A (4:56)

Yeah, I mean there's. Okay, so there's probably three big ones. Right. And two worth diving into. Cause one we've been with for a while. So one is just obviously moving out of the ZIRP era and the speed and pacing and scale of the rate hikes that we all have been through over the last few years has an undeniable impact on the industry. And not an awesome one. Right. It's just easier to raise money in a low rate environment than a 5% risk free environment. But we've been dealing with that for a few years now, so let's just sort of set that aside. But that's obviously one of the kind of factors that's been impacting the industry. And it hits different sectors differently. Right. Private credit's been thriving. Commercial real estate, you know, very capex heavy has really been struggling. You know, venture outside of AI seen a big correction especially in the early stage. So pockets of, you know, difference in how that rate environment has impacted private markets. But, but that's obviously a big one. But I think the, the two that are worth exploring probably are, you know, GPS are getting hit right now with like these two tsunamis that are cresting at the same time. One is AI and you know, we're in the middle or really the early stages of kind of computing platform shift that is bigger and more profound than any that any of us have been through in our lifetimes. And we're all, every business, every industry, but especially GPs are going to have to totally rewire and retool their business at warp speed. So that's a huge one. And then the second one is the rise of the retail investor. And just looking ahead at private markets growth and whether you think the industry is at 30 trillion today or 50 trillion and depending on who's counting, everybody sees 15, 20% CAGR ahead. And of course the two stories there are, we look at the next decade as an allocator, thinking about where to put your money. The return opportunity ahead is in privates. It's easier to find alpha than it is in the public markets where all of that has been really armed and kind of traded away. So there's great opportunity in private markets. And then the other huge driver is that the industry is really built on the back of institutional capital, public pensions, defined benefit plans, sovereigns, family offices, et cetera. And there's $100 trillion of household wealth that's really sort of sitting on the sidelines that especially now with products like private credit that are yielding liquid open, really good fit with what the retail investor and their advisors are looking for. Trillions of dollars of inflows into the industry will come from a new type of investor. And the difference for GPS here is we have a lot of GPS where they might have grown to billions of AUM with dozens of LP relationships because you know, you fly around the world, you go to nice dinners, you build meaningful relationships and you're getting checks in the hundreds of millions of dollars. And it's a totally different game if you have tens of thousands of investors writing you checks in the tens of thousands of dollars. So you will also have to retool your business and rethink the operating platform, rethink the products that you're creating to be a fit with what this new class of investor is looking for. So I think it's an incredible time for gps. It's awesome to be operating a bit, so much change, because if you kind of lean into it and embrace it, you know, there's great return potential. But it's not a time, I think, where you can sort of sit still and do what you've always done and expect that's going to continue to work in the decade ahead.

B (8:33)

Okay, so if we take the second first, the retail, are these companies that are, that have been like gps, that have been primarily institutional, that are now going to have to retool their business to start taking in retail, or more of these going to be like the new companies starting, are just going to start retail. It seems to be easier to maybe start there than have to shift the whole business. Like how you think about that?

A (8:57)

