you sort of unpack it a little bit, it's a more nuanced.

Speaker C: Give me some examples of infrastructure.

Speaker A: So infrastructure, the aws, uh, you know, with the platform, the cloud platforms. And so I think of that as infrastructure, of course, Azure from Microsoft and Oracle, infrastructure, uh, and then application software like Salesforce or Hotspot or uh, SAP or whatever that sits on top of a cloud, I would, uh, consider that application. But then even looking at the application software, the market, uh, went up during COVID Then it started going down, it went up and then down again and now we saw the same level as it was back then. And there are people that I've talked to, uh, in the US in uh, sort of hedge funds that believe it has nothing to do with AI, it's just a market correction. Uh, and uh, I'm not, I, I don't think it's that extreme either. But AI clearly has an impact. Uh, but some of it is also a market correction.

Speaker C: A correction from what Sort of heady days of.

Speaker A: Exactly like, uh, theoretically, if you think about value, value is really about, um, how secure are, uh, future cash flows. And then you add all those up. That's your value, which of course makes sense. You know that that's how you calculate value and then how secure they are. Uh, that comes from a discount rate. And the discount rate is higher if risk is high. M And uh, of course the longer time takes, it's also if you forecast Something that's 10 years out, it's worth less than something you have in the bank account today. M because I think 10 years out has a risk associated with it. So that's sort of the fundamentals of how you think about, uh, valuation. Uh, and if that's your model, uh, and you looked at cash, uh, flows and so on of these companies, they were overvalued at that time. There was a bit of a bubble and it corrected itself. But right now I would actually argue that most of the application software companies, on the basis of future cash flows, they're actually undervalued, um, right now. Uh, and I know a lot of priority funds, they think of it that way as an opportunity to buy into the market.

Speaker C: Um, but, um, if you were buying as a private equity investor, what would you be looking for in the shape of SaaS assets today? What would tick the box for you rather than.

Speaker A: I think that there are five things that I would look for. The first three are what a lot of people have talked about. It's uh, companies that have, uh, systems of record, uh, which is a big fancy term for they have data that has value and it's proprietary. Uh, and it's companies that have deep domain expertise or they really understand the context. Uh, and it is, uh, domains where it is deterministic. Uh, and what I mean by that is, as an example with bcom, where I'm a CEO, if our software recommends a bonus or a performance assessment or salary increase, and it's an AI agent that actually came up with that recommendation, uh, what if that employee sues the company because he or she is unhappy about it and the Company is going to be like, well that's bad. But we don't actually really know where they came from. It was this AI agent that did it. Obviously you can't operate that way. So there are certain domains where it has to be explainable and has to be auditable. Mhm. Uh, uh, and clearly those domains cannot be disrupted in the same way. So I say those three I definitely agree with. But I think they're missing two that are much more important than those three. Uh, which is one about are you latched onto the market in the right way? You know, do you actually get ideas or do you not get ideas? And the ideas you get, are they good? And um, will those ideas, if you act on them, lead you down a path?

Speaker C: How do you get good ideas versus bad ideas?

Speaker A: A lot of companies look at their customers. Who are their customers? If they have a bunch of customers that are sort of forward thinking, chances are that they'll get better ideas. And SaaS software is really an accumulation of ideas. That's really what it is. Uh, and the value of getting on a SaaS platform is that you benefit from all the other customers and their ideas that have accumulated over the years. So of course if you are a company with a bunch of customers, uh, that are very forward thinking and some of the best companies in the world, chances are their ideas are good and chances are that that's going to flow into the software. So that's why that's so important that uh, I call it market poll, that you latch onto the market in the right way. So say it polls you to a good place and then the fifth one, um, you know, it's obvious, but it's about velocity, you know, how fast can you move? Which of course is really bad if you move in the wrong direction. But if you do move in the right direction, then of course you will want to move as fast as possible, um, in a very deliberate way. And for software companies, that has a lot to do with the tooling they have and which of course some of, or a lot of it is AI based. Uh, you call that software development? Life cycle is the process of developing software and for that you need tooling. And Claude, of course being an example of a new tool. Uh, but the best companies, they combine that with a few other bits and pieces. They do some agents themselves, uh, and then they uh, end up with better tooling than other software companies.

Speaker C: Would you put your company in that category?

Speaker A: I would venture that to be the case.

Speaker C: God, tell us what you've done there.

