Recall Sessions: Most SaaS Companies Won't Survive This - Jake Saper (Emergence)

23:21Pivotal. And Rob was the CEO and founder, just had this incredible ability to recruit just the most crack engineers and figure out how to do pair programming and everything else. Rob started a company during COVID called Mechanical Orchard. And what Mechanical Orchard does is it is an AI native service that does mainframe modernization. So they go to really big enterprises, Fortune 500, 100 companies, where the mission-critical workloads or workflows are still written in COBOL and housed on an IBM mainframe. Which is nuts. Which is nuts. 75 % of Fortune 500 companies are still running their mission-critical applications in a mainframe.

23:55We think that the cloud is dominant because CRM, ERP, whatever. The reality is the easy things were the ones that got to the cloud. The inventory management system that you built in 1995 that powers everything because you're the retailer, you are so scared to move that into the cloud. It is so hard. It has millions of lines of code. The people who wrote it are retired or dead. You've got to figure out a way to get it in the cloud. Historically, you've hired consulting firms who have spent years and tens of millions of dollars trying 50 % success rate. AI does coding really well. Now, the foundational models don't have all the access to the COBOL data that they would need to be able to do this fully out of the box, because by definition, that COBOL data, it's on mainframes.

24:34It's not cloud data they can scrape. So Rob's insight was, I will use the foundation models and pair it with the data I'm gathering from working directly with these COBOL mainframes to basically build a cursor for COBOL, for lack of a better term, but to sell it as a service. So he's not selling a software product that says, hey, Target, you figure out how to do this yourself. It is, no, I'm going to do the entire service for you. And what is the service? The service is moving your application into the cloud. So rewriting it. Is that a one-time thing? How do you price and package that? Yeah, that's a great question.

25:05like it's recurring, because that sounds like a one-time event. What's interesting, and this has surprised me a lot, is that once you get into some of these organizations, it's not just one application that is in the mainframe. You do it, you spend six months moving their core application, then you realize, oh, this other one was tied to it, and so now I'm doing this other one, and I'm doing this other one. So, so far, we have found that when you tap these big organizations, the work just continues, because there's so much code that's been written over the years in this language. Now, over time, there's all sorts of software that you can sell to support that.

25:33So observability tooling is really important. So you are the author of their core software application, and you generally also write some sort of observability tool to understand, like, how is it performing, what's the uptime, etc. That's a product you can sell as a recurring revenue basis thing. Anyway, to answer your question about how I got to AI-native software, I made this investment, and I started to realize that this concept of selling a service and not selling the software was more than just one company, that this was the new business model. It emerges, like the history of the firm is we have aspired to be early experts in emerging business models.

26:08When mainframes, when on-prem went to the cloud, the emerging business model that got created was SaaS. We were the first experts to understand how to build those companies. I think we're in a similar moment right now, where AI is a crazy, groundbreaking technology shift, and there are emerging business models that are being enabled by this for the first time, and no one knows how to build them. We are in this foggy phase where it's like, what is this new business model? Is it even a thing? And if it is a thing, how do you build it? And so what I and we have chosen to do is to say, AI Native Services is going to be the next big emerging business model, and we will be the emerging experts to understand and teach the industry how to build these companies.

26:50Because the way you build the software, the way you think about going to market, the way you hire for the team, the way you price, all these things look very different than software. Let's use, like, when I think of services, I'll start with IT, like consulting firms. Like, are you saying that there's a world where there's a new type of consulting firm? Yeah, but the consulting firms own the outcome. So it's not just a, like, I'm going to tell you how to do something. It's like, I'm going to do it for you. So I think the more compelling way to think about it is accounting firms will become AI-native accounting firms.

27:20Law firms become AI-native law firms. Insurance brokerages become AI-native insurance brokerages. claims processing companies become insurance, AI native claims processors, customs brokers become AI native. There are trillions and trillions of dollars just in the U.S. economy spent on these kind of professional services types of activities that are staffed entirely by people and not by technology. And the reality is there's an opportunity to create an entire new industry that is AI wrapped around by humans who provide the throat to choke and the expertise to deliver the service. I mean, historically, putting the prior iteration of this prior to the AI kind of world, they would have said, oh, services-based businesses are not venture outcomes.

28:03Yeah, they saw. What is the counter to that right now? The biggest reason why a services business was not venture-backed before was margins. By definition, it grows linearly. The reason why tech has been such an attractive asset class for so many years is that it grows nonlinearly with cost. So if you're building Facebook and you've got a bunch of users that start to use it, and you get all the advertising dollars that come with it, your cost to support that is not scaling linearly with the advertising dollars. So these businesses look really, really good. Historically in services, it's been linear.

28:35You close a new deal, you need to hire three more people to deliver on that project. So yes, revenue goes up, but margins don't go up. So you have these lower gross margin businesses, 10%, 20%, 30 % at the max. It is now possible, if you have AI that performs a lot of the core services with the human layer on top, to deliver these services at 50 % and 60 % gross margins. And what's interesting is that the addressable market for these companies is much larger than software ever was. Because they are not selling a tool, they're selling the outcome. When you look at this, there's a lot of ways to go about it.

