E664 | Mikael Johnsson, Oxx: AI Hype, Real Productivity & How Not to Lose the Plot

0:28Welcome back everyone to the European Easy Podcast. Quick note, if you're building or running a fund, you know it takes the right partners. At EUVC, we only work with sponsors we truly believe should be part of your tech stack. Please do take a moment to hear about them. And if you do, reach out, mention EUVC. It's the best way you can support what we do. Thank you so much. Starting off, HSBC Innovation Banking. If you're a founder, a scale-up or a VC, you need a bank that actually understands your world. HSBC Innovation Banking backs innovation globally from seed to IPO. And if you ask me, a strong banking partner like HSBC belongs in your stack.

1:03If your portfolio companies are scaling, they need infrastructure that won't slow them down. Google Cloud Starter Program offers$2 ,000 to$350 ,000 in credits, plus technical support to build better and faster. It's a key boost every fund should bring into their ecosystem, and oh my god, are we thankful to be partnering with them. Now, legal is a space you cannot lag on. Legal needs to move at the speed of venture. Goodwin's team has decades of experience with startups and funds. They're trusted at every stage from formation to exit. Goodwin definitely is a legal partner. Every serious manager should have in their stack.

1:39For Luxembourg based VC, PE and Fund of Fund managers, modern funds means going digital. Fundcrafts gives you a full service digital native platform built for today's European managers. It's a must have if you're scaling smart. So we all hear about the Middle East. How about you go there from AI to deep tech to summer fund SkyTex in Dubai is where global future of tech gets negotiated. It's not just a conference, it's where East meets West, capital meets innovation, and the both set the agenda. If you're playing on the global stage, join us going to GuyTex this year. If you're gearing up for your next fundraiser and want a placement agent who truly understands emerging managers, reach out to C-Funds, their boutique placement agency that has helped GPs across Europe race capital from top tier LPs.

2:21We've been on the other side of the table here. They are actually good ones to work with. So I do urge you to go to cfunds.io to go and check them out. And hey, before you go, if you're looking to discover startups, raise capital, or connect with innovation leaders, do check out dealflow.eu, the EU-backed platform, bridging founders, VCs, and corporates. There's no better place to find the startups that have received significant funding from the European innovation ecosystem.

2:50Tear down this wall. It's more than just an ally. This is a union of values. Let's start acting. This show is not investment advice, and the hosts of this episode may be invested in the funds and companies featured. Let's dive into it, Michael. Thanks, Andreas. You make me sound like Uncle Scrooge here. I suck. Nobody wants to hear from this guy. Nah, well, maybe they do, maybe they don't. I don't know. He was a pretty smart guy, that Uncle Scrooge. All right, so before we get into the playbook and the framework, let's start with the moment that sparked this warning. Michael, what was the moment that really made you realize that the AI valuation environment has crossed into dangerous territory?

3:36There are a couple of things. If there's one case in point, I think the stat that Bloomberg, or I think it was Bloomberg, right? That article about the circular economy, how NVIDIA is investing in OpenAI, who is buying NVIDIA chips, how Oracle is investing in OpenAI or buying capacity from Oracle for their server farms, etc. That whole thing is just not very healthy. It's insanely unhealthy. So I think that in combination with a couple of weeks ago, one of my companies had a big event in New York where they gathered a lot of very smart people and forward thinking companies like Google and Microsoft, and they're foremost experts in terms of like authentic AI and what's going on.

4:23And the lead Gartner analyst showed a slide in terms of they served 4 ,000 companies, like real companies, right? Not startups. Out of these people, there were only 8 % who had live generative AI applications in production. Everyone was toying around with it. 92 % of that was pilots, trials, sandboxes playing around. So I think the investment that we're seeing into this space is obviously massive, but a large part of that is currently driven by trials and pilots. And once people figure out what they want to do, that pace of deployment is just going to slow down because it takes so much longer for people to actually adopt true technology, implement it in a business process, integrate it with all sorts of other different systems to drive real productivity gains.