Speaker A: So we have our own software development life cycle. We think of it as, uh, it can code, it can test the code and of course look for bugs. Um, it can also performance test to see can it keep up with just activity. Uh, and it can do so called pen testing for penetration for security, where an agent is trying to be a hacker. And it can deploy and all of that with interconnected, uh, agents that are orchestrated across that whole chain. And then to start that chain, uh, you feed it with a very long document which you can call a business requirement document, or it's basically a spec for what it is you want to develop. But then we've also developed our own tool for putting this back together. So that's another. So it's not just that link of agents, we've also built another tool that actually creates the document itself.

Speaker C: Um, is that like a designer, all of that.

Speaker A: So what I tell the companies I'm involved in is that as a company now, to succeed, you need to treat your SDLC as a product just like any other product, take it as seriously as if it was a product, because that is so important for your velocity in the market. So, um, yeah, so that's sort of the fifth one. And there's a very important wrinkle to that, which is, um, you can relatively easily build something that is almost good enough, you know, as a very viable, credible prototype. But to make it fully good enough sometimes can be quite complicated. Because that last step, uh, from almost good enough to fully good enough is based on code that you didn't write yourself. You know, so maybe, I don't know, 100,000 lines of code that Claude wrote, not one of your engineers. And then to take that code to fully good enough, of course requires you to understand code somebody else wrote. So that can be a challenge. And then the other challenge is after you've then done that, then you need to, whatever that new functionality is, you need to fold it into the code base you already have. And of course the code base you already have has a database structure, it has a hierarchy and who has access to what information, who has what access rights, has workflows and has visibility in terms of who can see what data and all that. And so of course what you plug into it, it needs to comply with all that, you know, all those rules as well. And that can be quite complex as well. So, uh, I feel like a lot

Speaker B: of people are probably building stuff and they just don't know how it works. It works, but they don't know how it works.

Speaker A: That's the first bit, the prototyping is you can basically sit with a customer and you can listen to them, um, take input and then within three to five minutes you have prototypes and then you can ask them is this what you meant? And then they say well yes, but actually I would like this to maybe be moved a little bit and whatever it may be and then you can spin up another one. So I think that's a big, big um, value from AI is that, that customer interaction, it can be a lot more um, reactive, uh, you know, with a, with a good feedback loop. But, but also it also has the ability for you to um, have each of your customers actually have different software. Where that's almost been a cardinal sin in software up until like a year ago. You really want everyone to have the same because it's a lot more efficient that way. But I think now with AI actually that may not be the case anymore. And actually that, so it takes me to another point. So the promise of SaaS, um, for some companies stated explicitly and for most other companies, sort of an implicit promise in terms of the value you get from it, uh, has been I talk about the accumulation of ideas, that software is really an accumulation of ideas and you want to get the best ideas from the best customers and so on. But the promise that a lot of SAS companies have talked about, they call it workflow optimization and some companies have called it workflow automation, but I don't think anybody has actually done that, you know. So I think that is an unfulfilled promise pretty much of all software companies in, in my opinion. Uh, and it's just follow following simple logic in the sense that if you, if you look at software it's, it's standardization of how you do something and um, there's a workflow built into it and so on. So it's really a standardized workflow. Mhm. And if it's standard and they all have the same, how on earth can it be optimized? Ah, logically speaking there's just no way it can be optimized. To find an optimum requires you to vary, can be fully standardized. Uh, and if, let's say that uh, optimized means that you can do something quicker or if optimized means that you can make better decisions, whatever your measure of it being optimized is, uh, there's no software today that really does that. Uh, but I think right now with

Speaker C: AI they wouldn't claim that though, would they? Most of the HubSpot wouldn't agree with that.

Speaker A: But look at it what it.

Speaker C: No, I understand what you're saying.

Speaker B: Yeah.

Speaker C: For us to really automate we've got to do the automation.

Speaker A: So the automation yes, but the optimization no. So in other words, so what I'm talking about just as an example, so CIM system, maybe I shouldn't call houseboat out but let's say just any CIM system because it goes for all of them. If it's optimized then that should mean that uh, as a customer of that software that my bookings per salesperson must increase otherwise it's not optimized. Is it just if you think about it that must be the value driver or the measure of whether it's optimized. Of course. So how do you do that? Well you learn, you have some intelligence across how do you manage a pipeline. You learn when to qualified deals out which deals to take through, what are the signals of a deal that make it a good deal. So you only work on that and that way if you focus on the deals that you're going to win then you can do more of those and then you book more and everybody's happier. So it's that sort of thing that's not in any of the CIM systems. No, uh, you know, so that's an excite. But I think with AI we're going to be there, we're going to get there. So I think actually AI will finally allow SaaS to fulfill his promise which

Speaker C: is conscious of it. Are we going to get to a position where actually like at the moment we're so conscious of HubSpot, we're so conscious of our, yeah our CRM, uh ah, our SaaS systems. But are we just, are we going towards a world that we're just going to be interfacing with our, whatever our hard tech is our laptop, talking to it, saying this is what I want.

