Niels Martin Brøchner

This transcript was generated automatically (YouTube auto-captions) and then edited for readability: names and obvious mis-hearings corrected, filler words removed, speaker labels added where the speaker is clear. Passages that could not be understood are marked [unclear]; short omitted passages are marked [...]. Timestamps are approximate. The recording is the authoritative version.

Niels Martin Brøchner 00:00:00

I've done, I think, like more than 100 interviews with CEOs and C-level people and the 2% problem is a problem, there's nobody who hasn't said it, directly, indirectly. Ultimately, like, the theory of constraints is basically you look at what constraints have been removed for you to actually take a completely different task. Like the best people in the world kind of deal with like heartache, setback [unclear]. It's like if you're a student or fairly young, like new in your career, yeah, I would go for the learnings and for the educations and what I can get out of it more than I would go for the money short-term.

00:00:37

We need to know what somebody needs to know before they know that they need to know it. The aha moment was probably

Konstantin Girst

Welcome back to a new episode of Campus To Capital and before we jump into today's conversation I think there are two important remarks to make. So, first of all, I'm sure all of our listeners will either see or hear, maybe even both,

00:00:58

right now that we upgraded our equipment. So, hopefully now we will be able to bring some more pleasure to your ears through our episodes. And then maybe kind of as an intro for today's episode, the idea for this episode kind of came to life when we talked about a recent podcast with Jack Dorsey, who's by the way the only entrepreneur ever who was standing CEO in two S&P 500 companies at the same time. And in that said podcast, Jack Dorsey criticized that we haven't really challenged and effectively changed the way how we thought about organizing people and organizations since the Roman Empire and that new ideas, such as being agile or flat hierarchies, have not really disrupted any sustainable long-term change. And today we're sitting down with the man who currently thinks about flipping all of this upside down and going about organizing not with a new dashboard or not with some new narrative but with something that changes how organizations fundamentally challenge and understand themselves. So welcome Niels Martin Brøchner. It's so nice to speak to you today.

Niels Martin Brøchner 00:02:08

Thank you for having me here. It's going to be fun.

Julius Ambrosius 00:02:11

Thank you also from my side. I mean the visuals are, as you can see, super nice. So we are actually in your office today. So thank you for that as well. And I mean we did a fair part of research before this podcast to also understand what you're doing and we figured that it's not that easy to grasp and to understand. I mean it's probably also hard to explain but if you would have to explain it to a 12-year-old, what essentially are you doing at Paradox?

Niels Martin Brøchner 00:02:40

Yeah. So, like to a 12-year-old, that's always difficult. I would probably, I explained it to my nephew the other day who's 15 and he plays ice hockey

00:02:51

and on a fairly high level. And basically the way that I explained to him is like if his coach was able to give him context or ideas, share ideas with him like ongoingly, while he would have a conversation with our product and it would automatically give him the information before going to practice or before actually meeting that person or before going to camp, he would know what his trainer was thinking about and he would be able to actually have a conversation with his coach's ideas and he would suddenly understand, because like it would be like, the coach believes you're super fast on the wing, you're able to do X, Y, and Z, you can penetrate with the ring like to the right and that makes you like have a competitive advantage on your, you know, your opposition, then he would be suddenly excited about doing it. Whereas if he was told that he's coming to practice and he thought he was playing center and now he's going to play right wing, he'd be like, "No, I'm a center. I don't understand why we're doing it." Coaches like, the team is set, go play. That is a very different kind of way of thinking about like coaching and mentally preparing people to do work. And he's obviously there to kind of, in the service of the team, which means that he's obviously interested in, like when he's playing for his team or the national team or whatever, he's interested in actually helping the team win but he needs to understand like how he contributes to it and what the coach thinks about his role, like actually helping. So that would be for a 15-year-old.

Host 00:04:19

I think it's also a good explanation for two 20-year-olds to be honest. I mean I think what's interesting here, maybe as a question to that, is [...] So for a 15-year-old who hasn't had that already, did it come to him very naturally or

Niels Martin Brøchner 00:04:50

I, 100%, like for him it made all the sense in the world. It was actually really interesting because in his opinion, his world, AI works. Yeah, like he uses AI all the time when he goes to school, when he prepares for training camp, whatever he does. Even when he prepares like a night out or something, he might, you know, like if he's going to watch a movie, he'll probably ask AI about that movie, what should I, where am I going, what's the setting, etc. Like so he uses it on a daily basis versus maybe my generation, like 40-year-olds or 60-year-olds or my mom who's 80, like my mom, she's 80 plus and she would basically understand it in a way where she saw it on TV as some expert explained doomsday scenarios and for a 15-year-old it's like the most natural thing in the world to understand that this is smart, like I understand how I could use this, this would make my life easier.

Julius Ambrosius

That probably also introduces one of your challenges, right? To make it accessible for like different kinds of people in an organization, which is something we will get to later on in the conversation. But I mean Konsti already talked about Jack Dorsey, that he addressed, yeah, basically the exact problem you're trying to solve. But I mean for every founder there is one point or one moment where they also have that moment of identifying this problem, of finding something that needs to be fixed. When was that moment for you and, yeah, how did you then come up with Paradox?