29:04One is to say, I could go build technology and deliver it to the services firm and say, hey, I'm going to create some efficiency in me. There's some unique business model. The other is, I will become the thing. I know you have some investments that are like this that you can talk about. What are the different approaches to building it? When you say AI Native Services, is it the first? Is it the latter? Or is there something completely different? It's definitely not the first. The first is a software vendor. You're selling software to a service provider. Maybe there's a business model where you're like, hey, I'll get a percent of the outcome, and I'll service you.

29:38And you're kind of like the distribution arm of it. But this is kind of happening in healthcare historically. It's like, hey, there's all these billers that are out there that are trying to recover some money. And you could sell to them, or you could become the biller. I think that there will be hybrid models. So I work with a company called Prosper AI, which is in the healthcare billing. It's really in the healthcare benefits verification and prior authorization space. Historically, this is a job that has been done either by BPOs, or it's been done by the hospitals and clinics that call the insurance carrier and say, hey, can Shomrod get his x-ray?

30:15Is that approved? Whatever. It's incredibly human. It's crazy. I've listened to a lot of these conversations. There's a number of humans that sit on a line and exchange information back and forth to say, yes, you can get the x-ray. AI obviously can do this way better. This company started with an AI voice agent. Now it has a bunch of other agents that do this. And what's interesting is they started by selling to the BPO's because the BPO's needed this work. Well, imagine this also because of distribution. They're like, hey, I mean, unless the founder had a bunch of relationships. So in his case, yes.

30:45And he got some of the largest BPO's to start, which gave him huge credibility in the whole industry and particularly with the direct customers. So now he's got that business model. And then he also has a direct business model where he's selling into clinics that were too small or weren't working with BPO's yet. and he's just going direct and saying, hey, look, if I can do it for these guys, I can definitely do it for you. And he has this kind of partnership model with some BPOs where they jointly go to some customers and say, we're going to jointly market our offering to both of you. So to your point, there are going to be examples of this that are kind of hybrid-y.

31:17The thing to keep in mind is that an AI native service business is owning the outcome. And that is the core insight or the core North Star, I should say, about these businesses. So if you are a software vendor that is selling software to a service provider, but you are not also capable of owning the outcome that the service provider provides, you are not an AI-native service business. You are a software company. If you're just selling a technology and saying, hey, you can use it. You can use it to help you do accounting better. If you aren't capable of doing the accounting yourself, you are not an AI-native services business.

31:47But the cool thing is there are now a bunch of companies that are doing the accounting themselves. I work with a company called Hanover Park, which is a fund administrator. Fund administration, for the non-VC listeners out there... I definitely know a lot about fund admins. Every time I talk to someone who's an investor, that's their reaction. Everyone knows how painful this is. For those folks who are not investors, fund administration is a vertical accounting firm. It's a vertical accounting firm focused on private equity and venture capital funds. The way we do accounting is very bespoke, and there are some legacy services firms that do this.

32:20They are not well-loved. They don't have great margins, but they're very large businesses, because it's a really big industry. Hanover Park decided to do this from the ground up, and they did something to start with that I thought was really bold. They built their own custom-made ERP just for this use case. They built the core system of record themselves, but they chose not to sell that ERP to existing fund admin. That would be a software company. They said, I'm going to take this ERP I built, and I will become the fund admin. I have the unique ability to deliver better fund administration because I literally built the core tool that I need the agents to come out of.

32:55Whereas even if you took an existing fund admin company and said, I'm going to make it all AI native, but if the core system they're working on is some legacy SaaS system that you have to figure out how to get your agents to interact with, it's never going to be as good. So this company built the technology and then is doing the services on top of it, and so they're able to do it way faster. Do you believe that all services markets will have an AI native venture backable opportunity? or do you feel like some market? I mean, healthcare is so big. I mean, we can, you know, the number of people. There'll be a bunch in healthcare.

33:27There'll be a bunch in healthcare. Like, you know, but like, are all markets considered as an opportunity from a venture perspective? Because I struggle with this. Even in healthcare, someone might say, oh, I'm going to be going after a specific sub, you know, specialty area. And it's like, I'm like, is this sizable enough to become an independent company? Or like, you know, how do you think about that? Well, let's compare it to vertical SaaS, right? Because the argument would be the same. In vertical SaaS, was vertical SaaS ever big enough to build large companies? And the reality is, when we invested in Viva in the mid-20, 2008, 2009, the addressable market for a CRM in pharmaceuticals was$400 million.

34:03If you read our investment memo, it was like, that's the market size. Right, wow. So that was kind of a crazy bet. The thesis we had was that the market would expand over time, and that the founder, Peter, would find ways to sell more stuff into that market. I don't know what the current market cap is, but it's probably$35 to$40 billion. Obviously, the market size grew a lot, and Peter figured out a way to lay it with the cake. We've learned the lesson from vertical SaaS that if you are selling something that is very valuable to an industry that has a lot of money, and you're a great leader, you can find a way to expand it.