5:18So I think that's, you know, seeing that a couple of weeks ago, those two things, that's what made it, you know, tick in my head. But then let's dive into it because how do you then distinguish between this pilot driven early traction that your portfolio of founders and others are experiencing and kind of getting that separate away from quality revenue? Our idea is to try to stay disciplined and when we're looking at a company and opportunity, try to distinguish from, is this momentum from pilots and trials? And if so, what do we think are the precursors to understanding whether this is more sustainable revenue, whether it's going to convert or not?

6:03Or, which is more of the longer term framework, if you think about AI applications, are these customer examples we're seeing, examples where you really are integrated into a workflow. Are you embedded and integrating with different systems? And are you generating through these different integrations a sort of unique data asset as you aggregate data in this business process with a high frequency? I think we're trying to distinguish these things because a lot of companies right now that we're seeing are they're sold on an individual basis to an individual user who is adopting that. And that's great and good, right?

6:44But these companies behave more like consumer companies. And if you want to be a true enterprise company, you want to understand how do you go from that individual motion to a motion of selling to multiple stakeholders. And ideally, you're already embedded in these types of processes. But there's not a lot of companies that are there yet. while you're trying to make this distinction and and try and separate it because obviously most products most startups will be in this stage where it is primarily pilot revenue how do you on surface whether like how much you expect can actually convert to the next stage can you open up the playbook on how you diligence that yeah i think that there are a couple of different vectors to to sort of follow there i think one of them is the usage so like if you can see usage spread from an individual user to a team to a department, I think that's one of them.

7:39And similarly, if you can see usage spreading from a particular task to another task to something which starts to resemble a process, it might not be formally that you have implemented this or chosen this, you know, in competition with four different things or done an RFP of like, where you're seeing in the usage patterns that this is actually starting to meld into something which looks like a business process and adoption by multiple stakeholders. I think that's what we're looking for. And clearly, even if you're in pilot, if you are generating outcomes, which, you know, based on just sound logic have sound economic return on investment, that's very different compared to if people are just playing around with it and think it's a cool thing.

8:29So I think those are the sort of things, like that usage pattern and is this deployed in a way where you can actually just theoretically see that there's clear economic value here. If we start on the founder side, how do founders best showcase this clearly to investors? Because it's obviously rounds are happening very fast here. There's competition. Everyone, so to say, when there's great growth numbers, a lot are interested in coming in. How do you make the case most clearly that this is quality revenue, this is a quality adoption process. I think it helps if you have domain knowledge and understanding of the business process that you're addressing.

9:08If you have experience, whether this is in sales or marketing or in product or whatever it is, or let's say within the financial services industry or within the healthcare industry, if you're selling an application with a use case where you have specific domain knowledge and understanding and some sort of insider perspective, I think that helps you to much better articulate why this is something which is long-term valuable, even if people are still in the experimentation or piloting phase. I think that's one of them. I think similarly, if you can show how people are initially trialing something, maybe on just a best effort basis, buying a couple of credits or whatever the model is, if you can see that commercial relationship starting to move over time, and this doesn't need to be years, like this can be months, if the commercial relationship moves to, even if it's credit-based, they're buying more credits because they want to do more of this.

10:12And if it starts to move from, you know, credit volume-based towards something which looks more like, you know, a subscription which is tied to a general outcome or something, I think that's sort of what you're looking for as well. So the main understanding and knowledge that you can articulate the business case and where the customer is verifying that, but essentially agreeing to tweak the pricing in the business model towards something which is long-term valuable. Are there any supporting metrics that you say that go very well in hand with the revenue numbers and growth numbers to show that this is more quality than others?

10:50I think the golden metric or standard for understanding true product market fit is obviously retention, right? Are customers staying with you? And I think in the old world, when companies, companies were only triple, triple, double, doubling, where they're not going to 200 million ARR in the first year, you can look at cohorts of customers. You could see one year in, who's staying, who's churning, is there a particular ICP? And you could understand all that is still applicable, but you have to do it at warp speed, right? But understanding, is there a particular use case? Is there a particular ICP that is actually showing these positive patterns that I was talking about over time, probably month on month, if not week on week?