Speaker A: Yeah.

Speaker C: And it will be, it will be pulling from our Ah, SaaS applications but we're not actually consciously working with hosts?

Speaker A: I think for some part I would say yes, but there's so many things you can't do that way. Um, and uh, of course an example can be let's say well let's talk about BCom not to bring up my company too much but then let's do a second plan. So it's not so much exactly not so much about bcom but the domain we in that domain it's really decision intelligence. So what the software does, it has in it workflows and so on that allows the users to make the right decisions that hopefully, uh, have the best positive impact on the companies that these people that use the software, um, are trying to get to. Ah, his decision. Decision to who? Like for instance, some people talk about pay for performance, which of course is, you know, you see performance and then you pay accordingly. Which, yeah, that makes sense. It's sort of a fairness principle in the sense that if somebody performs then that person should be paid more than somebody who is not. And that of course makes a lot of sense. Uh, but actually I think the reverse thing is more interesting in the sense that um, what if you actually get performance from pay? So in other words, you pay people a certain way and then from that, because you paid them or compensated them a certain way, their performance actually increases. So I think that's a much more interesting correlation between the two. And uh, to do that you cannot, you know, talk your way to that sort of situation because it's managers making decisions and, and decisions about who gets what increase, who gets what bonus, who gets what performance rating and so on and so forth.

Speaker C: Is this how you think we overcome the challenge with the SaaS pricing model twice per person?

Speaker A: Yes, yes, it's an interesting thing because right now I think AI is a tax. So what I mean by that is that most software, uh, initially the first thing they implemented in terms of SaaS or sorry, AI functionality was different things that just did what the software already could do, but just in a different way and maybe a nicer way. So for instance, uh, example is inside of the software there's maybe uh, a search thing where you can ask IT questions, you know, in terms of how to do certain things that didn't mean the software couldn't do it, do that thing, uh, but it was of a helper in terms of how to do it is one example of early functionality that's based on AI. Uh, and of course it's useful, but most companies that put that in there, they didn't increase their price, price is the same and now they have an increased cost because people are actually using that bit. And that's why I call attacks. So that was the first wave of AI functionality. And then second way was AI agents and so on that companies could actually charge for. Uh, that sort of provided for incremental revenue. And of course with them the challenge very quickly became the cost of the tokens. Uh, and most companies think about it in one of two ways. Either make it usage based, so you pay for the tokens you consume, or make it outcome based so you pay for whatever value you get out of it. Uh, and I think if you Google and so on, read different articles, uh, it's almost like outcome based is the holy grail. Uh, but all the customers I've talked to, they don't like it.

Speaker C: Uh, how would the price be set on an outcome?

Speaker A: That's part of the issue is that how do you budget for that? If you have an annual budget, you commit to using something, that thing has an outcome, you don't know what that outcome is yet because you haven't used it. So how do you budget for it? Uh, so a lot of the customers I've talked to, they don't necessarily like that. Um, but of course the same applies to usage base, but you can buy it in chunks. So in other words you buy a bag of tokens or you buy a bag of whatever the measure of the usage is and then you can budget for that and then hopefully you're not going to run out, you know, throughout the year.

Speaker B: Yeah, but then I guess there's also an issue there with how do you communicate how many tokens you're going to be using for a certain activity.

Speaker A: That's exactly. And there are some tools that can help you with that such that uh, they actually give you an assessment of how much the data center. But uh, it's a very difficult question definitely. So I think as often is the case the answer is probably a combination. So in other words something that's a bit uh, seed but almost like quasi seed based, uh, plus usage, uh, based. And between the two, um, what do

Speaker C: you think the KPIs ah, will be

Speaker A: for SaaS businesses and moving in terms of valuation or.

Speaker C: Well, yeah, I mean you've run businesses on the basis of like retention rates, recurring revenues, price per seat, number of seats.