I think a lot of it depends on the, the product that the GP is offering to its investors. And you know, if you just stick within real estate as an example, you could be doing super high risk value add, like buying totally bombed out office buildings that are yielding nothing and do three years of redevelopment in the hopes that you're then going to go sell that thing at some multiple of the equity you put in it. But there's going to be no cash flow along the way. That's very analogous to venture investing. You put money into a company and you say goodbye to it for five or 10 years, you hope to make some significant multiple of it back. That type of product is really tough, a fit with what the retail investor is looking for. One of the things that's been proven to be really clear at this stage is that to reach the retail investor, you gotta be going through the wirehouses, you gotta be going through the broker dealers, you gotta be going through the RAAs. It's not really a story of going direct. You know, we saw that play out with all the crowdfunding companies and the Jobs act that was passed in 2013. And it's just not a very big market to try to go direct to the consumer. And it doesn't pencil out the money you have to spend on trying to acquire them. And it just, it's just too difficult. So you're going through the Advisor, you know, you're trying to get that Merrill lynch advisor to have the conversation with her or his client about why, you know, Ford's product in this category is, is a really great fit. And one of the things that's clear is those advisors want liquidity, right? They're used to putting clients in mutual funds and open end vehicles. And it's just really hard to have a conversation about locking up your money for 10 or 15 years. So you gotta have a product that offers liquidity. That's hard. With a lot of investing strategies, they just don't, some of them just don't support it. You gotta have registered funds, right, versus funds relying on exemption, which most GPs are traditionally done. So there's a more onerous reporting requirement. It's more expensive to operate. But what that has the advantage of is the Advisor can go offer that fund to any clients in any state. They're not worried about state level restrictions and blue sky restrictions and things like this. And so I think a lot of GPs are looking at this and going, and we have a lot of customers that have been hacking away at trying to attack the retail channel for years, spending tens of millions of dollars building distribution teams. These are customers that have like absolutely stellar brands within their asset classes, but they may not be globally recognized names because there's this third piece you really need which is you need air cover, you need marketing, you need a brand, you need the Advisor to have heard of you. I think where the industry goes is you sort of see this bifurcation where gps are going to have to make a choice. And the choice is, am I either going to go mega cap and get big, you know, or be rolled up and be, be part of something big and really just become a fee based asset manager where you're building these big expensive distribution arms. You are, you have many different products because once you spend the, the distribution cost to reach that advisor, you don't want to just sell them a multifamily fund, you want to sell them venture, you want to sell them private credit and you want to sell them every asset class you can and amortize that distribution cost. You need to be spending money on the marketing and everything else. Or do you become a sharpshooter? Because the law of large numbers is going to limit the mega strategy. It's just harder to make great returns on $100 billion than it is on 10, than it is on 1, than it is on 100 million. And so I think GP's kind of have to make this choice, you know, like, are we going to get big and get huge, or are we going to really focus on differentiated strategies where we are unique, where we offer sort of a compelling and different uncorrelated, whatever the characteristic is, type of return profile. And both of those camps are going to be super viable places to be. I think where it's going to be really difficult is you're mid size and your products aren't super differentiated. And you know, so I, I, that that's sort of how I see it playing out. And we've already seen a wave of consolidation hitting gps. And what's wild is when you look at the scale and the size with which the big GPS have gotten bigger over this cycle. You know, I'll give you an example. Blue Owl was formed in 2016 with I believe, $5 billion of seed capital. Right. And along the way they've merged with dial big GP stakes business, big private credit, direct lending business. They're now at 290 billion of AUM. So 5 to 290 over less than a decade. That just swamps the size of the entire early stage venture ecosystem across every manager, all concentrated in the capital raised by one manager. So there's this crazy power law in effect here where among kind of like the top five to seven managers that are attacking retail, they're basically raising all of the capital and then most of everybody else who's trying is kind of not doing very well. To be honest with you.

B (13:58)

You're so spot on. Obviously you see it. I couldn't have said it better. And the truth is there's only room for so many mega fundraisers. I mean you can't have hundreds of them.

A (14:09)

No, no.

B (14:11)

Here's Blackstone, kkr, Carlisle, I guess, Blue Owl. Now that's insane. Although I do see Blue Owls now attached to almost every big deal that's getting done.

A (14:21)

Yeah. So then you talk about that Apollo is another one that's grown a lot.

B (14:26)

Well, then you, but then there's only so many niche strategies.

A (14:30)

Yeah.

B (14:31)

So like there is a lot of people that are going to be stuck in the middle. Yeah.

A (14:36)

I mean, I think, and you know, we often, like I was in an event with some of our biggest customers and we were talking about this, like, well, what is the middle? Right. Like, what do we mean by middle? Like just put Aum Aum bands around it. It's sort of really hard to characterize what that means. But I think if you're running a strategy where you're like sub 10 billion, sub 20 billion AUM. And your fund sizes are whatever a billion dollar fund sizes or something like this. You can conceive of staying differentiated. You know, it depends on the sector, right? Early stage ventures different than real estate is different than every asset class. But you can conceive of like, wow, okay, I can intelligently deploy a billion dollars every two years, right. And stay within this profile and be disciplined and make my money from promote. Or you could be like, I'm going a hundred plus and. And. But the. That danger zone is right in the middle, right? Like, I'm a $25 billion manager. I can't quite afford the. You know, Mark Rowan of Apollo is on record saying they spent more than a billion dollars over three years in building their retail distribution team at Apollo. And, you know, so, like, you have to kind of pick which camp you're gonna. You're gonna be in, I think. But I. I'm actually pretty bullish that around the world, across all the opportunity sets, across all the different ways that society is gonna change, especially driven by this huge upheaval coming from AI, there's gonna be sort of countless pockets for this sharpshooter strategy and that we're really not going to be limited by that. And that it really. Actually the limitations would be more of a search and discovery problem. Like, if you have thousands of amazing managers out there pursuing these really differentiated, interesting strategies and you're an lp, how do you find them? How do you connect with them? How do you build trust with them? So I think there's a big opportunity there, but I'm actually pretty bullish for the sharpshooters out there because I think it'll be a good decade ahead.