34:32The difference between vertical SaaS and AI Native Services is that the dollar budget you're going after is much larger in AI Native Services, because you're going after the labor spend. Mm-hmm. Like, let's take pharma, for example. If you were to build an AI-native pharma sales agency, there's way more money spent by the sales reps going out there trying to sell these drugs than there is on the technology underpinning that. And I don't know if it's even possible to build an AI-native pharma sales rep thing. I mean, that's not possible, but the size of the budget is actually much larger for labor.

35:03So, I'm actually even more bullish that there can be venture-backable, niche-y outcomes in AINS, AI Native Services, than I am in vertical SaaS. What do you think about... So I meet founders that are pitching this in two different ways. One is build an independent brand. I want to acquire customers. In the case of the example of going direct and so forth, others say, hey, I want to go and acquire. I want to go do this PE. What do you think about these different strategies? Is there natural opportunities to lean towards one or the other? Do you have a perspective of whether one is better than the other?

35:35I do. In fact, that's why we named it AI Native Services, because we were trying to distinguish it from roll-ups, AI roll-ups. The work the word native there is doing is that the company was started in the AI era to be an AI Native Services vendor, not an AI... Well, you can still start an AI Native Services company, but the way you go about it is to acquire... I don't think that's an AI Native Services company, because you're acquiring a legacy services business. That's legacy. But you can... This is the point. The you know thing that everyone's saying, there's a lot of work being done by the you you know thing.

36:06If you buy five services businesses that have been around since 1985, and you try to slam them together, and you have all these legacy people that were definitely not AI native people, and you say, okay, now, everyone, use AI. Do the thing, but do it with AI. Do you think that's really going to work? Well, okay, just to play, I mean, not that I've done this before, but, like, what has been positioned is, you know, a lot of these folks have existing client bases. They're experts in their workflow. We can kind of study the workflow. Then we're going to build the software around that workflow. And then we'll figure out what's our true need of capacity there.

36:43We will downsize. And that's how we'll do it again. Yeah. So it is definitely true that there is an advantage from a go-to-market perspective by buying an existing book of business. That's obviously true. And there may be an advantage on the data side, although I think that's overblown. Because I think people assume that these legacy services businesses have lots of data and it's structured in a way that you can make use of by building an AI tool. And the reality is most of them probably don't. So I think that there might be something there, but I think in many cases it's overblown. The go-to-market motion or buying customers, I think there's some value in it, but I would argue that should be step two, not step one.

37:19Step one is build an AI product. Find a way to, without any legacy, gorp-y stuff that you have to keep managing, find a way to build it truly from the ground up. Build your ERP for fund administration, whatever service you choose. Get some customers organically. And the reality is, I've seen this now firsthand, if you are selling an existing service and you're selling it either faster, better, and or cheaper, demand is not the constraint. Like to use the Hanover Park example, he has so much demand that in Q4, he stopped selling completely. He said, I'm not going to take a single new customer because the challenge in this business is actually delivering and deploying.

37:53And I just had a call with him an hour ago and he's threatening to do the same in Q2. The challenge with these businesses isn't actually, like people think it's go to market. I think for the right CEOs and you choose the right market, you know there's existing pull because there's a massive services entry for it. So it's not a question of like, if I build this, will people like it? It's like, if I'm selling something either faster, better, or cheaper than my competitors, people are going to buy your thing. The challenge, the biggest challenge in AI native services business is being an AI native service.

38:19The biggest risk in these businesses is being a service, right? The biggest risk, and this is why I think buying, doing the roll-up thing from day one is really risky. Because by definition, you are starting as a service and you're hoping to over time migrate to become AI native. I think it's a much higher likelihood for success if you start with an AI tool and you figure out a way to build that to be something that can deliver a lot of the value with AI. And over time, as you do more work and you build the systems in a way that there's a closed loop, more and more of that work will be done by AI.

38:50And then you can figure out ways to scale go to market from there. I think organic is working really well in a lot of the businesses I work with. I'm seeing these partnerships model work really well. There's this really interesting frenemy dynamic for a lot of these businesses where the legacy service providers are like, shit, I need some help. And so they're willing to do these pretty favorable deals with these AI native service providers to say, we'll go to market jointly and we'll split the revenue 50-50 and whatever. And then you get a lot of this go-to-market benefit without the headache of the acquisition stuff.

39:19Now, this isn't to say that over time there's not a role for, okay, I've gotten to some meaningful scale where I've solidified my culture, I've solidified my product. Now I can do a tuck-in acquisition to get exposure to the Northeast market or to the Canadian market or whatever. That, I think, makes sense. I think what I don't fully understand is the idea of trying to do that de novo. I'm always curious about is the sort of tension between outsourcing a service to a services firm that might have a better business model to serve the client versus saying some of that value is going to go back in-house.

39:48There's a lot of examples of this accounting. You can say, oh, well, people have worked with accounting firms at the mid-market. Well, if all of a sudden you have these capabilities, why can't you just not work with a services firm and do your own accounting? Using the same AI, and maybe there's a company that's selling directly to the client. How do you determine whether there is a real market shift towards, are all services-based businesses going to survive in the next 10 years? It's a great question. I think that the businesses that are most likely to survive in the services realm are those that have really high trust need.