Speaker A: Yes, yeah, yes. I think right now some of the key, I guess valuation metrics or performance metrics of SaaS companies up until a year ago, you know, six months ago have been of course the rule of 40 that talks about the growth rate plus EBITDA margin still important. I don't think that's ever going to go away. Uh then right after that then retention rates, gross retention and net retention. Gross retention is basically how much do you retain from your customers, uh, but not including upselling additional things you sell to your existing customers and then net retention is where you include the new stuff you may sell to your existing customers. So I think all of those are still going to be there. Uh, but I do believe because it already is easier to develop product, I think the product set each of the customers will have, will be more dynamic. Uh, so right now, like in the old SaaS world of 12 months ago, um, most companies would have 1, 2, 3, maybe 4 products from one company and then they would sit on those 1, 2, 3 or 4, you know, for 3 to 10 years or 12 years or whatever without much change. If the company is now the supplier can release two new products every quarter and some of those new products are going to do whatever the existing products do, maybe better or in a different way. Uh, I think one key metric is going to be, uh, sort of customer net retention. So in other words, for each customer, how does the revenue develop and how much dynamism is there in the product set? Uh, and I think that's going to increase, uh, over time. And of course which also means that there's going to be um, already is, uh, but increasingly so, uh, value migration away from your ability to do something to your ability to come up with the best idea of what to do. Uh, and actually a lot of companies will be struggling uh, with do they have enough product marketing? Because if you're going to release all this, how on earth can you do that if you don't have enough product marketing and so on. So I think that the optimized organization today, even today, uh, is very different from what it was eight months ago. So it's already changed, uh, and increasingly so it will change. So combination roles will start to combine. So two different roles will merge into one role, uh, where one person can do both by using an AI agent.

Speaker C: Give us an example of that.

Speaker A: I think a very obvious example is between the BDR role and pre sales. So in other words, if you have a pre salesperson with good tooling, or bdr, um, person with good tooling, chances are that that person can do both jobs, um, and actually maybe do it better because they become broader set in what they do. And that's why. And actually I think there's too much focus on replacing people rather than augmenting people. And let's face it, it's a much simpler problem to augment a person than is to replace a person. Uh, because what a person does is a lot more than just the tasks. It's also the ideas they come up with. It's also the impact and have on the rest of the team and so on and so forth.

Speaker C: Um, what about new roles, New position?

Speaker A: There's one. I'm not gonna claim uh, any originality on this one because if you Google that. So on one role that people talk a lot about is the fde role or the forward deployed engineering role. Uh, and depending on who you talk to, it's defined a little bit uh, differently. But think of it as a role where somebody who understands the language of the domain that the customer is in goes to visit that customer, maybe brings that prototype tooling I talked about with him or her, sits in front of the customer and then spins up a solution right there in front of the customer and then uh, and then uses tooling. That's the same tooling as basically the SDLC I talked about before. Uh, and then whenever that session is done it can fall into the.

Speaker C: That's not a developer though, is it?

Speaker A: Probably. Well it's somebody who is. So that person does not, customer facing that person does not need to know how to code. Uh, but of course that same applies to a lot of the uh, developers. They also no longer need to know how to code.

Speaker C: Yeah, but that's going to require a skill that's almost like a consulting, it is skill set, advisory skill set with some process understanding, engineering application knowledge, the

Speaker A: deep domain and to have that dialogue back and forth with the customer. And of course that's with one customer. But uh, you know one of the challenges of the product management role of course is to make sense of input from several customers across a customer group. Uh, but if AI actually enables you to have different solutions within each customer, then that's no longer as needed.

Speaker C: Is that the job that rescues SAS from the scenario of a CIO CTO in a large enterprise business saying, you know, this sort of CRM software, we've got HubSpot, why don't we just develop that ourselves?

Speaker A: So I think, so I think short answer is probably yes, but, but uh, I think companies, software companies will be faced with three different risks as it pertains to AI. So the first one is upstart, uh, company AI, native company vibe code, something that blows you out of the market. Yeah, and I think for most software companies the risk of that is quite low. Uh, and it has to do with the domain knowledge and how complicated it is and so on. But there are some solutions like uh, I don't know, there's some communication tools without mentioning actual company names that are quite simple and you can actually do it and, and build a pretty decent tool yourself. So I'd say that's one risk, but probably quite low.

Speaker C: Then second, on that basis there are going to be quite a lot of businesses. There will be some platforms that are in that bucket. Yes, I mean we've probably subscribed to two where we're sitting there. Yeah, not HubSpot, by the way. But yeah, ah, we've not made where we think we might be able to build something like that.

Speaker A: Yeah, but, but of course the best for those companies who are so unfortunate to be in such a category of company, the answer is the same. So it's the five things I talked about before. Move fast, add to it, do other things, just move with pace and then it becomes a lot harder to do. So I think it's the same imperative for everybody in terms of what to do. But anyway, that's sort of the first risk and then the second risk is, um, that um, you end up in a situation where your customers, it's a CIO thing you talked about, uh, where they decide to do it. And I think that will happen if

Speaker C: they're doing their jobs properly.