B (16:35)

Okay, real quick, you said liquidity, and maybe Juniper is doing something or maybe they're not. I've always said, like, whoever creates the true secondary market for a lot of this stuff to where you actually have liquidity. So my. I guess I don't know if Juniper is building this. I don't know how you've thought about it. When I think of like, what is the greatest.

A (16:54)

What's the.

B (16:54)

What's something that would just fundamentally change forever is Even if the $50,000 LP could go find a way to transact on their interest relatively quickly in a way that's not cumbersome to the GP. I mean, we've had lots of LPs over the years come in, you know, we're getting a divorce or we're doing this, or can we trade out our A hundred thousand and it's just like not that simple, especially for a small gp, without a lot of resources. Have you guys even thought about how liquidity is going to be created if all these retail come in and you've got Morgan Stanley that wants the option to create liquidity, whether it's in a venture fund or a real estate fund.

A (17:35)

Yeah, we thought very deeply about this for quite some time now, because when we founded the company more than a decade ago now, my interest was always about making the private markets more efficient, making them more transparent, making them more like the public markets, right? Where you've got more efficient discovery, more efficient trading, lower cost, access to the really incredible returns that the industry generates. It wasn't about building investor relations software. Now we're a fund administrator too. It wasn't about that. Those were all waypoints or means to this sort of bigger vision that we're building towards. And so enabling our customers to offer liquidity is squarely in the bullseye of where we're going in terms of the vision. We are being very careful about it because. So there's some things that are different about the public markets and private markets that one really has to take into account when thinking about liquidity. One is that it's a repeat game, right? Like the great relationships between GPS and LPs last over decades. They're in every fund. And then there's this element of trust and sort of you're there in good times and bad, right? And this is especially true in real estate. Like sometimes there is the capital call or whatever. Sometimes one vintage is just the outlier and the otherwise great stream of a manager's track record. And so you really want to be deploying across all the vintages versus just picking and choosing. So it's a repeat game between GPs and LPs. And that's different than public markets, right? That's one element that's different. The second is there's this adverse selection bias at play where it's really hard to get into the good gps and it's really hard. Therefore, if it's not a good gp, it's probably going to be really hard to sell the position, right? Who's going to be the buyer? So you have a challenge there around that, that sort of fundamental bias. And so that's a challenge. A third one is that, you know, I have seen so many startups come and go over the last 12 years. I've been working on this general problem where they conflate and they confuse building a tool that would help facilitate liquidity. With a market mechanism of liquidity. So let me explain the difference. Like, there have been a lot of companies that are like, oh, we're going to just put the fund positions on the blockchain and then once they're on the blockchain, boom. Liquidity, right? And you're like, okay, like I see how, you know, having a distributed ledger can facilitate moving, you know, the records around a little easier. But like you're conflating a problem of records management with one of who is the buyer when you have everyone who wants to sell. And that's what makes liquidity really hard. And so a great example here is that like, look at Blackstone and their flagship real estate fund that they offer to the retail clients. And if you look at kind of when we first started going through this rate hike, they suffered a bit of a, I don't want to call it a run on the bank. Those are too strong of words. But they were facing a lot of outflows relative to inflows and they had to go do this deal with the UC system to provide this backstop that said, hey, we will fulfill redemptions and we got this $5 billion facility. I don't remember the exact amount of what it was to backstop that, to sort of settle the market and for people to say, okay, I'm going to be able to get my money out when I need it, so I'll keep it in there. And that's the really hard part about liquidity, is that you have a buyer when everyone wants to sell. So the actual problem to solve in liquidity is this two faceted problem. One is it is structurally a pain in the ass for the GP to facilitate transfers and to offer liquidity. It's a huge headache, right? It's not your business, it's not what you want to be doing. So you do have to reduce the paperwork and administrative burden. So there is a role for tools in that. And we're building all that infrastructure, right? Make it really easy, click a button, whatever. And that's not to say it's easy, but the actual harder problem is like, how do you actually have a functioning market system where buyers and sellers are where demand is clearing. And that's not just about putting stuff on the blockchain or having the tool available. You have to make a market. And so it's a much more complex problem. And we've been studying it, we've had teams that have looked at it. And what we have concluded basically is that we're building into the vision of creating all the tooling infrastructure, right? All the kyc, the AML doing, all the compliance, all the subscriptions, getting to a place where you can click a button to come into a fund, click a button to get out of the fund, all the queuing, the redemption, all of that infrastructure. Because you need that anyway if you're a fund administrator. And we administer thousands and thousands of funds, so we make this big investment in technology. And then there's the harder problem of figuring out how do you make the market and do you make that with partners and who has balance sheets and who wants to be buyers, right? And it's just one of the challenges is you just, you have this dynamic where it's very often the case that when everyone wants to sell, no one wants to buy. And when you have these really deep liquid markets like the US public equity markets, like there's always sort of a buyer at some price. But the problem you have in private markets, there's often you get these liquidity shocks like we've been through, where it's not, there is no market clearing price, there's no buyer, it's just, it'll go straight down to the bottom. So it's a really hard problem. But I think it will come. And our philosophy on it is that the GPS always have to stay in control. Like there's never going to be like a market of trading interests because of this relationship, repeat game, the trust, everything else. GPS want to know who their LPs are, they want to stay in control of that. So the whole system has to be architected around the idea that the GP is sort of the nexus point in the center. And the other thing too, Chris, is like you should be able to borrow if you're an lp, you should be able to borrow against your private markets positions. And the reason you can't today is because it's not efficient. It's way too costly for banks to underwrite that. Like how do we know if for capital's NAV that they're reporting reflects the underlying NAV of the. You just can't figure it out. You're not going to figure it out for a million dollar position. So there's another way you can offer liquidity to LPs too, which is you should, it should be the case that you can borrow against these illiquid private markets positions. And it's fundamentally, it's a data and underwriting cost problem. We think we're a pretty good position to solve that over time.

