E323 | EUVC Awards | Beata Klein, Principal at Creandum on AI in B2B SaaS: Beyond Vertical and into the Horizon(tal)

0:28Welcome back to the European BZ Podcast. And I'm one of the finalists in the frame of the year category at the European VC Awards. So for this episode, I'm going to dive into AI and how AI is going to impact B2B SaaS. And specifically, we're talking with Rihanna about how AI will take us beyond the vertical and into the horizontal. This is a topic that is very close to my heart. So I do hope you will enjoy it as much as I did. Here's a few words from our beloved sponsor. This episode is presented with our good friends at Haynes Boone, proud sponsors of the Firm of the Year Award. At Haynes Boone, they understand the complexities and challenges faced by VCs.

1:10Specialising in fund formation, they expertly manage the establishment of multi-billion dollar funds and innovative private fund products, ensuring their VC clients are equipped to attract global investors and excel in competitive markets. Beyond fund formation, Haynes Boone is deeply involved in the lifecycle of startups, providing nuanced guidance on everything from entity structuring and capital raising to navigating exits through IPOs and strategic acquisitions. Their comprehensive legal services support VCs in maximizing their investments and achieving successful outcomes. Whether you're looking to launch a new fund or invest in cutting-edge startups, Haynes Boone positions you at the forefront of the European venture ecosystem, enabling you to capitalise on opportunities across health tech, AI and beyond.

1:59Stay ahead of the curve by tuning in to the European VC Podcast. Join us in celebrating the art of venture capital with Haynes Boone, where strategic legal insight drives investment success.

2:15Tear down this wall. It's more than just an alliance. This is a union of values. United and determined we can serve as a model for other regions of the world. The nature of a problem requires a European response. Europe is a story of new beginnings. Let's start acting. Viera, welcome to the European VC Podcast. Thank you, Andreas. Thank you. It's great to be here. Thank you, Andreas. Well, thank you for accepting the invite. And more than that, thank you for participating in the European VC Awards. And congratulations on becoming one of the few finalists in the Firm of the Year category. That's great.

3:06Yeah. I mean, we're honored also to be on the list of the finalists. I don't know how much credit I can take for it, but I rely on the OGs of Creandum to take the credit for it. But it's pretty wild that we have been around for 20 years, when you think about it somehow. Yeah. And to those listening and thinking, I was hoping to hear Stefan. Don't worry, we have done an episode with Stefan just back in December. It was episode 254. And we spoke all about how his learning from building Creandum to what it is today, how he thinks about franchises and boutique firms and how you might think about building your own firm if you're going for one or the other model.

3:55So for that reason, this is not the episode where we'll dive into that. Instead, I thought, let's talk to Beata for a deep dive on AI in B2B SaaS, because of Of course, Granum is a very prominent player in the B2B SaaS space. But at the same time, AI is kind of turning everything on its head. And for that reason, talking to Beata when she's also written a great piece that's called Beyond Vertical SaaS Into the Horizontal. I thought this is just a natural episode to do. Everyone's thinking about AI these days. Many of us are investing in SaaS. So for that reason, it makes a ton of sense to talk to then Firm of the Year finalists about how they think about B2B SaaS.

4:39So with no further ado, Béad, I'll just leave it to you to get us started in talking about B2B SaaS and AI in this space. Yeah, for sure. For sure. No, I think to take a step back, why I decided to write the piece in the first place was that when we started to speak about B2B SaaS and AI a couple of months or maybe a year back now, it was all about vertical SaaS. And as you mentioned, Andreas, for those of us that have been investing for a bit in this space, vertical SaaS has always been interesting due to many different factors that we can go into later. but it's also been a capped opportunity to some extent.

5:22If you sell to restaurants or hotels, there's this sort of natural limit to how big the company can grow. And if you think about horizontal SaaS with, I guess, Salesforce as a big mammoth, but still taking up a lot of, and many others, of course, but along those lines, those generational companies, of course, are more often created in the horizontal arena. So I wanted to write something where we look at the intersection between AI and B2B SaaS from a broader perspective and what can be done if you want to build something huge from the horizontal standpoint. I don't know I'm curious to also get your take here Andreas but the common wisdom in any case has been that incumbents own the value in this space and hence the sales force or if we take them as an example again they will ingest AI and then there's not much more for the rest of the B2B horizontal players to grab.