40:27So if you are performing a service that has some regulatory oversight, certainly any sort of legal oversight, if there's some sort of sensitive healthcare angle to it, anything where it's like, if you mess this up, it's really, really bad, it's probably something that you're going to want to throw to choke. You're going to want a third party who's an expert who stays on top of the technology and is willing to take on that, as I said, that kind of insurance-like responsibility. On the flip side, if it's a service that maybe everyone doesn't die if it fails kind of thing, it may be the case that you just build your own AI to do it.

40:59I think about marketing agencies, for example. And there will probably be some marketing agencies that survive and thrive in the new era, but I think a lot of people are realizing, oh, actually, I don't need to outsource that anymore. I can just use cloud. Are you saying like, you know, publicists, which is like, you know, tens of billions of dollars and Omnicom and, you know, WPP, these are like massive agencies. Do you think of some of those, some of those, there's gonna be a shift of rather than paying an agency to like bring it in house? I think the role, and I'm not an expert in this space, but I think a lot of the role those folks play is like ads.

41:30There's like a network element to it. So I think they'll probably survive because there's still this kind of like, I'm in the middle of the ad flow thing. I'm thinking more like the web design agencies. Like if you're a web design agency and like you've made a lot of money that way, like you should use Bolt or you should use one of these companies and just do it for you. Well, let's talk about a category that I know you have deeper familiarity with, like law, legal, right? You know, I've been wondering about this because, you know, there's one model which is like the ultimate value. Who has the budget comes from the client, the general counsel, legal teams that are paying the outsourced legal firms.

42:02You know, there's a bunch of them. And I'm like, well, isn't most of those dollars going to be wanting to come back in-house and just saying, hey, we have a smaller team? I kind of wonder whether law firms in general, and there's a lot of them which should even exist. Sure. And yet there are AI native legal firms that are getting created. There's more and more. Yeah. I don't think it's an either or thing. I think it is true that GCs will look to realize, like, oh, I can do more of the stuff I used to outsource myself internally with tools. And I think that there's still a role for the external law firm.

42:38Back to the same point I made before, which is, like, part of the reason you hire a law firm is a CYA tool. It's like I had someone who was, you know, past the bar say this is blessed and it's kosher and so we can do it. So I think there's still going to be a role for those folks. But I'm on the board at Ironclad. We sell to general counsels. and we had a record Q4 and outperformed last year. So the business is doing well, I think in part because GCs are realizing I think tech spend by GCs is going to continue to increase. Yeah, I mean, I have not made a lot of investments in legal because I don't know where the puck's going to go.

43:15But I do think there are categories like accounting and healthcare. I think it's hard right now for VCs to figure out what makes this team kind of unique. Because they all kind of like, how do you determine, there's five companies coming to you and they all kind of say, I'm going to create a new AI native XYZ services business. How do you distinguish between, is this the one? Because they're all saying the same thing. I think that - They all sound great. In an AI native service business, I would argue that having some domain expertise in the team is even more important than a software company.

43:50Because in a software company, you're selling a product. In Ains business, you're selling yourself. You are a service. You are saying, I'm going to do this for you. You're not selling them some tool. And so you, whoever the you is, you yourself, the founder, or you, the team that you're representing, becomes incredibly important. And so one easy way to do it is to say, do you have DNA that's relevant to the service? And that's relevant both in terms of figuring out how to build a product, although sometimes having too much legacy DNA can actually hinder you from building the new innovative product.

44:22But it's more important from a go-to-market perspective because your customer is going to want to say, is going to want to see like, oh, yes, this is a startup, but like they have this woman who like ran fund admin for this thing for a long time. So like I trust them because like I know her, I know her brand or where she came from, etc. So some of the best AI Native Services founders come from the space. I mean, Rob is a great example. Like Rob is like the king of the nerds in terms of engineer hiring. He was a pretty logical person to start this business. But in the case of Chris, who runs Hanover Park, he was not a fund admin.

44:52But what he did, I mean, he went, I was skeptical when I first met him, because I was like, how do you know about this space, which is so bespoke and random? And he was like, I've had 115 conversations with CFOs of venture firms and private equity firms. Ask me any question you possibly could about the way fund admin gets done. And he just became the savant in it. And then he hired a bunch of people from legacy firms that all have recognizable brands. And so when he went to customers, he was like, look, it's AI, but I have these people who you know and trust are the human rapper on top of it. So I think to answer your question in one way, understand what the team's DNA is.

45:27And I think that you can't just be some hotshot kid out of Stanford without domain expertise to sell the service, because ultimately the buyer doesn't care how smart you are. The buyer cares, like, do you have relevance and why trust you? And so imagine there's a listener out there that's building an AI Native Services business. What does traction look like? They raise a pre-seed round and they want to prove a story to the world. What do they have to show to you as an example to be like, they got the right team and now they're executing? So it's totally different than software. In software, if you raise a seed or pre-seed and you go ahead and raise in Series A, what the Series A investor is looking for is, are you growing quickly?