B (23:52)

It'll be fascinating. I mean, you know, somebody like in my world, you know, we're not huge. We managed at the top a little over a billion. You know, somebody comes in, they got a quarter million dollar redemption, then you got to go offer it to the rest of the LPs.

A (24:07)

Yeah.

B (24:07)

Which even if everybody takes their pro rata share, it's like not very much. It costs more to do the legal work than it. Then it's like, well, nobody really wants it. So then it comes on my desk and it's like, well, I don't want to look biased as a GP now that I'm qu quote unquote, the only buyer. So I pass purely because I don't want to look like I'm taking advantage of a situation.

A (24:28)

Right.

B (24:28)

Then it's like, okay, well then you LP can go ask somebody out in the open market to do it. And they're like, cool, but we don't know the gp, so we got to get to know them. And I'm like, yeah, I don't have time to spend time to convince somebody. So it's just like this never ending rat race that ends in just a bad situation. A lot of retail capital coming in. It'll be exciting to watch that all play out. Let's pivot to the second big fundamental change or opportunity, which is AI. And maybe you can take that question from the broader case. And then I do want to know what y' all are doing about it with your new product. Juni.

A (25:07)

Yeah. Okay, so I think at the broad macro, this should not be at all a surprise to any of your listeners, but I spend a lot of time with executives at our customers, and I can tell you everyone's at varying stages of acceptance of this reality. Right. You can hear something a dozen times before you really truly internalize it. What we see is that, let's say in venture, with Venture gps, there's a real realization that profound change is like not coming to us at the future, but is upon us right now because they're out hunting for these AI companies to invest in. Every day they're hearing pitches. Every day they're seeing how these startups that are AI native are doing things differently. And they just can't help but see, oh God, this is so much bigger than the shift to mobile telephony or from rich clients to cloud computing. And so with the exception of venture, they are at the bleeding edge of this. Most managers and most asset classes are somewhere between not yet interested, somewhat skeptical, or just not even paying attention, in my experience. And I think that's just profoundly misguided because if you look at the macro, it's basically the case that these, this transformer technology has taught us that there is mathematical structure to our language that can be represented in really big computers and that every, you know, every word relates to every other word in a mathematical way. And it turns out like that mathematical structure contains the insight of language and of communication. And since all of our knowledge as humans is represented in language, at the end of the day it all boils down to symbols somehow for us to communicate to each other. We don't, you know, maybe some of us communicate telepathically, but most of us, you know, we're writing stuff down and we're whether math, symbols or languages or whatever. It therefore is the case that basically all of knowledge, that all knowledge that humans have has this pattern to it that can be discovered by these machines, that can then be utilized through this sort of next token prediction methodology. And they can seem like magic to us. They can, it seems like a sentient being that you feel like you're communicating with another human being. The reason this is so important is because all knowledge work is going to be done by compute. All, all the knowledge work that we conceive of as, or let's just say almost all of what we conceive of as knowledge work today, which is to say Almost everything a GP does is going to be done by a computer 10 times to a hundred times better than a human can do it. And so that's just happening. That's going to happen. That's not a, that's not a controversial take. That is just what it is. And it's just a question of timing and sequencing and how long does that take and where's the resistance and where are there hard pockets and easy pockets? And what we're seeing is that there are certain classes of work that are much faster to be disrupted by this. One of the biggest impact areas that AI is hitting is software development, software engineering. And software engineering is working with programming languages, just like English or Spanish or French is a language. And so that is a really good fit with what these AIs are capable of. And so one of the first areas to go is software engineering. And we run a software engineering shop. We've got a couple hundred software engineers inside of Juniper Square. And so we're at the front lines of seeing how this is disrupting our own operation. And what you're seeing is that the entire job of being a software engineer is changing right in front of our eyes. From writing code to having the AI generate The code for you and steering these agents that are doing work on your behalf and then trying to solve problems like how do I know it wrote the right code? How do I evaluate that it's accurate? Right? So overnight right now, the job of being a software engineer is changing radically. The job of being a product manager is changing radically. The