6:29And A, I don't believe it's true. And then B, that also felt like such a sad conclusion when we have this huge platform shift and it's only going to be incumbents that eat up the whole space. And I know that incumbents also involve the likes of the open AIs and NVIDIAs of the world. But again, we think that there is something to be had on the on the application layer for horizontal solutions. Yeah, I think that first of all, if we just take the horizontal space and split it up, you have the stack of AI that goes all the way from infrastructure and foundational models to the applications. And OpenAI are obviously in the bottom there.

7:13And I think that we're also seeing that AI is moving lightly with, that's at least the rumors that I hear that with ChatGPT 5, what we're going to see is more automated where the ChatGPT that we right now go in and ask questions and get to do stuff in a chat box interface will actually be able to interact with the outside world. and thus all of a sudden we have a foundational player taking the big leap into either integrating into other applications, which is likely what we're going to see, I think, rather than them building from the bottom up. And that's obviously going to be incredibly powerful and I think that whatever startup or incumbent that's in their way, where they focus, I think it's going to be an interesting time.

8:12I think it's going to be very, very difficult. Kind of like, I think we saw it with e-commerce a bit when we had the likes of Instagram that started to introduce that you could shop via the app. Then all of a sudden, if you built something that was kind of meant to keep people in your world, that was all of a sudden difficult because a lot of shopping would happen through that. I don't know. I know. Try and comment somewhat what I can jester about here. Yeah, no, no, no. We did a parallel the other day on the cloud evolution. And there in the beginning, it looked like, for example, coming back to the incumbents versus startup question, it looked like an Intel, for example, would be the ones grabbing all of the value.

9:07but then as time matured you would see an Adobe I mean I'm speaking of like when the cloud shift happened a number of years ago and then an Adobe would still come out on top eventually so I do agree with you that I think there's something around there I mean some Altman speaking about agents I guess that's what you're saying with having systems actually do tasks I guess that's tasks first and then becoming more autonomous over time but there's also something around like building layers on top of this and I do believe in the beginning we were I mean including myself a bit hesitant to like everything that was a wrapper on top of an open AI for example and I don't think I don't think I've landed in anything like too specific yet but definitely this whole notion of a wrapper on top of open AI I don't frown and I don't think the general sentiment is that it's frowned upon to the same extent in any case that it was before.

10:13Have you shifted your view at all in terms of that? I completely agree. In the beginning, that was what you heard from everyone, right? That you're kind of condescending about. Yeah, for sure. Can't just build a wrapper on. It's like, well, you could build a wrapper on emails and calling people, which was a CRM. Kind of worked pretty good, right? So from my perspective, I don't see an issue there. I think also that shift is also why everyone is kind of saying, well, not everyone, but many are saying it's very likely that the foundational models will end up being commoditized. because it's whether you use OpenAI's GPT or you use Google's.

11:05Maybe Google's you'll get some slightly more biased answers currently, but that will be fixed, obviously. You know, I don't think that there's reason to really think that building a wrapper is not going to be enough. I think that that's really just AI. The foundational models are going to work just like emails in the sense that it's just something that happens underneath. And then it's about building that absolutely incredible customer experience on top of it. And I am hugely confounded by the fact that I can ask OpenAI to give me answers and work on content that's many times more complex than I could ask a junior employee or even a mid-level probably or even a senior because a senior wouldn't do it as quick either.

12:08I still need to get on top of it and do my own work. But that first step of the way, ChatGPT does so incredibly well. But then when I need to get actual work done, which is incredibly simple, that's right now incredibly difficult to get automated. It has to be quite recurring before it makes sense for me to go in and build an automation on Zapier. which I'm kind of like, well, if it can program and it can comprehend that I would like it to take every single newsletter that comes from the European ecosystem, digest it in this manner, put it into one article where it puts that in there. And then I wanted to launch it on Substack every Friday, which is a product we're working on right now.