Speaker A: They probably have to look at it, it will happen. Um, but I think it's also going to not happen in a lot of cases. And I think the uh, fundamental question is who owns the problem? If you are a CIO and of a 300,000 employee company, do you want to own the platform that you wipe coded with your team? So one thing is to do it, but you also have to maintain it and so on. Is that what you want to do or not? No support desk, there's nobody to blame, stops with you, and so on. I think the psychology of the situation, if I was theory, I would say absolutely not, I would not want to do that.

Speaker C: That's not what we're in business for, is it?

Speaker A: But that's it. And some of course will choose to do it. But that's also where this, uh, SaaS being an accumulation of the best ideas goes a little bit against that. Because whomever then does that, is that going to be the best ideas that they come up with? And two years from that, are they going to look at that and say, well actually no, we should have done it completely differently and so on, and then they have to maintain it all time. And of course as a SaaS company, when you have a customer, you land a customer constantly, they come to you and say, hey, can you add this? Can you change that? And so on, because they get ideas all the time. And then as a vendor you sometimes react to those, other times you don't, uh, but you accumulate at least some of those ideas and then everybody benefits from it. Ah, I think another sort of analogy that probably will cause some people to pause is open source with Linux, um, as an example, as an operating system, did Linux lead To the death of, uh, Windows. I think we all know the answer to that question. That did not happen. And why is that? Because nobody wanted to own the problem. So now it's on for free. Was not a better option than something you actually had to pay for because they didn't want to own the problem. Now, where it actually became something that people use was to get a bit red hat because then they could own the problem, and then it was no longer for free, you know, So I think that's, uh, I think that analogy carries over to a lot of software. Yeah. So that's all. So I'd say that risk is maybe low to medium. There will be some that are going to do that. What to do about that risk is exactly the same. Move fast, you know, have your customer, uh, that's it. Move fast, sell more. You make sure that you are dynamic and so on. And then the third one, which I think actually that risk is high. Well, that's your existing competitors. They move faster than you. And that one is real. And that one, that's the one we should worry about. Um, which again, has the same conclusion as the five things I talked about. It all boils down to those five things.

Speaker B: Do you think you talk about moving quickly in urgency. Do you think the market is doing that? Are they feeling that?

Speaker A: I think everybody wants to, but Most of the CEOs I talk to, they don't know how to.

Speaker B: Okay.

Speaker C: Um, and why not?

Speaker A: Sometimes for different reasons. Uh, there's a lot of resistance in the system in the sense that, of course, uh, if you're in a software company and the job of being a software engineer fundamentally has changed, which. Absolutely, it has. That's pretty scary. Uh, as an engineer. So do you then embrace the new way of doing things, or do you just put up a block? And there are definitely cases where people put up a block. Uh, so that's one obstacle. Another obstacle is just a way of thinking. You just mentioned the fd or we talked about the fde. It's a different way of thinking to have that sort of dialogue with a customer than somebody who sits on their own and writes software and so on. So it's different mindset. Um, good news is that they both have to be very smart, you know, so at least that carries over, you know, from one way of operating to the other. Uh, but I can see why some engineers are nervous about that, you know, so. And that, and that goes even for leaders of engineering teams, not just the engineers themselves. So sometimes that makes change difficult, uh, to implement. Uh, and then a lot of CEOs may not know enough about themselves. They know you have to do it, but they can really sort of articulate it. And then there's a lot where if you haven't seen it done, sort of really done, uh, then either you don't believe or you think you can't do it. It uh. So um, it's a bit like you know, back with cloud or the Internet. And how long time did it take for Microsoft to react? They knew they had to react. Why didn't they react? It's probably because of the things I just talked about, you know and they probably had some of the smartest people on earth, you know, inside of Microsoft. But sometimes you just can't move, um, because of those kind of things.

Speaker C: Yeah, but in those, in, in those scenarios it was almost sort of safer to wait and let yeah time play its course and see. Sometimes you should react, but it doesn't feel like. No, but sometimes this is a bit more cataclysmic if you don't, if you don't move fast enough.