way you do prototyping, all that's like massively changing. We see other pockets like legal where AI is having a near and present impact because that summation use case of like let me read a million documents and, and reason on them or give you a summary of them is a really good fit for those industries or radiology or whatever. But that pattern is coming for every industry. The investment analyst inside of the gp, you know, the executive assistant, the, you know, everybody who's doing any work that is about managing processing knowledge. So like if it's a job that sits in front of a computer and knowledge, you know, information comes into that computer and the human is like doing something with their hands, right? And then sending that information off somewhere else that's gonna be done by a computer, you know, within the next five years without question. And the technology is there today to do the, to, to pretty much automate all of this work. That's the level of profundity of the change that this represents. And I think most people in the private markets industry, again with this exception of venture, they don't see it because you just, you go to like ChatGPT or something and you're like, huh, I'm going to try and load this like super complex financial model into it and see what it can do. And it's like, oh, it doesn't know financial modeling, like this is just a silly tool. But what you realize is actually most of the limitation of these LLMs is in the user. It's in your ability to prompt the LLM to do what you want it to do. And so what I urge our customers to think about when I talk to them is you have to really get smart about what is the human work inside of your organization and what is the computer work. And you're, today you're going to have a lot of humans doing computer work, right? But you're going to have to accept the fact that all of that work is going to move to computers over time. And if you don't do it, your competitors will. So you're not in a race against the AI, you're in a race against your competitor who's embracing the AI faster than you are. That said, there's a lot that the machines don't do, right? What these LLMs is doing are going out and building a network of all of this knowledge. They're relating all this knowledge to each other and then they're retrieving it for you and they're giving you answers based on that knowledge. But at the moment, they still aren't generating new knowledge. Right. And so only humans, as we, as far as we know right now, can generate new knowledge, new hypotheses about how the world should work, new ideas of what a great investing strategy would be. That kind of creation of knowledge is still only coming from humans, so that'll only become more and more and more important. And so I think people have to really get this mindset of like, okay, what is the human work that in my business that only humans are going to be able to do? And then really lean into that and embrace that. Because there's going to be a power law here, just like everywhere else, right Where. And it's, it's a sad thing for, I mean, long term, I don't think it's a sad thing for society because I think we'll think of other jobs for humans to do, but there's going to be massive disruption coming and it'll take time. So that's what's happening at like the biggest level. So, like, when I think about it at Juniper Square, I think, okay, every single job of Juniper Square, including my own, is going to be rewired around AI. And if, if I just tell our employees, accept the fact that there's always going to be a role for the human, right? So, so, so be, feel secure in the fact that the humans are still going to be needed in this future society. So hold onto that tightly, but hold onto very loosely the idea that you know what your job should be and accept the fact that you're just gonna have to grow and evolve what your job is as these machines take on more and more and more of it. And if you can stay in that kind of growth mindset, you're gonna do just fine. And if you have an organization of people that can really adopt that mindset, you're gonna do just fine. The problem is that's not human nature. Human nature is like, I spent 20 years to learn this craft, I know how to do this craft. This is how it's done. And the machine is coming for me to take my job. And you get defensive and reflexive about that. So every facet of being a GP is going to get rewired around AI and We have gps, Chris, who are recruiting, not many yet, but we have some that are recruiting for full time AI engineers to be working inside of their business. You know, we're offering a lot of this capabilities to our customers too. So you don't have to go out and hire software engineers to do this, but the ones that are out of the edge are doing this because they see the potential right now here and now to, to, to embrace these technologies. So I think that's like the big picture thing that's, that's happening.

B (33:54)

One, one thing you said is there the AI is you're going to have to get better at how you prompt it. But we also have to probably acknowledge also that the LLMs or the AI that you're talking to is going to have more context and is going to be better at having context of. I should have to prompt it actually less long term if the context of the AI is learning more about me. Fair.

A (34:17)

Yes. Yeah.