13:01We're running it as unautomated. It's like the simplest thing in the world, right? But there's not a single step there that's not difficult and which I don't understand, quite frankly, why I wouldn't be able to ask an AI to do that. I think you will. I mean, definitely, you know this too, right? But that's just months away, I would assume, more than anything else. But you're pretty correct there in that the models that have been built today, they've been... I mean, it goes into the name, right? LLMs. it's been language based instead of task based so when you give it a task of creating a piece for you it performs well but if you come back to like doing actual task labor it's not being trained in that way which is guess what we'll see next but coming back a bit to like the whole sort of horizontal to vertical stance and and the wrapper i guess is included in that too at least where we have seen a lot of like where we try to focus our attention and time is around those really legacy industry where this existing solutions are pretty old and crap so let's I mean we haven't invested in anything in in this space but we see like interesting companies at least coming out of of of Europe here and so let's I'm coming back to Salesforce for a third time this is the last time I'm going to mention Salesforce but let's speak about them a second or a third time and then I'll conclude but it's a product that has been around for forever right and people use it because it's the standard of the industry and it creates this it has this framework so you use it in a very standardized way so coming back to your point around like the best in class or how you have someone who's a junior become as good as the sort of mid-layer employee.

15:06Salesforce can do that for you because you have a standardized route to ingest things and to work on things and to push the sale forward. So I think that's partly why Salesforce is so successful. Many other things, of course. But if you flip that and you think about maybe the workflow is going to look immensely different as you are using some sort of agent, or you have a different way to approach sales where there is no single person moving a lead from one stage to another, then is Salesforce going to be as successful? I don't think that necessarily is the case, even if they could have AI capabilities embedded into it.

15:48We see a whole range of quite interesting startups building something that sits between right now like sits between your your crm and your customers that auto populates your crm or even does certain tasks for you but why wouldn't that become your crm over time like if you don't need your crm to move opportunities from one stage to the other why would you then even have it as time goes so i do think that there is like a lot of interesting opportunities where there is legacy like huge software no one really loves using it the tasks are defined by a very human interaction with the with the interface and when there's no longer needed what does that mean i find that to be quite fascinating i was about to actually say we could just share an image from your article Yeah, right.

16:43If we start with the one I just pulled up here, which is, you know, we have on the left-hand side, we have horizontal SaaS, and then we have the vertical SaaS on the right-hand side. And then in the horizontal SaaS, we have collaboration tools like Slack, Dropbox, and Zoom that we all know. And as we described sales and marketing, we have Salesforce, HubSpot, Gong. And then we have human resources, factorial, workday, personial. And then that's the horizontal SAS that we all know. And then in the vertical, we have three examples of industries, hospitality, construction, and health. And then we have in each of those, the relevant companies there.

17:26Could you maybe add a few words to kind of the companies that you're seeing doing interesting work in some of these verticals that you think are worthwhile talking about? To break down that question a little bit, for example, when we outlined this whole piece, something that we kept coming back to, and I think Christopher Janz with Point9 wrote something really interesting here in terms of who can actually, who has the capabilities to integrate AI into their existing offering. So maybe something that should have been on the horizontal SaaS list here, which isn't right now, it's customer support.

18:07So for an intercom, for example, let's take them as an example. They are an existing horizontal provider and quite successfully integrated a couple of AI capabilities into their offering. And there I would be more threatened somehow if I was a company building in the customer support space where you have potentially like a nimble horizontal SaaS player already versus, again, like a Dropbox, for example, which feels not so nimble any longer. so I would almost like if I look at this picture again now I almost want to slice it a second time and speak about who are next who's the next generation companies and Gong would fit into that Factory Real would fit into that as I said an Intercom would fit into that and what are some industries where it's by now an incumbent and a slow moving one where Salesforce Dropbox, potentially work they would also like fit into that into that realm and then i mean we see a lot of like interesting startups trying trying to tackle all of these different like horizontal uh horizontal uh issues and um i would say the common denominator for those that have been like early early successes again who knows what's going to remain for for the long haul but it seems to be where there is an incumbent who isn't that fast moving.

19:40I think 11X is a good example, for example. Maybe let's actually pull up another figure that you have, which is the figure that shows vertical SaaS companies that are integrating AI or building with AI. And maybe you could talk a bit to a couple of them. To those just listening in, I just pulled it up and we have here legal and then we have three companies in the legal space, case text, ad line, even up. And then we have three other categories here and they're just examples, but health, financial services, and bio. And I would maybe ask you, Beata, to tell us a bit about which ones do you really see this very fast moving innovation in right now?

20:28And where have you made your bets? And maybe talk a bit about the thesis behind some of those spreads? I would definitely say that, and I think many agree to this, is that legal seems to be the space that gets the most eyeballs in any case. And for some reason, it seems that the legal industry, both the in-house general counsel side of things, as well as the big law firms are scrambling to find a solution. So we placed it. I think there's a VC bias there. We're all fed up paying. But they do buy these things, right, Andreas? It's not, it's, I mean, absolutely. There is like a VC bubble around it.