46:08Is there an incredible pull for what you're selling? And do you have good customer retention? Now, you could argue that in the past few years, the Series A investor hasn't been able to ascertain in a second because founders are raising so quickly, so you don't know if customers are going to renew. But let's just assume a world where customers are renewing. Generally speaking, in SaaS, that's product market fit. If you're growing quickly, your customers are renewing, that's a good profit for PR. It's even better. Exactly. That is not the case in Ains. In AI-native services, I think one of the biggest risks is something I'm calling Mirage product market fit.

46:36And that's a world where you are growing quickly. and your customers are renewing, you've got good NDR, but you don't actually have product market fit because the service you're providing is not being done by AI. It is being done by people. And in that case, you are just a worse services business. It's taken a worse financing model. And so I think that's the head fake that's happening or will happen. I still think this AI native services thing is still quite nascent. While maybe it seems sort of buzzy because I and some others have been talking about it more recently, it is still definitely the minority of companies that are being started this way.

47:07This is still, we're at the very, very early stages. We're still trying to figure this stuff out. I wrote this thing called the AI Native Services Playbook in part to track the way this is growing over time and be a resource for founders. So I'm still obviously learning myself. But I think that there's this Mirage Product Market Fit thing. I think people are going to try to raise, and they're going to show good growth numbers. And when you double click and see how much of the service is actually being provided by the AI versus by the humans, you're going to see that it's still majority human. MARK MANDELSKI But is it, but are we, I mean, is there a metric that you're looking for?

47:34MARK MANDELSKI Yeah, great question. MARK MANDELSKI Because in some ways I'm like, there's AI everywhere. But is it really meaningful to say, hey, we're actually moving costs or improving top line? And it takes time to show this. I mean, I tell founders they've got to show it. And I'm like, well, I mean, that's not easy to go do. For sure. So a bunch of thoughts. So let's start with metrics, and then I'll talk about strategies to actually achieve it. So on metrics, there are input metrics and there are output metrics to understand if an AI native services business is working. Input metrics, the most important thing you need to do as an AI native services CEO is figure out what your product North Star is.

48:10How do you know that the product you have built, the AI platform you have built, is delivering real value and the value is compounding over time? And the reality is that metric is going to look different based upon the service that you're actually providing. There's no universalizable metric for that. In the case of Ryan at Crosby, which is an AI native legal firm, he tracks this thing called Hurt, or human review time. and it's how long does it take for a human to review one of their contracts. And then he pairs it with some quality metric to make sure that it's above a certain bar. And over time, if hurt is going down, then theoretically that should be a good input metric.

48:43But every one of these AI Native Services businesses needs to figure out, what is the North Star product metric that is the most important thing for me to track? So that's the input. And the output to know if it's working, there's kind of intermediate outputs and there's one most important output. So the intermediate output are things like revenue per employee. How efficient are you as a business, which is something that becomes really important in a services business. So revenue per employee in a legacy services business is, by definition, going to look much worse than revenue per employee in an AI native services business.

49:13It may be the case that as you start out, your metrics don't look that good there because you still have a lot of people as the AI is getting smarter, et cetera. But as you start to go through your growth curve, you should see leverage there. That's something that you and the VC should be tracking. And then, of course, the most important output metric is gross margin. So are you at a 20 % gross margin? Are you at a 60 % gross margin? It's a completely different story there. I do think that there's a lot of fudging going on in the gross margin calculation right now. And I don't think it's necessarily ill-intented.

49:42I think it's just that founders don't know. We're still learning. We're still brand new in this. But if you're calculating COGS as an AI-native service, you need to include the human labor required to deliver the service. That's obvious, but some people are putting an R &D in. I'm like, no, no, no. This needs to be there. It's COGS. Obviously, the spend, the figuring out how to allocate token spend. And this gets a little tricky because some of the token spend is R &D, like if you're actually developing the core platform. But some of the token spend where you're actually using the platform to deliver the service, that's got to be COGS.

50:13And so if you have an honest look on what is COGS and it's approaching 50%, 60%, then you probably have a pretty good AI Native Services business, and I would like to talk to you. The thing that I wonder, though, is folks that are raising rounds today, Like, there's this, like, I'll raise a pre-see and I'll raise, like, a million or two. It feels like it'll take a long time to prove what you're just describing. Yeah. Maybe, and maybe in, like... Especially the gross margin thing. Like, let's put that out, because, like, that's, you know, you can calculate it and it just looks really bad in the beginning.

50:47So here's my preference, which is, I think, an unpopular thing to say as a VC. I would rather see a business, an Ains business I invest in, grow more slowly, but really perfect the AI product such that those leading and lagging indicators are showing up and to the right, than grow super quickly and figure out the AI later. I think the grow super quickly and figure out the AI later thing is a... We had some of these issues in previous tech cycles where it's like, I'm going to sell$1.50 and it's going to work. I believe, because I've now seen it in a bunch of the business I'm involved with, that an AI-native services business should not struggle for go-to-market.