B (34:18)

So if I have an AI for Fort, it should know FORT pretty well. And so when I'm talking to it, it has years and years of context. Let's say we're just taking your typical gp. They've got some investors, principals, asset managers, some investment analysts, they've got accountants, they probably have data entry. Could you even like pick a job or two? And y' all have spent a lot of time thinking, if you're sitting there right now going, this couldn't possibly happen to me, or maybe it could, but I don't know how do you have like a prime example that even though maybe it hasn't happened today, you would almost be willing to go to the mat to say if you're an investment analyst, this is almost guaranteed going to happen to your job. Or is it still too early to imagine how the jobs might change?

A (35:08)

No, I think we can imagine how the jobs might change. I think where the error bars exist is putting a time window down to say by this point the way you are doing things today will be obsolete. That's where there's big error bars, right? So as an example of this, if you'll just take, you picked it. Let's just pick the investment analyst job, right? So again, this varies by asset class, right? But by and large we know what it is, right? Which is partner develops some interest in an asset, right? And let's be honest, inside of most GPs, at least the ones that I know, partner develops interest in an asset, partner has a napkin where they, you know, input the like six or seven numbers that they need to make the decision. Right. And then they make the decision largely based on their patterned experience over decades of doing this and them knowing what the critical variables are, and that's how the real work gets done. And then there's this giant exercise to make everyone feel good, including your LPs, including the rest of the IC and everything else, which is justifying the decision that you made. And that is, by and large, what a lot of investment analysts are doing. It's like, well, we're going to buy this company, we're going to buy this real estate asset, we're going to do this, whatever. Now go evaluate how they're positioned relative to their competitors. Evaluate what the market TAM is, evaluate, you know, and yeah, a computer operating on, you know, like five petaflops of compute is going to beat the smartest Harvard, you know, graduate every day of the week and twice on Sunday and happily working, by the way, on Sundays and around the clock and for 2 cents an hour kind of thing. So if you define the job of investment analyst as your job is to go produce all this rote analysis and then package that up into an investment memo for the committee to consume to make a decision on AI can already do that. Well, prompted AI can already do that, I would argue, almost better than any investment analyst, because you also have this ability with AI to be like, you know what? We really want to beef up this investment memo. Make it a hundred pages, you know, go. Actually, no, like, make it 10 pages. Write it in more colorful, flowery language. Write in a more precise language. It just, it turns the content that is produced into something that is functionally free. Right? So if you define the job of investment analysis, producing the content to support an investment decision, then that job is going away, you know, like, just as soon as you can really deeply adopt the AI into the process. Now, the thing to recognize is that what I like to tell our teams internally is a good mental model. A good way to think about AI today is that it's like 60 to 80% good, meaning it still hallucinates. It's, you know, getting better and better and better, but it's, it still hallucinates. It's very confidently tells you the, the wrong thing, right? And if you don't know better, you, you believe it, you get misled, you go in the wrong direction. It misses stuff. And so you have to look for the applications inside of your business where 60 to 80% is good enough. And don't try to, don't waste any time trying to get 60% to 100%, because just wait six months, you know, and then the next model update's going to come out and get you to 70% and 80%. And, you know, so there's companies going back like a couple years. We're trying to do fine tuning, you know, changing model weights and, you know, building out models into specific verticals. And that all got washed away by the pace of the general foundational model improvement. So just bet on the fact that the foundational models are going to continue to get better in near perfect competition, by the way, Amongst all the OpenAI and Meta and Anthropic and Grok and Gemini by Google, and bet on that technology curve accelerating, and then figure out where 60 to 80% is good enough in your business today and how the human can take the 60 to 80% and bring it to 100%. And that's the right mental model. And you supply that all over the business. So it's like, well, now the investment analyst is not producing the content because that's functionally free. The investment analyst is qaing the content, running all these bots, like, thinking of how do I write evaluations or evals to make sure that what's being produced by the model is accurate? Well, that's a totally different job. That's not going away anytime soon. So you can call them both investment analysts. And so that's what I mean by having this mindset of what can the computer do versus what can the human do? And I think what's so critical too is that there's just a period of learning and experimentation, and you kind of have to get into the learning curve to start to figure it out. So if you have this mindset of like, oh, the AI is not there, I'm going to wait two or three years. I'm going to wait till I see it so dominantly on display and my competitors or whatever else. Then by the time you actually pivot to taking it seriously, you're going to be two or three years behind everybody else that's been on the edge of the curve. So I just think it's. It's something you can't afford to not be have. If not your top priority, then it should certainly be the top three of anybody running a gp, in my opinion. Okay, so.

B (40:29)

So AI hits and you have a very successful technology business. And basically, I think it's fair to say it's kind of a broad mandate amongst most software companies. Like, you got to be. You got to implement AI.

A (40:45)

Yeah.