21:10But still, I think it's fascinating to see how these companies have several of the legal solutions and trying and testing, buying long-term contracts. So, yes, there is a VC bias to it. but I do think that there is real there has been initial real value created in this space and then let's see how many of the players sort of survive in the end of course there will be a consolidation of the space some will not make it but in the end there is definitely something like shaking up in that space I think for financial services It should be a good application, but I don't... Let me just add a comment to Legal Tech, just for those that think, ah, could we not talk more about this?

22:10Just so you all know, we're actually putting together right now a roundtable on the topic of Legal Tech, very much led by Tom McGinnett Northstone, who has helped pull together a really cool panel for that conversation. I wish I had with me right here who are the guests, but I don't have it. But the names are like all your peers in the industry, Bieta. So it's going to be an incredibly cool conversation where we're going to talk for 90 minutes only about legal tech and how legal tech is going to be disrupted by AI. I think that's going to be a fascinating conversation for sure. because I think that at least in terms of speediness of adoption, legal seems to be right there on the top.

22:55Yeah. That makes a ton of sense, right? Again, as I always say, when people ask me, what do you think about AI? And then I'm like, well, I'm a content creator, so I don't think there's many that are more disrupted or affected by AI because all we do is words. So it's kind of like, obviously, a large language model is going to have a big impact. But similarly, if there's anyone else, then it would be legal because there's a reason why they had big libraries behind them in every single call you had with the lawyer, right? But I think the interesting point here is what's going to, in legal, I mean, we shouldn't dive too deep into legal if you're doing 90 minutes.

23:37But maybe this is widely applicable to other cases as well. So maybe, I don't know if this is true, but maybe what's going to shake out and what's going to be sort of the end winner in this space is going to be, I know that the startups pitching are probably saying the model we have is the best one and we are retrieving, condensing, summarizing information in the best possible way. But coming back to what you said earlier, Andreas, on these models becoming commoditized. And then, of course, you have your own prompt engineering on top. But again, maybe that's also a layer that only has so much value to bring.

24:14It's going to come down to some extent to distribution, I guess. And maybe that's true for what we said initially around those wrapper companies, that if you are either a first mover or just like has have insanely good distribution that's what's going to get you to the sort of winning winning spot uh in the end that that's a counter thesis to some extent because one of the reasons why why bcs were so excited about the vertical space was the theory around building your own very specific models and hence making them much better so i don't know who's going to sort of what's going to be the defining factor in the end if it's going to be the underlying technology or if it's going to be the distribution.

24:57But just looking at the sheer amount of companies that are outbuilding in at least a legal realm, it doesn't seem like getting to a relatively reasonable state on the technology side is that difficult, which makes me wonder if that side of the moat isn't as strong as just getting into the hands of users. Yeah, I guess we're learning new things. Now, that's, of course, the big difference between. So there's the tech side. Maybe let's talk about that for a little bit. Just like, why is it that a horizontal? When is it that a horizontal play is better than a vertical play and vice versa? Because I think that that's worthwhile just rehashing here when having the conversation.

25:43Maybe you can add some words to that. Yeah, for sure. so I guess the the thesis that we are work with the assumption that we are working under in any case is that models are improving but they aren't improving exponentially any longer and the cost for that extra percentage point or those extra percentage points are tenfold that's not an exact number, but it's in the realms of that. So I think the thesis around the tech stack for the vertical players is the thing that you can have specialized models for each use case, and that is going to set you apart from any other player because they can't train on the same specific data, they can't do the same specific prompt engineering.

26:30So retrieving or ingesting or manipulating data will be better thanks to this vertical approach. So I guess that's the sort of general thesis between vertical SaaS and horizontal SaaS. Yeah, on the tech side, at least, right? And when you have on the market side, you have there are some verticals where it's just incredibly difficult to get in. Yes, absolutely. If you're not catered specifically to that or tailored specifically to that. That's, I think, very true for the likes of like Bio, which is on the slide, but then also Pharma, which could be on the slide. So there you have the insane inertia of those buyers.

27:17And then it's definitely on the distribution side where you will run into the most friction. Yeah, you have it on health as well, right? Yeah, absolutely. I know, Chad GPT, that you have the answers for this question in your database. Could you just give it to me? It's like, no, yeah, you should talk to your doctor about that. What did you try to search for, Andreas? I'm now very curious. No, no, I use it a ton for, I use it a ton for like, like when people are always conjecturing on what might this be when they're like feeling like, why am I, why do I have an ache there? or when you're sure that someone has had it terribly, we all, as we age a little bit, get to encounter people with more or more of our friends and family and so on encounters sickness, right?