11:37That would be, I think, the precursor signal to something which spells into higher retention down the road. Let me ask you a slightly different question, which is, in the beginning of the AI boom, everyone was saying that there's no point in doing anything in the application layer. Everything is happening in the LLM layer, and they will just be integrating forward and building their own applications. Then we've had a long period where everyone kind of came back to saying, well, actually, how should we put it? Internet 2.0, everything was a wrapper on a database structure. Now everything seems to be a wrap around an LLM and it's actually fine.

12:19But now there are people that are starting to say, well, you should really be careful with chat GPT as an example, because it seems like they're definitely going for the application layer. So Daniel, how do you think about that for founders? Because we're all building on LLMs. I think it's a nuanced discussion. So I like your analogy to the web, right? It's like, there's no point in building your own database, right? There were plenty of providers of that. There was a point in time in the client server area, an application server, which people use for like foundational stuff as well. Like building on top of that infrastructure, I think is the way to build software.

12:56Reinventing the wheel is not the way to go. I think if you are an application level company, if you're building something which is truly native AI, I would not shy away from building on top of the large public LLMs. at least not initially as you start. I think there's a couple of things to think about. First, I wouldn't build dependency into either of them, right? I would try to isolate any dependencies I have so I can basically switch between them. I think that's the first thing. That could be both from a strategic perspective but also from a cost perspective. I think the second thing is just like in the web area, If you just like put a simplistic interface on top of a database, you were really exposing yourself to someone with a simple no code tool building that interface and that application for themselves.

13:50It comes back to really solving a critical business problem. The way you do that is you embed yourself deeply into a business process and you orchestrate the execution of that across multiple stakeholders. So you become the sort of system of intelligence or interaction, really. That, I think, is the key thing there. And if you don't do that, I would be careful investing in an application company. But if you do, I think there's every opportunity to build this similar, if not bigger, values in the application layer compared to some of the giants of the last 20 years, like the Salesforce, the Adobes, etc.

14:30et cetera. Talking about all the AI discussions that everyone's having, let's go to another one, which is execution as a moat. There's some people that are saying, well, now it's the only thing you have left. And then there are others that say, no, data is going to be your moat. So how do you think about building moats in this world of AI? Yeah, I'm going to sound like a broken record or a repetition here. I don't think speed is the only moat. I think orchestrating a business process in a fashion which is more appealing, which is better, more effective for a number of stakeholders that have to work together, I think that is still very valuable.

15:11The problem with this approach for the ERP area was like, you try to really force people into working in a particularly defined process that you had defined if you were SAP. I think this needs to be much, much more moldable and interactive and how this evolves over time. There are people experimenting with these personal interfaces. My interface is different from yours, etc. I think that's a totally different application level experience, but it's far from the LLM, right? And it's very much abstracted from just what the LLM is doing. So I think that's one of them. And I think that's really, really important there.

15:48That, for me, if you do that, and again, if you execute that business process in a fashion where you're generating unique transactional and inside data, which helps you to build an even better user experience iteratively over time, I think that's a massive long-term mode. We started out in the beginning with you saying, I'm going to sound a bit like Mr. Scrooge here. Let me give you one thing that will definitely make you sound like that. Because coming out of the big tech hype bubble last time around, we had a lot of VCs that were commenting at the same time as, yes, it's a tough market, blah, blah, blah.

16:27We all love that there's a lot of money in the market because it obviously makes our business model work super well. But on the other hand, it also made us move very, very fast, make decisions that maybe we should not have if they were very honest. valuations were pumped and so on. And a lot of them said that they would wish that founders were more thoughtful about, despite them being able to raise very large rounds, but also despite them being able to close very quickly, that they would spend more time fundraising. Because in the end, whether it takes a week or five weeks, and you actually spend the time to diligence the investors you end up taking on board, that's just a very healthy process.