Niels Martin Brøchner 00:06:21

I think Paradox formed in a very natural way based on experiences. I led a company before called Contractbook and at Contractbook my co-founder always told me, you only know 2% of what goes on in the business, like so when you make decisions you're making them on a very undereducated level in reality, and I always thought that was interesting and then I understood that I didn't know everything but I didn't understand that it's actually a more common problem and then when I was no longer with Contractbook, I was interviewing a lot of peers and I was talking to, like, what should I do, like other CEOs and founders and stuff like that. It's like, what would you do if you're in my position, and kind of figure out what you want to build, or if not build or start a company, or go to work with another company, be a VC, whatever, right, what should I do, like take advice. Simultaneously, in parallel, I was going a lot in saunas. Sounds weird and

Host 00:07:27

well we are in Copenhagen after all

Niels Martin Brøchner 00:07:29

it's a hot topic, right, but a couple years ago, and I met a lot of CEOs in these [unclear] and like sat there day after day and I indirectly interviewed them about their problems. I didn't really know I was interviewing them, but I was just sitting having conversations with them and what I did and where I came from, my perspectives and stuff. And ultimately ended up being like, hm, I think there's actually a huge problem here that we might be able to kind of go after, which is an even more zoomed out problem than what Contractbook was. So contract management seemed like an interesting problem when we started; in hindsight I don't think that the problem might be big enough in reality for like the executives on the highest level wanting to work with it. So the aha moment was probably the analysis of why did people come to Contractbook first of all, and they came for a million reasons, but when they stayed, they stayed because they actually come to upload documents and get notified with information in the documents, right? And then it occurred to me that basically all alignment in companies is like non-physical contracts. Like you align, like we're doing this podcast, there's no contract, but it's like you would do the same. So basically all alignment is indirect contracts. So what we were focusing on in the paper

00:08:59

way or non-paper way at Contractbook, on contracts, to notify people about the most important stuff that they should be aware of in this contract, you could do that about everything, right? But it requires a different system. And then, so I had this idea to build a company brain, basically, didn't really know how to do it. I was selling to, I said to the investors, Speedinvest, I said, Tim Cook and friends [unclear], I was selling to, and I said that we're building, you know, with AI and it's going to be not a wrapper product, it's going to be a product where we're going to be a better product with AI progressing, so we need our own IP. So we had like some first principles that we're working around and that made it pretty clear for us like how to build and what to build and for whom to build. And then I got introduced to this book called Theory of Constraints by Tom, my co-founder, and then from there on it added up itself and then he explained basically world models to me at some point in like a side sentence and I was like, whoa, what we're working on right now, maybe we shouldn't do that, maybe we can't do, and I called him, was like, do you have a whiteboard, can I come to your place, and I drove to him and, obviously he has a whiteboard, he's an engineer, right, at his home. So I drove to his place and he explained it on a whiteboard and we talked about it, said that's what we're going to do, like we're actually going to build that thing, and at that point I didn't know it was called a world model, it was basically just like a proposition model that found the proposition via reference in

00:10:29

a very simplified way and that works a bit like a brain

Host 00:10:33

and for listeners, could you maybe expand on the term world model and kind of conceptualize it for them?

Niels Martin Brøchner 00:10:40

Yeah, in a simple way, world [unclear] from physical environments with sensors. So you have robots for example that would work in an ever expanding world model and when one robot learns how to use one finger it can apply that learning to all robots. In the same way we try to translate that, so when some employee has learned something we want all employees to know it, if they need to know it. So the idea behind is that we need to know what somebody needs to know before they know that they need to know it, in reality.

Host 00:11:15

So it's all about context, right?

Niels Martin Brøchner 00:11:17

Yeah. It's all about knowing what other people need to know and also knowing what it doesn't know, right? Because it's a self-learning model. So if it doesn't know something, so let's say one of our team members has added a half truth to the system, it would ask that person like, can you elaborate, or ask other people, do you agree with this statement, because like is this a truth or is it just random nonsense? So it informs me what I don't know. It asks me questions about stuff I should share information about. But also tells me stuff that I didn't know I need to know, right? So it tells me when two of my employees, I think I've done a stellar job communicating, obviously. So I think that I've done my job. I've communicated super clearly. The day after, I get a brief, we have these morning briefs that take about a minute and you get like four of them. You get that information and I can hear that two employees are definitely not agreeing. They don't have the same perspective. Right. So at the base level, what it does when it's worst, which is basically now, it actually just leverages the conversation for you to have it. So like, so let's say it's you two guys and you have a perspective on a session you're going to do with somebody. It suddenly is going to be able to flag that you're actually not agreeing on a part of the podcast and you thought you were agreeing on, but like the way you speak about it, they can actually figure out that you're not agreeing and then it kind of just pushes you to have that conversation. And that happens all the time in businesses where people think they have the same

00:12:47

perspective. The leader said the goal is this. We're going to do this. Then employee X's perspective and interpretation of that is one perspective and the other one looks at it from a different angle. And if we can solve that we're far ahead I think. But ultimately what it does now, it does that already. So it's the worst it's ever going to be basically right now. So which is, you know, fantastic news for us.

Host 00:13:14

It's fascinating. I mean, I think that's definitely something [unclear] oftentimes misaligned and, I mean, I think the fascinating part about this is, you know, identifying this misalignment although the people think they're aligned.