Speaker A: I guess we will find out. But it's interesting. Uh, but, but I think that's a very, very good comment because I think there's so many cases where the fastest end up being the losers. You know the ones that, or you can also say the best, the ones that are really on the ball and they move super, super fast. Uh, and then they get into trouble. So a good example of that is uh, like with the Internet all the way back, the so called sea legs in the US I don't know if you know that term but it's think of them as local gone over our heads that one. Well it's think of them as local, local operators, uh, communication operators and they popped up all over the US and uh, because there was this bonanza. Fiber, uh, optics. Yeah. Sort of during the Internet, you know, the emergence of the Internet and of course all sorts of investment capital flowed into that. And uh, if you just called it a sea leg you could attract all sorts of money. So it sounds a bit like you know, AI. Yeah. And uh, the ones that within they uh, went in deep, installed a lot of infrastructure. Sounds a lot like AI. And uh, in terms of fiber optic cable and communication equipment and so on, they ended up with all sorts of debt and then there was a valuation reset in the market and then a lot of them went bankrupt. And the ones that went bankrupt were the ones who moved the fastest. So I think there's a lot of analogies. Um Here. But, but when it comes to the, the fundamental difference with AI in the two fundamental differences, if you compare it with cloud or mobile or the Internet or some of the other sort of big technology shifts is uh, although there's lots of similarities but I think that the two big differences is that AI is not just about a technology. It's not just about doing something in a particular context in a new way. It's where work is being done by non humans. It's a very big difference with the agents and so on doing actual work. And then I think the other very, very fundamental difference is um, with the tooling and so on where you can just move so much faster than you could back at the time of cloud and mobile and all that stuff.

Speaker C: I just wonder if we could talk about the dynamics of investing and selling businesses in the SaaS space at the moment. So as a CEO of uh, private equity backed SaaS businesses, do you look at your exit options now? Have your exit options changed? I suppose is the question on the basis of this. Probably needs to go to trade now versus another private equity round. Has private equity in the short term lost appetite to be investing in?

Speaker A: Yeah, so I think if you look. So I think short answer is no. Uh, but their bar has raised in terms of what they'll buy. Uh, and I think you see the same Sam from your network and the people you talk to. The deals that happen now for the most part are actually even higher valued than back in the heydays of 2021, 2022, uh, is. But those are the very best companies. And then the not very best companies, not necessarily bad companies, but just the not very best companies, they don't sell. And then for those that don't sell, uh, they carry on, you know, and then hopefully they grow and hopefully they improve their margins. And then by doing that and just holding on, then there's going to be a return on those companies as well. It's just going to take a longer time. And I think that's the way most funds think about it. And then of course that's for companies they already have. Then there's companies they can buy. Uh, and of course for companies they can buy, if there has been a valuation reset, then that's actually not bad news, it's good news in the sense you can then in principle, yeah, you can in principle buy into something at a lower point and that's not a bad thing. Uh, but I do think that the challenge of being private equity uh, has changed at a fundamental level in the sense that it's not just about buying and selling and financial engineering. It's also very much about value creation, um, and more so than it was in the past.

Speaker C: Um, and that's not just SaaS, that's everywhere.

Speaker A: No, that's all companies and so on. So they need to get closer to their companies. And uh, the ones that will do the best are the ones who not just know how to buy, know how to sell and do financial engineering, but also how to improve the performance, uh, of the company. And of course that's always been the case, but I think increasingly so now m than it was, let's say two years ago.

Speaker B: How are PE going about doing due diligence around AI? When they're buying a business, like obviously a new space, they might not be that clued in on like, what does that process.

Speaker A: So there's, you know, naturally there's not been a lot of deals that's happened, but, uh, but it's pretty clear what people talk about. There's a lot of two sides to it. You look for all the bad news and then look for the good news. Yeah. So the bad stuff is how, how protected is that company from disruption? Which again gets back to the five things I talked about before, you know, which is both good and bad news. The first three think of that as to protect you. And the last two is the good news. So they'll look for those five things and they'll look for the companies that will know how to not just automate a workload, but also optimize a workflow. Uh, so I think that's probably the change. But, um, uh, I think a lot of funds, they actually think that now is a, is a good time to do everything apart from the financial side, um, in terms of the, um, debt side of things, you know, where that of course has, with interest rates and access to capital and so on has worsened. Uh, so how much they can finance deals and so on is less than it was before, which of course affects their returns. So even if they could buy at a good price and sell at a much, much higher price and improve performance and so on between those two, uh, it may still affect their returns because they can't finance it to the same degree. Uh, but that funds are created very differently in terms of how much leverage that they've operated with. And some funds have never really done it that much ever. So for them it's not a big change. And the ones that have done it a lot, they're going to have to look at their model and so on change the model. Model a bit, uh, to accommodate for that. So.

Speaker C: And management capabilities going to be highly scrutinized around.

Speaker A: Yeah, yeah.

Speaker C: And, and I mean, I mean capability around AI. So I mean, and that I'm, I'm making this comment on the basis of. Across sectors.

Speaker A: Yes, yes. Yeah.

Speaker C: Not, not specifically SaaS, but I, you know, but from what you said earlier in regards to like the CEOs that can adapt to change.

Speaker A: Yeah.

Speaker C: Versus those that can't. And that applies to, you know, CEOs running.

Speaker A: Yes. Yeah.