B (40:45)

So the question I want to know what Juni's doing. But I also want to more just hear from a CEO and how your mindset works. So this is all coming. Investors are probably calling, going, hey, what are you going to do about AI? Walk me just through how you thought about all this, what was necessary to.

A (41:02)

Build.

B (41:05)

Why you got to a point where you believed it not just because people told you AI Big, but you're like, no, it is big. And now I gotta almost not change your company, but evolve like you've told everybody else they need to evolve. Like, I'm more curious how your mind reacted to it because y' all just raised a lot of money. I do want to hear what it's doing, but I'm first curious, like how you began the process to put together an idea that you raised a lot of money around because everybody's doing it. And candidly, I think you're the first venture backed tech company that I've had on the podcast really, since this whole wave has started in the last year, year and a half.

A (41:41)

So, yeah, let me walk you through my just as a CEO, before we get into what Juni AI is for gps, let's just talk about that evolution. So I'm really into meditation. So I have a. Don't worry, I'm going to come back to it. It's not going to be a weird, random tangent. So I'm super into meditation. It's been this amazing tool for me as a CEO. It's been this amazing way to, I don't know, build an understanding of the universe, understand how things work by just sort of paying attention to what's happening internally. And so I go on silent meditation retreats. I spend a lot of time thinking about the mind and the senses and things like consciousness and the brain. And I'm interested in topics like neuroscience.

B (42:17)

And like, what is somatic programming?

A (42:20)

Sure. Yeah, exactly.

B (42:21)

So I did my research.

A (42:22)

Yeah, somatic. Somatic.

B (42:24)

Yeah, Somatic.

A (42:25)

Yeah, exactly, exactly. Somatic programming. Yeah.

B (42:27)

Yeah.

A (42:29)