Niels Martin Brøchner 00:13:34

Yeah.

Host 00:13:34

Yeah. So, it's super interesting.

Niels Martin Brøchner 00:13:37

We've had, we're a small organization now, had it multiple times where I've been able to listen to my morning brief, which is basically a gap alignment, or it could also be consensus alignment depending on how you want to frame it, and I can hear that two employees that sit 2 meters from each other are not agreeing or something, based on conversations they've had for like a couple of minutes, because of a hesitation, or that one person used one word to explain the situation, the other person used this, and then it digs into that kind of wording and that frame of that perspective and then it would bring up other information from the business, like from, let's say, employee, you know, to further out, and it would bring that into the conversation. And it brings some of the first principles of the business, how we work, like how we see the world, etc. And that just helps steer people into where they need to be. Like we try to say we want to prevent the spark before the fire starts, right? So if we can make sure that these businesses prevent the spark, no big fire is ever going to be, which means that you need less big pivots. Yes, obviously there's going to be stuff that's out of your control, but you need less big pivots because of internal stuff and misalignment.

Julius Ambrosius

And I mean, if we now look at this alignment issue historically, in the introduction, Konsti said this has been a problem for over 2,000 years. And I mean there are a lot of companies that

00:15:07

work with business intelligence. I mean there are the SAPs, there are the McKinseys and they've been trying to fix that issue for the past, I don't know, 60 or 70 years. So what do you think, why did they, or like why don't they actually solve this problem or why hasn't this been solved already?

Niels Martin Brøchner 00:15:29

Theory of constraints.

Julius Ambrosius 00:15:30

Okay.

Niels Martin Brøchner 00:15:31

Like you need to look at what constraints have been removed for what they did to start with. So like the way we apply technology and how the generational technology moves, like on a cadence, currently there's just been moved a ton of blockers for how you did it. So if you look at how computers started, how they wanted to work with computers 50 years ago or whatever, the idea was to do this constraint as organizations but you couldn't because the technology wasn't there. So we created all these infrastructure systems, frameworks, basically binary systems, zeros and ones, like data systems, for it to translate into something that we could do at scale but it's actually not how humans work. It's not how agents work either, if you use an AI agent, but ultimately the theory of constraints is basically you look at what constraints have been removed for you to actually take a completely different task. So when ERP systems came, that allowed people to work from a cadence from let's say half a year to maybe a month, because like suddenly, but the reason why you wouldn't do less than a month would be because of accounting. Like it would be so expensive if you wanted to account on a daily basis, that wouldn't make any sense. You would still keep a month or quarterly cadence because if you have somebody to do the [unclear] every day, that would be more expensive than running a system. So you can only do that on a quarterly basis. That's why you have basically accounting, right, like even everywhere basically. The question now is, if you don't

00:17:01

need human power to do that, like to do the accountant, could you do it on a daily basis?

Host 00:17:07

Theoretically you could do it.

Niels Martin Brøchner 00:17:09

Yeah, so that applies for everything. Like so there's a ton of things where you could look at it, and if you look at which constraints have actually been removed and then look at the same problem and then you apply the newest technology, and it's, to use a more lame analogy, it would be like an engine on a horse wagon.

00:17:27

Like that doesn't make any sense, right? Like then you need a different steering system, but suddenly it's way better. But it requires a dramatically bigger shift.

Host 00:17:36

Yep.

Niels Martin Brøchner 00:17:37

Like because you suddenly have the combustion engine. You can actually put it in and then it makes sense, but you wouldn't put it on a horse. It wouldn't make any sense.

Host 00:17:45

Paradigms have to shift. Yeah. No, definitely.

Niels Martin Brøchner 00:17:47

So I think that's also just to kind of like full circle on that. Sorry to rant here.

Host

No,

Niels Martin Brøchner 00:17:51

that would be why you would try to not just apply AI to old data systems, because it is old logic, it's not how you would do it if you were redoing it with the constraint removed.

Host

I feel like that's super interesting because

Niels Martin Brøchner 00:18:11

sure

Host 00:18:12

[...]

00:18:19

[...]

Host 00:18:21

Exactly. [...] So database is everything. AI models have to work on that, so on, so forth. But it's, yeah, fascinating to hear that like that isn't actually the forefront and that there's something

Niels Martin Brøchner

You don't have to listen to me, right? You can listen to Sequoia and Jack Dorsey and then you could, you've been working in world models for a long time, like how you train one thing that applies to everything. So yeah, I do think that you need to kind of rethink how you do stuff. I'm not saying that what we do is going to be like the certain truth in the future or something like that. What I'm saying is that the current format of data logic and how we're storing data is not perfect for AI. And I also know that for my old company we took a direction on a foundation, like data layout, called JSON, which is binary data format, so we could use that for AI purposes. Ultimately we all know the [unclear] story, right, where it ended up being LLM models and not, you know, JSON models basically trained on, which means that the frontier models are not working with data systems like

00:19:51

binary systems that like the lectures are referring to, they're actually working with what we're doing, they're working with propositions. Like currently the setup is built on propositions, like everything when you're chatting, it's propositions. No,

00:20:06

like so when you're trying to build an infrastructure for agents to work in, it's actually better if you have, let's call it, a proposition model or a world model with propositions in it because it's actually easier for agents to work in. They understand it better because how they're trained is like our brains, and the limitation of the current setup is, that's why we're called Paradox, it's the information paradox. So we have complete information overload, data.