Speaker C: I suppose the sort of next sector is techni services and then it's professional services and then you get all the way down to blue collar and engineering at the end, at the end of the scale. But CEOs need to be able to demonstrate how they are adopting, implementing and adapting their businesses around AI for a, a efficiency, um, margin and then value creation gain.

Speaker A: Yes. So I think. No, I definitely agree with that. And I normally use the analogy of um, a train on tracks. You know, in the sense that if you, if you're a company in a stable market, which we're definitely not in right now, but in the olden days, 12, uh, months ago, you know, uh, if you're a company in a relatively stable market, then it was about getting up on the right tracks, in other words, moving in the right direction and then move as fast as possible in that particular direction. Um, and then stay on those tracks until you sell the company. That is no longer the case. Because I think following on from that analogy, I think there's a lot more sort of ongoing off roading that's required. Uh, and uh, in the sense that you may be on tracks, but how do you know they're the right tracks? In other words, how do you know you're moving in the right direction? You don't know that unless you go a little bit to the left and to the right and you have a look around both to the left and to the right of the tracks you're on. So I think for CEOs and leadership teams and so on, you look for somebody who can deal with change, who can think, who can be almost like designing systems. And I guess at a level a company is a system, a market is a system, an economy is a system, and so on. And somebody who can think about that in a creative way I think is going to be needed. Um, uh, but you also need somebody who can drive your train. So after you've found the traction, you want to move down that, in that direction as fast as possible. So you need both, uh, but you probably need the off roading more now than you did 12 months ago, um, which is a different profile, um, but with the same component pieces but just a bit more you know, emphasis on the off roading than what it was before.

Speaker C: What are the other things that you think are gonna, are gonna be coming down the pipe? I mean I, I suppose I'll give, I'll give you your idea that you just said to me back before, before we started recording is the IPOs that are coming.

Speaker A: Oh yeah, yeah, yes.

Speaker C: Uh, and the impact that's going to have in a negative or a positive way and maybe some of the other sort of changes you might.

Speaker A: Yes, I think with that the, you know it's pretty obvious that you know, with anthropic OpenAI XAI SpaceX, you know, sort of related, um, from that point of view with those IPOs that are coming up, if they're very, very successful then they'll probably just carry on what they're doing. Uh, and uh, which includes price levels in terms of the tooling for software development and so on. It may go up but not necessarily in a dramatic way. Uh, on the other hand if those IPOs are uh, not necessarily a failure but a bit disappointing then they're going to have to get money from somewhere and I think that money is going to have to come from software companies. So in other words we're going to have to pay for the failure of those IPOs if they end up being a failure. So I think what we should also be looking at out for is the difference in how good the models are. And if they're about the same then AI, uh may turn into telecom in which case it's like a commodity. Doesn't mean it's cheap, it just means that it's cost plus. So the logic behind the pricing is cost plus. If one company runs with the market then it's more value based pricing in which case we're going to have to pay up uh, as customers.

Speaker C: But coming back to all the way back to the beginning of this piece that was put out there by Citrini, the global intelligence crisis. One of um, their first predictions was that leverage in SaaS, um, assets was going to become uh, unpayable. There was going to be a credit crunch like Crisis because these SaaS businesses were going to be out of business and the debt, uh, the debt becomes bad debt. I mean that just that from talking to you for 45 minutes that just feels completely extreme.

Speaker A: Actually I think it's a bit I'm sure there's going to be some of it, and I think we're seeing some of it already. Um, but at the same time, that's a whole lot of money sitting with the private activities, you know, more than almost ever. Uh, and why is that? It's because they're hesitating a little bit in terms of buying companies. Uh, they buy the very best, as we talked about before. Uh, and so a lot of companies that would have sold, you know, three years ago, they can still sell, but not right now.

Speaker C: They'll hold on to them for longer.

Speaker A: They'll hold on to, on them for a little bit longer and they eventually they're probably going to sell. Yeah. Uh, but we, we shouldn't forget that there's just a whole lot of money there. But the thing is they don't obviously don't want to spend that money on cash issues with their existing assets. Uh, if they're, what is that for? Well, but the money is there.

Speaker C: Yeah.

Speaker A: Um, so for them, would they, would they let the companies go bankrupt and lose all their money or would they rather dip into those funds? And I think it's pretty obvious what the answer to that question is. So it's a bit of a power battle between the private credit and, uh, private funds in terms of when does that thing happen and so on. And the power has moved a little bit from the funds to the private credit people and they're getting a little bit tougher in that whole relationship and so on. Um, but I don't think that's going to lead to SAS apocalypse. I think Sasmageddon was another one. Yes. I don't, I, I would be very surprised that that is going to be the result because that means that all these paradigm are going to take all those losses and are, uh, they really going to want to do that?