51:27Ultimately, if you're selling something that has existing demand and you're selling your customers. Well, this is an existing market. People are already buying these services. Exactly. And that's the thing people don't understand. In tech, in software, you invent some new thing to solve some new problem. You still don't know if people actually want to buy this. And so you've got all these things you've got to prove on Product Market Fit. If you're selling accounting, everyone needs it. Everyone's going to buy it. If you can go to me and say, I'll do your accounting, it'll be more accurate, it'll be cheaper, it'll be faster, and you can show me evidence that the six other people that look like me that you served had happened for, of course I'm going to buy you.

51:56So the issue is, a great AI native services business, the issue isn't go to market. As I said, like Chris at Hanover said to me a couple hours ago, he's going to pause sales again. The issue is making sure the AI is good enough to deliver the service in a high quality way. Do you think this is sort of universally true? because I feel like, you know, it's like once you're on the show, you know, I've gotten two quarters to a million, like, you know, I've got to keep on showing the thing, you know. I've had a frank conversation with one of the more high-profile founders in this space about this very issue because once you get on the venture drug and, you know, your venture investors are saying, like, you've got to keep growing, you've got to keep growing, you're going to feel the pressure to do it.

52:33My hope, I mean, A, if you're working with me, I'm not going to do that. That's refreshing. B, my hope is that we're still in the early learning phases of this business model, and VCs aren't stupid. So if they do this with three companies and they see that they all blow up or they all get to a place where they have terminal values and aren't that attracted because they're traded on low EBITDA multiples because it's low gross margin, they're going to learn the lesson and say, okay, I'm actually going to do it the AI native version next time. I'm not going to try to force the growth over the AI native thing.

53:02It's going to take some time. This is why we're still in this foggy era for this business model. But I think those of us who are like real experts in or who are dedicating ourselves to become experts understand this and we'll work with founders to help them understand as well. The thing that I, you know, the humanity aspect and all of this is like we're really investing in like kind of this destruction of human labor across literally, you know, dozens of industries, if not hundreds. Like, this is millions of people that are, like, you know, going, you know, day in and day out doing something that can be done through AI.

53:39Accounting is probably the one that is seeing a lot of disruption here. Healthcare, obviously, you know, you mentioned the phone call, picking up the phone call, doing very basic things. There are millions of people working on this. Like, what is our responsibility in this, like, day and age? I know this is something that I know you think about, like, and I'm a little bit, like, nervous about it. But even though as a venture person thinks about financial, like where is there opportunities of value creation, I'm excited about it. This is the most important thing facing humanity. I feel so passionate about figuring out a way to help society make this transition in a way that isn't horrible.

54:16And I do not think that we as an industry are paying enough attention to it. Dario, to his credit, has been talking about this as have others. But I think most of us are kind of going along with, we have the economic incentive to do this. And I think we need to continue to follow this economic incentive, because that is the way this model works. And also, with great power comes great responsibility. We have to help figure out ways to make society shift in a way that is going to survive in this era. I don't have any answers to this, but I'm going to spend my time on it. So, next weekend, I'm going to spend four days at a summit focused on this with other leaders in tech and finance and business.

54:58We're all going to New Mexico. And the idea is like, we're just going to try, we're breaking into small groups and we're going to come up with like ideas to solve. How do you retrain? How do you like, we're going to try to come up with some ideas and then see what we can do in the real world with them. I have no idea if this is going to work, right? It's probably not. But I think all of us have a responsibility to be investing our time and money in helping society make this shift. I mean, you know, it's, have we seen evidence of this quite yet though? Because in some ways, if you look at any sort of technology shift over the last century, there's always been this fear of like...

55:35Someone gave an example of NASCAR and how the first time they built new technology for the NASCAR, there were like two people kind of in the pit working on the car, and now there's like 15 people. It's like clearly the technology almost created more jobs. Are we like... How much of this is just fodder versus actually being real? I think that it's back to my outcomes framing, which is, I think in this world, the technology that we are building can do so much more of the outcome. And to use the NASCAR example, like, that's not applicable there, right? I was trying to come up with something else here.

56:12And you could think about the Industrial Revolution, which might be the best analogy here. And I think it's not well reported enough, but there was this concept and there was a group of people called the Luddites. This is where the term Luddite comes from, that there was a big bloody revolution where people were really upset with the rise of these industrialization things because people were at work. Now, the reality is the industrialization only did a little bit more of the outcome than what happened before. And so there were still a lot of jobs happening in the factory. I think what I fear in this world is that the AI is going to do a lot more of the outcome, which is going to mean that the need to do work goes down.

56:50Now, there's all sorts of positive utopic implications of this. I'm a musician and so like one of the things I think a lot about is like what could this unlock if I or we all had time to create more and do things that feel more human like are there lots of like positive things for our kids and everything else but I think like there's going to be a really rocky transition that we got to figure out and sometimes I feel like when I'm talking to people that aren't so AI pilled that I'm a crazy man from the future who's like guys a big thing is coming and like no one we're not taking this seriously enough but it's starting to happen like I just got a text from one of my college friends who is not in tech today.