Host 00:20:35

Yeah.

Niels Martin Brøchner 00:20:36

And we're somewhat in decision paralysis due to that. So you should imagine that would also happen if an agent worked in our data systems. They would just spiral out of control and make a ton of data and a ton of systems that didn't make any sense.

Host

Let's dive into what you actually built based on top of that. You know, when you gave us some information to kind of grasp what you were doing before this podcast so we could understand some of it at least, you painted a picture where you talked about someone going into a meeting and us not necessarily being able to expect that everybody in the meeting will be on the same page. And I think it's a very interesting issue that I myself have not really thought of before because I think within a certain organizational context you always think, okay, if we're working on this everybody should kind of be in the know, but it is a very, I mean it's a very present issue, but it's nothing that I can fix by myself with my brain alone, right? I need some external tools, some external help, and the product that you built is called APA [unclear] and I mean there may be some VC GPs listening. But could you maybe give us not the pitchbook demo but kind of the user-friendly demo? If I log into the user interface, what happens and how can I actually improve what I've been

00:22:06

doing beforehand without APA [unclear]?

Niels Martin Brøchner 00:22:09

So the system is almost like a social network, first of all. So you have to imagine it as a social network where a new layer can't be rolled out before there's density enough, which means that there's enough context for this to make sense, because an empty app, empty feed doesn't make any sense. So what you would do is that you would upload a couple documents. So let's say you're an organization, doesn't really matter if you're a 10-man organization or a 500 or 250,000 man organization. But what would happen is that the leadership, which is the top of the social network basically in the organization, would add some documents, which is like your go-to-market documents, roadmap planning, some financial goals and plans, maybe some internal alignment documents with the board, etc. Then what happens quickly is that they figure out that the go-to-market and roadmap is probably not 100% aligned, first of all, and they also may be saying some stuff to the board that's not really aligned with the actual financial targets in the go-to-market model, etc. Not on purpose, it's just because like the human brain can't connect all the dots in the same way that an LLM would be able to do. So we basically build a mind map and can see how everything is connected and how it's connected. Then it asks questions to people, like so it gets them to [unclear], because it doesn't understand stuff. So it tries to learn. And then it gets all these answers into the proposition model

00:23:39

that lays, and then when the next level of employees get onboarded, the first thing they get next morning, imagine they get five briefs or four briefs, takes a minute to read, of information, of stuff where they suddenly figure out, now we're aligned, or we're actually underlined [unclear] here, there's actually some conversations we should be having over here that they otherwise wouldn't be having or would know before they met up in person, they had the quarterly meeting, or never. So basically, again, from day one you can actually start preventing sparks and then you can imagine that rolling through an organization. So it would ask like, do we need, you know, Charlotte's opinion on this, do we need Christian's opinion on this, etc., and then you would ask them questions about their specialty and what they know because it also understands the org chart basically. So that's how it rolls out from a very simple perspective. So you can imagine if you watch The Social Network,

00:24:33

how Harvard is basically C-level. Then you would have Columbia, which would be the VP level. Then you would have Ivy League that would be like directors and managers. And then you would have all colleges, that would be kind of everyone in the organization. But it's very important to say that the way Copilot differs from that is that, like, obviously they're doing a lot of fantastic stuff, but what they're not doing is really connecting to the top of business. So they're more of a, become more efficient down here, which for me creates more presentations, more emails, more double work instead of actually understanding what's the context of why am I doing something if I'm getting a job. Like you said, if you're working at Meta or somebody and somebody asks, like, why am I doing this, and you're like, okay, why, like you don't want to ask your manager why because that's kind of an obnoxious question, but here you can actually ask the [unclear], say hey, why, and then it says it's because [unclear] means X, Y and Z, we want to go into this market, X, Y and Z, so kind of the effect on this and how you actually affect this by doing this, like through the organization. So it actually connects to the organization, so it becomes more of a tool for meaning in many ways, and productivity, because people suddenly understand why they're doing stuff. So then they're not just moving email sideways and creating more presentations and saying they go to work, but they're actually, you know, doing stuff that's connected to

00:26:03

why, like the company objectives.

Host

So it also introduces like the meaning of psychological safety for those employees, because it's a back and forth learning, like the people are learning something because of the model and then on the other hand the model is learning because it gets more input, right?

Niels Martin Brøchner 00:26:22

Yeah, so a non-answer is an answer as well, right? Yeah, like so if you say,

00:26:26

they ask you about something, you're like, I don't know, I don't have an opinion on it, that's also context, right? Suddenly it knows that this might be something that's important for your role and somebody else could know. And then when it learns it, it will share it with you.

Host 00:26:38

Well, also next to psychological safety, I think just the way how we would, if every, I mean I think for this to be fully successful obviously it requires full compliance from all the people within an organization, as everybody kind of needs to contribute to leverage this to the maximum possibility. [...] I mean it's communication and turning over things over and over and over again is like one of the most inefficient things within financial services and I think this would actually make the life easier for a lot of analysts and associates doing a lot of grunt work late at night.