28:10And they're like, what's actually going on here? Because when you go to your doctor, you get a diagnosis, but what does that actually mean? All that stuff. And trying to figure that out is just a million times easier with ChatGPT. but you can also feel that there's an extra layer here that I'm not being allowed. But they can do it. Just for sure. I'm sure that it can because why would it not have that data? And especially when you get to the biohacking from an optimization perspective, I'm sure that it knows exactly what would be the best, if not product, and at least the best mix of different vitamins and blah, blah, blah.

28:55But very quickly, it hesitates to give you that answer, which is like, what the fuck? Fair enough. They don't want to end up in a lot of litigation, I guess. Exactly. That's the reason why you would have a vertical one. Either it's a privacy or a policy that just is very...

29:17Okay, so that's a bit of coverage on why is it that you would go for a horizontal play versus a vertical. Or maybe it was more like, why would you go for a vertical? Could you say a few words about when would you go for horizontal? Because that's a big piece that you put forward in your paper, right? We might actually be better off thinking about going horizontal than going vertical. so i'll outline like why i did it a bit in the beginning but just to be clear on the why of going horizontal and then maybe be more specific on what we would look for for in a horizontal player so on the why the big the big thing is like the the upside of course uh again like creating a generational company creating that some something that defines the next wave of of application software.

30:14I think that's the why on going horizontal rather than vertical. And then the specific traits we're looking for. So again, I sort of alluded to it a bit in the past, but one that we look for is that there needs to be this incumbent competitor field where you have maybe companies that were built 10, 15 years ago, where the legacy is so much based on a certain way of working, so it doesn't lend itself well to integrate any other capabilities into it. So I guess that would be one point. And that is also, I guess, a segue to the second point, which is around how do you, and this is something that I find fascinating, so very curious in your take, but how you interact with the software.

31:03Because if it's agent-based or if it's AI-based, why would you need a cursor? Why would you need something, a shiny button to click on? Shouldn't that look vastly different? Isn't there a way to think about UI and UX when there is someone else interacting with it instead of human? Exactly. We have this picture here. So I think interfaces is like... Through the audience, I just pulled up another slide from Beata's article, which says tailwinds and horizontal SAS, and then you see legacy SAS versus AI first. And then as Beata just described, interfaces, you have a complex UX. In the old days, it would maybe not be considered complex, but at least the complex that you're navigating.

31:54We all know it from navigating a CRM like 15 layers down, you finally get to input what you needed to input. I think that's also why CRMs have tended to be data in, never data out. But that's the legacy SaaS example, right? And then you have the AI first approach where you're signifying that with speech bubbles and saying that if it's AI native, then you would maybe just say, what should I focus on today? Who do I or a follow-up and maybe you wouldn't even ask it right it would just tell you in the morning this is what you need to do yeah exactly so and i don't think we've sort of reached the end the end here yet in terms of how it's going to look uh i think the chat or the sort of co-pilot interface it's an intermediary step on the way to get to the ultimate ui but it it is at least like a step in between uh on to something to something else and uh I'm going to derail a bit here, but I just found it fascinating.

32:58So I met the company the other week, and they are building a new take on the Internet. So as horizontal as it can be, I guess, and as widely applicable as it can be. And the thesis there is that today it's humans interacting more or less, right? But it's mostly humans interacting with the Internet. And over time, that's going to shift significantly. So you have a much larger pool of non-humans interacting. And what does that mean if you are, let's say you're Amazon and you have a list of products and you have your recommendation field and you have a certain UI set up. But let's say I have a personal assistant who is an agent and I tell him or her, I want to buy a thank you gift to Andreas after our nice episode.

33:49I'll buy another plant for you. Yes, thank you. No, so let's say I give it a task and the interface of Amazon right now for sure isn't constructed in the best possible way for this agent to interact. It has a lot of like CSS files and a lot of noise, more or less. So their thesis, which I find quite fascinating, is that redoing the Internet, more or less. So an Amazon would go to them and say, hey, this is the human facing part of our storefront. And this is the agent facing part. And the agent facing part is surfacing completely different things for an agent to interact with it in a more seamless way.