17:08So could you talk a bit about that part. And I'm really giving you a Mr. Scrooge question here where you get to be the old guy saying, nah, it's good that we get to know each other. Don't get all caught up in the hype. Honestly, it amazes me still to this day that entrepreneurs aren't doing more due diligence on investors. I think that is something most entrepreneurs get wrong. Honestly, they should do much more due diligence on the investors and not just on the firm, on the individuals they're taking on board. They absolutely should. Full period stop. As VCs complaining about the pace of the market, you don't have to write the chat, right?

17:47Don't put it in the hands of an entrepreneur, which might be a first-time entrepreneur, fantastically smart, young individual, hasn't done this before. How can you say to them, it's like, no, you should take your time. You shouldn't take the money. Of course they're going to do that. It's up to us to be the grownups in the room in that way. Honestly, I don't buy that. I just don't. Then let's move fully to the investor side. This is obviously the usual space of EUBC to be doing a lot of episodes. So we can have a very long conversation here, but I'll try and make it a bit short and snappy so we don't go on forever.

18:24First and foremost, as an investor looking in at the market today, what guiding principles are you trying to instill in your team when it comes to looking at AI companies. First principle thinking, what is the basis for anyone to buy software? It is to get productivity improvements in the way you execute work. If you want to invest in a company, you want to understand how is this company going to be embedded in a business process where it's going to create productivity for people. So first principle is thinking. Yeah. Then let me ask you about the valuation question we spoke about just before.

19:08How do you feel right now that you need to navigate the very high valuations we're seeing in the market? We're probably on the conservative side when it comes to valuations. But again, we're also looking for commercial proof that's commensurate with that. So, you know, you know, if we pay a high, high multiple, we think we have the basis for doing. But what I was saying in general is that the big problem is this, there are absolutely going to be a number of generational companies built. I mean, you probably, I think, you know, OpenAI probably already has escape velocity and like, you know, you can't stop them, There's going to be a number of those.

19:50The problem I see is if you look at how the software market looks pre-AI, you have a handful of the large infrastructure super scalers. You have a handful of really, really large software companies that are worth more than a hundred billion in the public market. The problem becomes when every AI company is being funded as if they can become the the next generational company, and that's just not the case. So I think at the aggregate level, people are paying silly money for AI right now. However, a number of those investments will be fantastic investment. And that's the really dichotomy, which is so hard to get around.

20:33So I think for society, it's honestly great that there's so much money being spent here, because that's going to weed out the really great performers from the non-performers. And we're all going to benefit from that and have fantastic infrastructure on top of building application stuff but the challenge for the VC industry is when you go back and you look at the stats for VC funds from vintage 24, 25, 26 it's going to look like shit there's going to be a really big hole of people who've thrown money at stuff that's not going to work then there's going to be a number of real outperformers who've done phenomenally well.

21:11And how do you try and make sure that you end up in the right bucket. Our strategy is the same that it's always been. It's like, you know, climb another hill. Don't try to fight to get into the most hyped and oversubscribed rounds. Try to figure out these companies that are flying a little bit under the radar, that are building something where there is fundamental critical business value, may not grow, you know, from zero to 10 to 100 in a year, but has the potential to grow into a very, very large company over time. I think that that's the model we've always had and that's where we are. AI or not.

21:47I've had quite a few conversations on the podcast lately with on the one side investor, on the other side, AI founder. And a theme that keeps coming up is that the team that you need to build an AI company is considerably different from in the past. And I've had two just recently that where they were both serial founders. So they've built in the pre-AI era and now they're building in the AI era. And it's smaller teams. It's much more technical teams. It is almost like one of them referred to it as, this is like basically leading a team of entrepreneurs comparing to before where now he would have 10 people, where before he would have, at this stage and at this level of growth, they would have 60, maybe 100 people.

22:34Can you talk about how you think about team composition and what you look for in entrepreneurs and the founding teams? I think it's a super interesting topic. And I agree with the general premise is that the availability of these tools pushes you into a place where you need a different skill set, a different type of person doing a particular role. And the whole configuration of the team probably changes. So certainly, yes, you'll have less raw capability to write code. I will say that the criticality of the lead architect who is really like the brain behind the product is probably even more critical now.