Niels Martin Brøchner 00:27:22

Quite certain it would. Yes.

Host 00:27:23

Yeah, definitely. It's also like, it's not really the purpose of your job.

Niels Martin Brøchner 00:27:28

No, for sure.

Host 00:27:29

Purpose of your job is kind of learn it the hard way and kind of to build grit and understand all the other stuff as well. Like they want to see who you are

00:27:36

like when they set up these kind of programs, because there's certainly more efficient ways to do it I assume.

00:27:43

Definitely. But it's not about me, but just as a personal reference to that industry.

Niels Martin Brøchner

But I think this is like when I referred to the 2% problem before at Contractbook. I've done, I think, like more than 100 interviews with CEOs and C-level people and the 2% problem is a problem, there's nobody who hasn't said it, directly, indirectly. It's like a problem in every business and if you talk to a McKinsey consultant or Bain or BCG consultant about this topic, they would say what I also have experienced, that it's not that the employees think there is a bad strategy. They actually think there's no strategy. So when you go in, and let's say a BCG consultant goes in and interviews a Goldman Sachs person, right, about the strategy, they would say there is no strategy, and that's kind of the most insane thing, that's what the consultants report. And like, there is a strategy. I know there's a strategy and there's obviously a very good strategy by super important, super smart people who create it, but there is just no system to actually communicate it properly, because people are tired of really getting the same every day, and what you learn day one at, you know, business management, or not business management, but when you come out in the real world, is you repeat yourself

00:29:13

until you puke [unclear], and nobody wants to do that. So if you have a system that doesn't care about it and doesn't start puking about saying the same stuff, like APA [unclear], then they would be able to constantly kind of guide you without being tired of guiding you.

Host

I think what you just described is also one conception a lot of people have about consultants, that they go into a company, they get a mandate of, I don't know, a couple of hundred thousands, and then after 6 months they leave and basically nothing really changed, maybe some layoffs. How do you make sure that when an organization onboards you that doesn't happen?

Niels Martin Brøchner 00:29:55

Like, I don't know enough about consulting, first of all, like the consulting industry, but what I can say is that what we do, we don't leave. So what a consultant does is, it might be more qualitative for the three months they're there, but they also leave, right? So suddenly that system that they created doesn't work anymore. So they set out like a brilliant OKR system of formal alignment and, now we report this way, build the org chart, remove these people, add these people, everything's built correctly, and then they leave. So the orchestration of the new system is also them and when they leave the system often breaks, as far as I understand, but I haven't worked enough with these kind of types to really know. But in the simplest format, we don't leave. Like we keep on being there and trying to orchestrate knowledge and learn why are these two people not fully aligned, how can we actually give them a common perspective, which is like align them basically.

Host 00:30:58

Yeah, this is also by no means any consulting bashing here.

Niels Martin Brøchner 00:31:03

No, no, there's not a [unclear] or bash consultants. I just want to make sure that I don't really know how they work in an enterprise.

00:31:12

But I do know that they leave.

Host 00:31:14

Okay.

Niels Martin Brøchner 00:31:14

Quite often.

Host 00:31:16

So you would say like one of your main value propositions is like the omnipresence of your model.

Niels Martin Brøchner 00:31:21

Yeah.

Host 00:31:22

Okay.

00:31:22

And tackling the issue from within I think is also a good idea to have. And now we talked about what's super exciting about APA [unclear] and what will probably also allow it to be a very successful product. But maybe on the other hand, to kind of face a potential threat or downside, is, I mean, with any memory-based and then data-based product obviously a huge risk is sensitivity through the data that it holds, because if you actually leverage it to perform major business decision-making, obviously it will hold very high security, high-risk, high stakes data and it will be a huge threat for any organization using APA [unclear] if there is any data or information leakage, maybe not just only on a technical level but also from a psychological perspective. How would you, or how are you addressing this issue? Because I think, I mean, you need to be able to guarantee full security but then also on the level of trust. I think people also kind of need to be fully aware that what they're doing poses no potential harm for the organization at all to be able to fully devote and engage with.

Niels Martin Brøchner 00:32:42

Yeah. Yes. [unclear] times and then you trust it. I think that the onboarding, a part of the onboarding is trying to create this psychological safety and awareness about how world models work and how the security works and what they can share and what they shouldn't share. There's obviously stuff that, we're, you know, a year and a half into this journey, which means that the companies working with us see the upside even though they know that the risk is, you know, versus SAP, 50 years old, right? Like obviously their security is different than ours, but the upside of what we can do with the data is different. So it's a risk-reward kind of thing. When companies go into it, they understand the risks with it and potential risks with it. And I don't want to go into details because it obviously is safe and needs to be safe, otherwise they can't do it. But more importantly, to your point, is that we spend a lot of time trying to say that people can't be onboarded in our system without actually having a real onboarding, like physical presence, where we're there and we actually educate them for multiple hours, and what we do, what they should expect, what they shouldn't expect. So let's say that a good example of something that humans react to is that if the computer says the wrong day, it calls yesterday today or the other way around. Then what we can see is that if we haven't made them aware that obviously it can do that because of the resolution