57:26And the text was, he forwarded the Mythos announcement from Claude. And he was like, I think working feels stupid now. So I do think people are starting to realize the implications of all this is becoming quite real. I'm kind of curious whether the, I mean, obviously we think about the business model. We think about the economic leverage that can be created through AI and so forth. But, you know, when I think about, you know, my friends that are doctors and, you know, middle America, I don't know if they're thinking that way. I don't I mean, maybe they are. And there's real pressure to be like, all right, we can automate our front office person.

58:02I think they actually care about. I don't know. You know, I do believe there's real technology shift and it's going to happen. I just don't know if it's going to happen as fast as we think it is. Yeah, I think that's like that's the positive spin, because the reality is like if this happens more slowly, it's better in many ways. Right. Because the society can adjust more quickly or adjust more effectively if it takes less time. But that company, Prosper, I mentioned, is selling to a lot of Midwest doctors. And if there was someone in the office whose job was just to call insurance carriers all day long, and now you can do it at a tenth the cost and higher accuracy, you're probably going to make that shift.

58:35Because doctors are very rational people. Let's take self-driving, for example. We always thought this was going to take forever, and there's so much money to spend on it. And now I got here in a Waymo. I'll leave here in a Waymo. My five-year-old was in the Waymo with me the other day and completely nonplussed by the fact that there's no driver. And here's the even crazier thing. At the end of the Waymo ride, it says, like, don't forget your keys and wallet. Right. And she turns to me and goes, what are keys and wallet? She doesn't even know what those things are. Right, right. And Waymo, you know, is going to roll out, you know, across the country imminently.

59:09I invested in a company that spun out that is from a Waymo founder, our Waymo executive called Bedrock Robotics, which is attempting to do Waymo for construction. So self-driving excavators and self-driving dump trucks and such, which would increase the speed at which we're able to build our country way faster because you could build at night, you can build more safely, et cetera, but also has huge implications on employment. And so I think that the people that are funding this stuff, the people that are building this stuff, have a responsibility to figure out how do we help mitigate the impacts of what we're doing.

59:39For listeners that are out there that want to be part of this conversation, what is the starting point of this? Whether you're a founder that are like, I'm just going to go and disrupt work. That's who we talk to all day, or VCs that are listening. Well, if you're a founder, the best way to think about this is, what business can I start to help address the change? An example, it seems very likely that white-collar work will be disrupted before blue-collar work. If that's the case, there's going to be some period of time where the demand for figuring out how to become trained in blue-collar trades is going to be a lot higher.

1:00:14There's going to be a lot of people who used to know how to code who want to figure out how to be a plumber, which sounds kind of ridiculous, but actually probably not. In fact, I read an article in the FT last week that applications for plumbing schools in the UK are through the roof. So one business I think would be interesting to build and perhaps to even fund would be, what does an AI native trade school look like? I don't know. I don't know if it's possible to build an interesting business that would be VC-backable. But if you're a founder, to answer your question, the best way to address this massive disruption is to start a business to figure out how to help people make the change.

1:00:48Yeah, another version of that is find markets that have labor shortage to begin with. Yeah. Because there's probably opportunities to create more jobs. And construction actually is that. So there are just not enough people that operate these excavators today. And so for some period of time, actually, this will benefit the industry. But over time, obviously, there are going to be people who need to find new jobs. I'm really grateful for you to take the time and kind of talk through all this. I know that we're very early in this AI Native Services, but I know that you've been at the forefront of this going deep, deep into it.

1:01:17I appreciate the value around the playbooks and all the services. I know I definitely share it to all my founders that are working on this space. Maybe we can kind of close this out with some sort of rapid-fire questions that I – Let's do it. Deep tech, you mentioned you did Bedrock Robotics and AI Native Services. What do you think produces better venture outcomes over the next decade? I think that AI native services will be a better adjusted expected value situation. I think these businesses will be more likely to succeed by definition because they're less binary. Deep tech is generally pretty binary as a tech work or not.

1:01:52There's obviously go-to-market risk there as well. But my read is that deep tech will have bigger outliers, but AINS as a category will be larger. Most overrated AI investment theme right now? I think if you are selling an AI co-pilot to a knowledge worker, it's going to be a tough ride for the next few years. What's an example of this? Well, like, if you were... Because, like, ITSM is getting a lot of... ITSM is interesting. I mean, like, one thing, actually, on that topic, like, one thing I would be interested in funding would be... I'm interested in this, like, in the concept of an AI native MSP, which is in some ways doing some of the tasks of an ITSM for smaller organizations.

1:02:27There have been a few attempts at this, some roll-ups, but I haven't seen enough yet. I'd be curious to fund. But as an example, if you are a tool that helps a marketer draft a better copy, that feels like something that's going to be pressured over the coming years. What do you think is the most underrated? Obviously, A &M Services. That was a layup. That was a layup right there for you. What is a pre-LLM SaaS company that you think survives and thrives? And you can't mention FIN and Intercom because we already talked about it without naming a portfolio company. I mean, one I mentioned before that's not a portfolio company, but I think I would be super long if I had the ability to invest in it directly would be LinkedIn.