Speaker C: Yeah.

Speaker A: So unless, yes, there are some really bad companies, you know, where they're better off just to let them go. But it's going to have to be quite extreme here for that to be a better option than for them to just keep sort that problem and move on.

Speaker C: Last question for me is just continuing on that sort of prediction theme. Think about where we were, okay. Two and a half years ago. That feels like decades in terms of AI and um, its capability. Think about where we were. Uh, just little old pep talks last summer. We, we said, Richard and I sort of said to each other, let's start building some agents.

Speaker A: Yeah, yeah, yeah.

Speaker C: And we have mostly Richard, not mostly. And team entirely Richard and his Team, not me.

Speaker A: Yeah, you showed some at the conference.

Speaker B: Yeah.

Speaker A: Ah, yeah, Stanley, yeah.

Speaker B: The AI chatbot.

Speaker A: Yeah, yeah.

Speaker B: Which you can access if you remember,

Speaker C: by the way, you've got to get on the platform. Stanley can give you Lars's answers.

Speaker A: Yeah, yeah, exactly.

Speaker C: But yeah, it's been incredible really. What we are the art of the possible today, by comparison to just six, 10 months ago, do you think that's going to continue?

Speaker A: So a short answer is I don't think anybody really knows. But if I was to guess, then I think there will uh, be a point of diminishing returns. And the reason why I'm saying that is that most of the underlying architecture, if you open up the hood and say of that AI thing, uh, so there was one big idea years ago that sort of kicked it off, uh, and that one big idea was, um, prior to that one big idea, uh, the logic was to build something that emulate the human brain in terms of the neurons inside of your brain and how they talk to other each each other. And that's still the structure of AI. Uh, but up until this breakthrough that I talked about, they actually didn't know how to calculate the factors that connect the neurons. And it sounds weird, but that actually was the case, uh, which meant that the ability to train was not there because they didn't know how to calculate that. Then they figured that out, how to calculate the factors and as such they knew how to train. So that was a big breakthrough. Uh, but then there have been a number of other things in terms of caching and how they look at data and you know, and so on and so forth and where you take some, almost like a feedback loop and so there's been some other, let's call them, tricks that they've added to it. But still that same fundamental idea, which is we're trying to emulate a human brain which is a, ah, just a bunch of neurons that are connected. Yeah. So unless they change that underlying basic idea, I think there must come a time where it's going to be almost like diminishing returns intelligence.

Speaker C: Isn't that AGI?

Speaker A: Well, that's where a lot of people, uh, they don't think we're ever going to get to AGI with the existing technology. So uh, and actually I'm probably in that camp, you know. But then there are some new, there's something called world models, uh, which is completely different principle than the LLMs. And that's something where like a former uh, meta researcher, uh, called Yann Lecun French, who just Moved back from, to France from the US and he just raised, I don't know, 2 billion or something like that to, to pursue that idea, uh, which is. It's AI, but it's a different AI principle. Whether that is going to be better, I have no idea. But at least it's different from the LLMs. And I think something different is going to be needed for us to get sort of all the way. That would be my guess.

Speaker C: But, uh, America and China in a battle to win the AGI.

Speaker A: No, no, clearly race and it's not

Speaker C: going to work well.

Speaker A: And for the most part what they're doing is they're just adding more, uh, to the same. Yeah. You know, basically more gpu, but it's the same underlying fundamental principle. Uh, and, um, that's where I think it's, it's interesting, you know, in terms of how's that going to work out with that team in France or there are actually also some teams ending UK that are looking at some, some other principles. And so it's kind of. I don't know if I was to guess it's going to be one of the other ones.

Speaker C: Right.

Speaker A: That's going to take us all the way. Uh, but it's, it's really a guess because nobody knows. Of course, nobody knows.

Speaker C: Well, thank you LS for coming back

Speaker A: to talk to us.

Speaker C: Love the conversation.

Speaker A: Yeah.

Speaker B: Yes.

Speaker C: Some of it went over our heads a little bit. Really interesting.

Speaker A: We hung on.

Speaker B: We'll have you back in six more

Speaker A: years and we can see what it's like.

Speaker C: I think it'll be like six months.

Speaker A: That's a scary thought if you think about it. Six years from now, what is everything going to look like at that time? Yeah, check. I'm still employed. Well, either we're going to be on the beach or dead or out there protesting. Yeah, yeah, yeah.

Speaker C: Great to see you. Thank you.

Speaker A: Well, thanks so much for having me. Um,