00:34:12

and where we are, that's annoying and that could create mistrust. But they have to think about that it actually surfaces some information that they otherwise wouldn't have had, that actually makes the company better, because like you two can sit and look in a calendar or a time zone and say the wrong hours, like when it's not on purpose, it's just because it's confused,

00:34:36

like it has a ton of information. So it's okay when humans do it, but when robots say something wrong people tend to be like, oh no, it's a robot, or they've taken over. Whereas, and we'll make mistakes in terms of like resolution sometimes, but more importantly we'll surface more information that's going to give them context to make better decisions and leverage the conversation in companies more than we're going to make mistakes on days. So we have to teach them about that stuff. We have to teach them about what they can say, what they should expect, how it acts as an executive assistant, how it acts as a consultant, how it acts as a coach, like what they should expect from that, what they should expect to be shared, like how a recap of their day works. So they can listen to what they just said, how it connects, so they can see how the world model actually works and how they can see what they've said and what other people say and how that's connected. How often they're going to get briefs, what they should expect from briefs. Transition coaching, like more kind of classical transition coaching, where they go in and actually have a conversation with a person who says, like in a management team, here's tension, this is how you should expect, we're going to resolve it like this, this is how psychological safety works in AI, etc. etc. And that's 10, 20 hours of work and we're going to be there every day when they onboard in the test period because we want to show we have a human presence and we are accountable for what we, like we want to understand if they like

00:36:06

their briefs, if they like the conversations they're having, and we don't expect them to speak more and listen more and spend more than kind of 10, 15 minutes a day on it. So it's not like they're sitting on it every day all the time. It's not a TikTok.

00:36:19

It's not, but instead of you sitting in Slack all the time trying to surface information, you're just getting the right information for minutes and you're having a conversation about that information, then you don't have to spend more time with that.

Host

Now I could also imagine that the error rate is probably pretty low, because I mean that's the usual misconception about AI, that people say, oh, if it makes an error, who's accountable? But I mean, let's take the consultant example again, a consultant can also make an error, but the error rate of models tends to be much lower, right?

00:36:56

And you just said you work together with some Danish companies I assume, and you're working together with them in co-creation of model three phases, right? So I would love to know what surprised you when companies started working with your model, because I mean you have to make some assumptions when you build that model, how people will work and react to it, right? So what surprised you when you onboarded your first customers?

Niels Martin Brøchner

Yeah. So I think the biggest surprise is actually the openness to what we do. I think that there is a lot of, in many of these big companies there's actually a lot of momentum. We also work with international companies, equal the amount actually, but that's more like, hey, we need to find the mature companies, and suddenly you figure out that a certain company in Germany or the US is, you know, looking for a solution like this and you can actually have a real conversation with them where they're pretty mature, even though they don't know us. So I think that from a surprise perspective, I think it's the openness and the willingness to actually live with non-SAP security, and they're fine with that actually because they can see the outsized value. And sometimes we see people leaning in and

00:38:26

the companies that are leaning in and wanting to do it, but they have a defensive mindset, like wanting to fire people. That's where it fits worst. So we don't want to be a cost cutting project. We want to be a human leverage, like humanistic AI project, where we get humans to actually work on something more interesting and not just use it and prompt the AI to give you a new presentation that somebody never reads. Like what you want is actually it to give you information so you can make valuable decisions. And then you can progress in your business. Because that's kind of what business is about, is actually about progress.

Host 00:39:07

And maybe to quickly look back at your sauna conversations that you had with numerous executives

00:39:14

when you were trying to hear about their problems, and then now you actually took the great initiative and are presenting us with a very novel, interesting and cool idea where we see great opportunity for organizational change. And when you talk to these people again now, what do they think? Are they all super convinced? And if some are somewhat hesitant, what would be the main reason why they are hesitant as of right now and would not maybe immediately think about, okay, let's get Paradox and APA [unclear] on board?

Niels Martin Brøchner

Security infrastructure. [unclear] thing as the surprise, like I would assume that people would be afraid of a small company like ours handling some of the information and trying to kind of surface and leverage the conversation within their business. So I'm equally surprised. I'm not surprised by these companies saying we're not able to do it right now. I think it's completely fair, but I think the conversation starts with, when they get it they're like, when can we start? Like so there's this curve and then when they get super excited about it, then they go into like this [unclear] kind of space where it's like, oh, procurement, safety, what can we do, and then it becomes a bit more difficult conversation and we haven't mastered that yet in terms of how we go through all these phases in terms of enterprise sales. I'm not an expert on that yet. So

00:40:44

it's still work in progress, but I think that the biggest surprise is the willingness to bend their rules to get something extraordinary and different, but also the biggest pushback is obviously that we're not SAP.

Host 00:40:58

Yeah. No, fair enough.

00:41:01

But I mean I think also, at this stage, I'm sure it's also super valuable just to have curiosity, right? Because I'm sure that even if people are thinking, okay, maybe it's not the time yet to adopt it, I'm sure they at least think it's a very interesting idea and they want to hear about it, right?

Niels Martin Brøchner 00:41:17

Yeah. I think that, and many also say like, hey, when you're at the security level that we need, we really want to have this conversation.