1:03:06I really believe in that network. I mean, maybe it's also I've ramped up my communication on LinkedIn over the past couple years and have found it actually surprisingly valuable. Some of the people I've met, some of the insights I've gleaned, some of the intros that have been made have been really helpful. Have they done a lot of AI stuff? No, but that's actually the point. That's actually what's interesting. That's crazy. I don't know if they need to because they have this trust network. You already answered this, but I'm going to ask you again. One metric every Ains founder should tattoo on their arm.

1:03:38There are three. The first is what is your North Star product metric that indicates that your product is getting better with AI? The second is revenue per employee. And the third is gross margin. True gross margin. And if you had to pick a vertical that produces the first$100 billion AI Native Services company, what would you pick? It's got to be a regulated vertical because I think those are going to have so much pull and value. We have already made three investments in insurance for AI Native Services in different parts of the stack. And so I'm hoping it's going to be an insurance. What are the three verticals?

1:04:13So one is a BPO, one is a claims processing company, and one is an insurance broker. Okay. All right. So different aspects of the stack. Okay. So your insurance. Insurance would be my answer. Yeah. Given patterns. Yeah, given my hope. Who do you think wins the foundation model race? I don't know. I think obviously Anthropic is the darling at the moment, and it's possible that they're now so far ahead that they'll stay ahead. Or is there even a race in the sense, I mean, you look at AWS, Google Cloud. Yeah. I mean, my bet on this is that the most underrated player in the foundation model ecosystem is just open source.

1:04:51I think that the open eyes and tropics Googles of the world will continue to lean in and be a few months ahead. But I think over time, we're going to realize that the vast, vast, vast majority of use cases for AI doesn't need the model that's three months ahead. You can use the model that's three months behind that's 80 % cheaper. And so as a result, we are super long on the open source model ecosystem. We did the Series A at Together AI, which is indexed on that ecosystem. And that company is now, I don't think I'm allowed to say, but it has got insane revenue growth. My wife is the president of Base 10, which is also very indexed to the rise of open source.

1:05:26And I've also seen how fast they've grown. So I think my answer is not exactly one of those companies. It's actually the open source ecosystem. Makes sense. Just for my education, is there a book or an essay that really shapes how you thought about the AINS kind of strategy? So the most obvious book is The Innovator's Dilemma, because that is basically what is happening to all these, both actually SaaS businesses as well as legacy service providers. The other book that I just finished, and I want to get the title right, is called To Rescue the American Spirit by Brett Beyer. And it is a book that came out recently that is a biography of Theodore Roosevelt, which is so fun because that guy is crazy and awesome.

1:06:07He obviously famously coined the concept of man in the arena. And the parallel I was thinking about, which was certainly not what Brett intended when he wrote this book, is that if you go from being a software vendor to an AI native service provider, you go from arming the man in the arena to being the man in the arena. You're actually delivering the outcome. And so in some ways, it's actually, it's probably a harder business to pull off because you have to both build product and do services and client service and everything else. But Teddy Roosevelt leaned into the heart. Love it. Love it. Last one, five years from now, what do you think you got most wrong about Ains today?

1:06:41I think what I will most have gotten wrong alludes to the question you asked before, which is, I think that there will be some services that exist today in our big industries that go away over time because the AI can just do all of it. AI Native Services is presupposed on the idea that there is a vendor that you want to have that delivers you the service. There will probably be some services where you don't need a vendor anymore and you can just have the AI do it directly. I believe that for the foreseeable future, there will be a number of industries where having a throat to choke, particularly in a regulated industry, having a third party validate something, there's still going to be value in that.

1:07:18But there will certainly be some services that we think about today that become in-house. Well, hey, Jake, I really appreciate you taking the time. This was a really fun conversation. We should probably do this in like six, nine months again. It's going to be totally different. Let's rehash what we just talked about and see if anything's changed. So appreciate you, Jake. Appreciate all your advice. And listeners out there, for folks that don't know Emergence, what would you like to say? We aspire to be the early experts in emerging business models. Business models like AI Native Services, physical AI with things like physical intelligence and bedrock, AI infrastructure like Together AI, agentic software.

1:07:51But in general, we just want to be early in helping people figure out how to build these businesses in this foggy era. I guess I would end by saying I feel more personally and intellectually energized in this era than I have in my 12 years of investing. because this isn't a function of like copy and paste. This is like, we're all figuring it out from scratch. It's super exciting. Well, absolutely love it. Thanks, Jake. Appreciate your time. Thanks, Matt.

1:08:17Hey, this has been Kaznoka, co-founder of Village Global. Thanks so much for tuning in to the Village Global podcast where we go deep on all of the biggest topics in tech. If you enjoyed this conversation, please subscribe to our YouTube channel. You can check us out on Spotify, Apple, wherever you get your podcasts. We'd love to see you for the next one.