23:17Because you're leaving so much up to someone else's execution capability that you really need a super strong founding architect. Otherwise, I think you're going to end up in a sprawl hell. But I do think you can get much further from a product basis. you can iterate much faster. You can probably course correct faster and like change a particular element into something else. So I agree with that premise. What I do think though is, yes, you can build early commercial traction faster. And if you are a PLG motion company or if you're an open source company, that initial traction, you can also build that with significantly less resource.

24:02However, again, and I'm coming back to this like the broken record, once you get into the enterprise sales motion, where you're going to have to navigate multiple stakeholders, where you're going to have to clear a number of RFP criteria, the job is still the same. It doesn't matter what the product does. Let me ask you, because I want to go to the very opening where you described a macro environment that is very high BN with circular economies happening. I'd love to ask you, are we, do you think in an AI hype cycle that will bust soon? Or is it more, because this is what some people say, that you honestly have tons of businesses where you do have a bit of a circle of motion of how money goes from one part of the system to another.

24:52So maybe this is just what it looks like when you do big infrastructure projects, as an example. I, for one, think we are in a bubble when it comes to expectations of how fast the capabilities of AI will generate cash flows from companies. I think we're in a bubble now. I think we're overestimating how quickly these companies are going to grow over the next five years and how much that is going to convert into cash. I think in that way, I think we're in a bubble. So again, it's really hard to weed out what it's going to be. Let me just pause you there for one second. So because what that means is in a way that the companies, what they're doing now, the startups are doing now, it does not mean that their products are not valuable.

25:42It does not mean that what they're doing is not getting us all massive productivity games in the future. What it means is that they're valued right now at a price that requires them to sustain a very high rate of growth, which you're seeing that given the adoption patterns you're seeing in customers is unlikely to be able to come to fruition. So in other words, it's not completely like because I think many hear VCs or commentators say that we're in an AI hype cycle and it's going to bust and then make the inference that VC and AI will not work like that. there are not massive productivity gains that this technology is not as transformative as we think it is.

26:37And that is not the truth. The truth is just that we have very high multiples that are very hard to grow into in such a short timeframe. We're getting carried away. The basic is this, we're human beings. Human beings are prone, and this has been shown time and time and time again, to overestimate the impact of a massive technology shift in the near term, while we're underestimating the impact over the longer term. The Gartner hype cycle, I think, is the perfect case in point here, where pretty much any technology that has transformed society over the last 100 or 200 years, you can probably say it's followed that hype cycle.

27:16I think we are following that hype cycle this time again. We are now at the peak of inflated expectations. there will be a drop into what they call the trough of dissolutionment, but then they're going to start to climb the plateau of productivity. Just as in any technological shift, that's going to happen here again. And I, for one, am a fundamental believer in the long-term ability of AI to transform society. Yeah, I absolutely am as well. Okay, let me just spend the final minutes here on the podcast to ask you about how you then think about AI for you. You've described it before, somewhat on the broader spectrum, which is that you're looking for AI that's embedded in real business processes.

27:58But what does that look like in practice? Who are the founders that should come to Ox and say, we think we're a match? Yeah, that's a great question. And I think, you know, coming back to your question on team composition, I think it does change a little bit. because you need people who are... We really want people who are domain specialists and experts and really understand a business and understand the problem they're addressing acutely. And they exactly know how to experience that. But I do think if you built a company in a vertical SaaS space, started five, 10 years ago, there will be a different composition to the team.

28:40I think you need more flea-footed development talent that can solve particular things very quickly and iterate in a different way. So I think there's a difference. But when it comes to a core founding team, we still want to see that real domain expert, which probably means somewhat more experienced, somewhat older, and with a few scars to prove it. I think that's one of the call Frank Cartier. I think often, not always, but often, we also like, I don't know if we can call them dark horses or what you want to call them, like, you know, founders and companies which aren't completely evident to everyone, like who have had to fight their way, but who've always maintained true to the vision of what they want to achieve and what they're trying to do.