00:41:25

Yeah. Cool. Like we think this is important, that what you're doing is different, but it's also like you have these traditional kind of curves of development in terms of maturity

00:41:36

where at this stage it is obviously super early adopters and people with a very particular mindset that buys a product and project like ours, and our best clients are going to be on the other side of the curve, like early adopters, and that's probably where the business is going to be very fun to be at, where the maturity is still kind of medium, but the early adopters are kind of flying in and being like, this is really exciting and we want to do this because now the safety and security and the infrastructure and everything is there. Yeah, so

Host 00:42:10

I was about to say I'm excited to see who's going to reap the biggest first mover advantage here. But maybe to move out into the bigger picture again, you have a stage model where you say that stage zero is where most organizations live today and that they get drowned in quarterly cycles and are trapped in fragmented context, and then stage one is 2026 to 2030, where there will be significantly shorter decision loops. And then stage two is post 2030, with decisions shrinking down from being done in quarters to maybe within days, within a single week. And I mean that's a massive claim because it's less than a decade for fundamentally completely changing the way we're doing things today. And how confident are you in this vision and how much do you believe in it and how much is it potentially also fundraising marketing language?

Niels Martin Brøchner

I can't say anything with certainty, right, and I can't predict the years and maybe 26 is too early, but I'm quite sure the 2030 is not too late. So the question is like when. Like you talk about, you know, AI within a couple of years we'll be able to do 25,000 years of work in 2 seconds or something like that. If that is the reality, it's not like, that's frontier labs that

00:43:41

talk that talk. It's not me and I don't know, right, but let's say it's not, but let's say there's a thousand years in two seconds. I would assume you can get a year worth of information fairly quickly consumed and processed. And I don't think that you need to work necessarily with a year wheel and quarters and stuff like that. I think you can compound enough context to make quicker decisions. Like in startup language you would call it OODA loops, where you make very short, where you basically, the best companies work the shortest open loops. So they basically just get enough context to make the next decision and learn from that. And in traditional businesses you work with data projects. So I think that the perfect analogy is, two departments are not agreeing on a number, let's call it CAC payback or something, right? They say, one says it's eight, the other says it's 8.5, right, and then they put down a data project and there's two heads of data from two departments that sit down together and the data teams work on it for nine months and they meet on a weekly basis and blah blah blah. I've seen it in my own company, it's not a great view. It's expensive. And ultimately, they'd never agree what's the correct CAC payback number. But the real question would have been in week one, if they agreed in the board that everything below nine would be go time, and they should just agree that it's definitely below

00:45:11

nine.

Host 00:45:12

So yeah, don't stick on the nitty-gritty.

Niels Martin Brøchner 00:45:15

Like the number doesn't, like the true data points don't make any sense. It's actually the context of what's the actual go time. So if you can make sure that the people, like in all cases, you can translate that to anything, like hiring plans, financial plannings, target sales etc. It doesn't really matter, if instead of discussing details that you don't understand the context of, you actually get the context and then you look at what you have. That's a very different kind of way of viewing, and I truly believe that, I think that world models or something equal to it is going to be like company brains in the future, like it's going to be something you can speak to and understand.

00:45:53

Is it going to be a proposition model like ours? I don't know, who knows, right? But ultimately this is the first glance into what I believe is going to be the future of companies and how the theory is going to work. And I don't think the 2030 is unrealistic, if frontier labs, and we have energy in the world and all that stuff, that will work.

Julius Ambrosius

[...] Namely, what happens to leadership when organizations get very good at alignment? Because, I mean, a huge part of what leaders do today is they arbitrate between conflicting realities, as they're the ones that hold the context. So if Paradox is actually working as you confidently say, what is left for the CEO?

Niels Martin Brøchner

Still decision making, just better, with better data. Like it's just going to be, you still need humans in the loop to make decisions, make the calls and understand the world and stomach feeling and all that stuff. So I'm not a

00:47:23

I think that you're just going to get better data, more clear data, where you should make decisions and not make decisions, and where people are aligned and not aligned, and that is going to produce better results for the businesses, ultimately. And our claim is not that you're going to get 100% visibility in data. Our ambition right now is 20,

00:47:43

right? They go from two to 20% with APA [unclear], like with our product. That's kind of our ambition, which means there's still 80% of information that's then untouched and still in people's brains and water cooler talk and that kind of stuff, but that's an order of magnitude better than where they are, right? So they get 10x better information to make decisions. So there's still a ton of work for people to actually work with that information. But if you boil it down to meetings, there would be less informative meetings where you would sit and discuss realities, like your perspective versus my perspective, because if APA [unclear] helped you understand each other's perspective and the business goals and perspectives, you always will be kind of within the frame of the business and the other people's opinions that you work with.

Host

And maybe one question in that regard, because companies have different hierarchy structures, right? [...] There's this famous example of the software company Valve.

00:48:51

They have basically a super flat hierarchy with no real leadership structure. Do you think that your product is suitable for any kind of organizational structure or hierarchy?

Niels Martin Brøchner

I don't see why not, but time will tell.

Host 00:49:12

Okay.

Niels Martin Brøchner 00:49:12

Like I don't see why not, because the reality is that what we want to do is we want to figure out how to get information from one part of a country to a different office in another part of the country without them having meetings, because they're working on the same stuff, that they actually indirectly influence each other and they should have the same perspective on it. And currently they're not even aware that each other exists. So how can they be aligned? And then that creates a lot of [unclear] work. So if you can make sure that even very flat hierarchies that are remotely spread over many countries, and they work on their projects and everybody's super happy and they're actually very productive, why is probably because they have a ton of meaning. They have a lot of autonomy. They have a ton of meaning why they're doing stuff. And what we try to do is create meaning, which creates a better kind of foundation for making decisions, which is going to create less formal alignment and create more autonomy, right? So ultimately you can probably say that the flat hierarchies are going to be even more [unclear] and the very strict hierarchies are going to be filled with more meaning and less, you know, less rules, where you can have more decision making given to the individual because they have enough context for actually making these decisions. Right now you ask your boss, what's the budget? Can I do this? But the reality, if you gave them context they would understand the budget and they would understand what they can do, based on them being human beings

00:50:42

and having a brain. Like they probably would be able to figure it out themselves if they kind of just tried, but we're not allowing people to think by themselves in businesses because we've set all these rules and hierarchies like the Roman Empire.

00:51:00

Like how does information travel? So he goes to that guy who has this and then he travels to this guy who has this, and the amount of perspective going multiple ways here, like up and down in the hierarchy, that can change is pretty extreme.

Host

[...] You know, it's a bit like that meme where you say like this meeting could have been an email, just like having a baby with modern technology. But maybe before we slowly start to round off today's conversation, stepping away from your product and also for a second looking at you as being the founder, I mean now you've already had some affirmation in your doings, for example by being backed by Speedinvest, one of Europe's most respected early stage funds, but being a founder is also super challenging of course. I think we don't need to tell our listeners about that; if they listen to these episodes, they've heard that from a lot of other successful founders as well. But then there's often maybe something where you think, maybe I should stop pursuing this, maybe I should do something else, whether that's like taking a long time to actually get positive feedback or, in the deep tech industry, that it's probably also very capex intensive. And was there a moment where you were also considering stopping to work on this? And then if there was such a moment, what kept you going?

Niels Martin Brøchner 00:52:30

It's a young [unclear] still. So I haven't met that moment yet. Well, probably at some point, obviously, you have them as a founder, as a leader of a company, even not a founder, just a person working in corporate life. I saw a study about what it requires to work in corporate, it's like the most insane level of patience and empathy, because you need to be so aware, like you need to deal with your own emotions on a daily basis of getting that frustration there is in a business. As a self-employed man or working in a small company, you still have the same thing, like you still work with the same frustration, you still have setbacks on a weekly basis, and you just need to kind of get over it, right? And I think what I say to my team quite often, I think what distinguishes the best people in business from the worst, and also in sports for example, is how people kind of deal with setback. Like the best people in the world kind of deal with like heartache, setback in 50 [unclear] minutes, right? And you can dwell with it for years. Some people dwell with a setback for years. You have people where they were a goalkeeper, a goalie in a football match or something, like they never came back. They literally never came back. They did like one bad game in a final or in the semi-finals or something on the big stage and they never came back.

00:54:01

So I think that yes, you have them all the time, but it's your own kind of ability to deal with that, both in corporate life but also in startup, is like how to deal with that frustration and try to kind of say, okay, that sucked. But then try to kind of figure out three good things. Like so let's say that I had a conversation with a VC and they said no. Okay, good. We still have runway. We still have a great team. Like we still have a great product. [unclear] So it could be, everything's, we need to close the business down, or you can kind of turn it into something where you look for like where are we and how can we deal with this problem, because quite often there's a lot of positives still. Like, I did bad on my exam.

00:54:52

Yeah, my family is still doing good. I still have my [unclear]. I still have all these things. I have my student job. I'm doing good, that I have my girlfriend. Okay, it's not that bad, you know?

Host 00:55:02

That's the right mentality. I agree for sure.

Niels Martin Brøchner 00:55:05

Like it's not that bad. Like it sucks, but it's not the end of the world. So yeah, that would probably be my solution for it. If I try, it's not like every time, like, if you ask some of my colleagues, you can also see me being pretty frustrated for

00:55:21

more than 15 minutes. Sometimes.

Host 00:55:24

Happens to the best of us, you know.

Niels Martin Brøchner 00:55:26

It's not like I'm perfect at all. [unclear] probably a pretty terrible example in many ways, but I think that's a good way to look at it.

Host 00:55:34

Definitely, yeah, I agree and I think it's also great advice for everyone that tuned in until now. [...]

Niels Martin Brøchner

I would do what I didn't do. I would go and learn from somebody else where the [unclear] wasn't, it was my wanting to be there and my wanting to learn from these people in something that I felt truly interested in. Because if you do a good job, you always get to money at some point. Like don't be, there's some money out there, like there's a good job that pays you well, but it's like if you're a student or fairly young, like new in your career, I would go for the learnings and for the educations and what I can get out of it more than I would go for the money short-term. Because you're going to be rewarded multiple times long term if you do that.

Host 00:56:55

Playing the long game.

00:56:56

Yes. Playing the long game. Exactly. I think that's a beautiful closing note to the episode today. Niels, thank you so much for taking the time. It was a pleasure.

Niels Martin Brøchner 00:57:08

Thank you for letting me.

00:57:09

Yeah, thank you so much. I enjoyed it.