And so I already conceived of the brain as a Bayesian active inference machine. That the way that we learn as humans is we come out of our mother's womb and we're curious and we have these sensors and eyes and ears and so forth, and we're paying attention and we are patterning learning. It takes a lot of tries, but eventually you learn how to stand up, you learn how to walk, you learn how to hold a cup without it falling out of your hand. And so you're just training your mind in the same way that we're training these models to build knowledge. And so I knew this about the brain and I intuitively understood it and I was interested in sort of frontiers of neuroscience and the difference between us as an active inference machine and our AIs at present as more passive inference machine is we are agents out in the world and we can go take action to do stuff, you know, so we can be like, I'm detecting thirst and I'm going to go walk over there to that coffee shop to buy a cup of coffee and satisfy my thirst. Right. Whereas an AI right now does. You know, we don't have any humanoid robots out there at scale yet, but they're not out taking agency in the world. They're constrained to an experience on these silicon transistors in our data centers. I think that will change, by the way. But so for me, as soon as I kind of groked what, not to use that word, understood this parallel, it was like this aha moment for me of oh shit. Holy cow. Everything we do is, you know, like these computers, these LLMs are working in the same way our brains work, only it's a 5 petaflop data center instead of the, you know, 86 billion neurons in the human brain. Oh my God. Right like this. So I was already tuned to that because of this experience of meditation. Now at the same time then, you know, we're using these tools internally, we're in the tech industry, so we're kind of early adopters as they're coming out. And you know, like, as amazing as the launch of ChatGPT was, which really kind of woke all of us up to this, you know, transformer technology that was developed at Google, by the way, not at, not at OpenHAB. It was an, both an amazing experience and comically limited. Right. Like if you remember, when ChatGPT first came out, it couldn't search the web. You were, you were inter interacting with knowledge that was like, you know, two years old because that's. It was trained on the Internet from two years ago. Still powerful, but you're not like worried for your job. And so then the next breakthrough for me was, I think it was ChatGPT's O3 model, was their reasoning model. Okay. That was the moment where I was like, oh, it's going to totally transform our company and every industry and everything else. So that was basically the ability for the model to teach itself how to think about new and novel problems. Okay? So if you ever see it thinking, like you type in a Chat GPT and it's like thinking or it's explaining its reasoning to you, it's like, oh, Chris has asked me a question like size the, you know, how many ping pong balls can fit in a 747, you know, some cheesy like consultant answer like that. It will teach itself how to answer that question. It won't just say, I think the answer is this spike because I've connected ping pong balls in 747 and my next predictive token says it'll be whatever it says. Ah, the user wants this thing. And I'm going to go build a chain of thought to analyze this problem set like a McKinsey analyst would and give you the McKinsey analyst answer. And what the model companies were doing, like what OpenAI and these foundational model companies were doing around that time that reasoning came out is like the wave one for these models is going out and traversing all of the Internet, all of the publicly recorded knowledge, and as we know, probably some of the not publicly recorded knowledge too, and building a neural net of that knowledge. But eventually you sort of consume all the knowledge that's in textbooks, all the knowledge that's in public web pages. And really what you want to get at is the knowledge that's locked up in people's brains. Like when I asked Chris about an industrial deal in Texas, what is his thought process? Like, what goes on inside his mind, patterned from decades of experience. And so then the foundational model company switched to hiring literally thousands and thousands of experts who math Olympiads, McKinsey analysts, you name it, who were then training the models, how the best humans in all these different fields approach problems. And so you could get this recursive runaway loop where the models could start to train themselves how to improve themselves, how to, you know, become smarter, become better. So for me, that O3 moment was the, ah, this is going to totally change everything at our company. And then like any, I think anybody in my position, leading companies or leading organizations, it's humbling how hard it is to get people to pay attention. And it's humbling how hard it is to get people to change. Like, it's just amazing how humans are just kind of hardwired to do what they've always done. So then there's a period of like, how do I get the company to really pay attention to this and take it as seriously as I'm taking it, you know, and you're yelling and you're pounding the table and every opportunity till I'm blue in the face, I'm like, yeah, but what about AI? What are we doing it with? AI over here. What about AI in our billing process? What about AI in our customer support process? Then I had a moment this spring where I said, okay, I'm not worried about this being a fad. I'm not worried about over rotating on AI. The risk of not moving fast enough is far greater than the risk of over rotation. That I have to do everything I can to get the company to embrace this. Remember Top Gun, the need for speed? It's like we all need the need for speed. Because that's actually, I think what's going to matter that everyone will have access to the same technology. I don't think there are any moats in data. So GPS that are out there, like, well, I've got my special sauce in data. I don't think there's any moat there. I think the models wash that moat away. So then the actual moat that GPS build is how quickly can you go from next breakthrough in model technology to figuring out how that's going to impact your business process, to figuring out how you can implement it in your business to your advantage and then changing human behavior to implement that loop and taking advantage of it. And the faster you can spin that wheel, the more competitive advantage you will build. And so, yeah, I had a moment. We had all leaders, top 75 leaders at the company globally. We gathered them in San Francisco for a week. And I said, we're going all in on AI. Every leader at the company has an imperative to come to me with a plan for how you're going to embrace AI in your business function. And then we are going to have a bunch of expectations of not just leaders, but every employee at the company that if you're not actively using AI to redefine your job, then you shouldn't be here. Because this is going to be an organization of people that are on this vector of, of change. Now we're in the process of incorporating AI competencies into all of our job performance frameworks. Every job at the company, and we have hundreds of them, different types. We'll have certain AI competencies that we're looking for. We're ensuring our performance is taking into account our performance evaluation is taking into account AI, and then we're evaluating for AI competency as we hire people. So if by now, at this point in time, you are not as a candidate, you can't sit down with a prospective manager and talk about how AI has transformed your life and the zillion different ways that you are using it, then you shouldn't be at our company. And then what we're doing is we are, you know, Rome wasn't built in a day, so we're, we sort of imagine like taking all the business processes that happen inside of your organization and cataloging them. Then imagine Stack ranking them for how well aligned to the current AI technology they are for rewiring and disruption. And then we're just patiently trotting through those business processes, trying to take a few of them a quarter, not trying to reimagine 2000 different processes overnight. Pick the high impact ones first and you build a muscle. I think the biggest, hardest factor for organizations is getting all of your team and most gp. You know, we have almost a thousand employees, so most GPs don't have, many do. But most GPs don't have the scale of employee base that we have. So it's easier when you're smaller. But it's getting all of your employees to adopt this growth mindset of being excited about the change and leaning into it and being curious versus this defensive mindset of AIs coming from my job and what am I going to do? Because every, you know, you see it in different pockets. But like if you work in support, like Mark Benioff's on the record, Salesforce CEO is on the record just like a few weeks ago saying that they fired 4,000 people out of their 9,000 person support team because of how well agents are now fielding these lower level, tier one and tier two tickets that are coming in. So now imagine you're one of the 5,000 employees left on the support team, right? And you're looking at this going, wait, the model's getting better at a compounding rate. What if the support team is going to be 900 people next? What if it's 90 people? What if it's 50 people overseeing a thousand different agents? Right? And so what can happen if you're not careful is you sort of devolve into this, you know, the opposite of an abundance mindset of scarcity. That's I think the hardest leadership challenge is how do you get people optimistic, skating toward a future when it takes so much effort to rewind, to rewire it and to write it. And human nature is to sort of start to fear for one's safety and livelihood. And I, I, I will admit I don't know that the answers, but I say that is the leadership challenge.

B (52:27)

Okay, and then how did you think about the product that you were going to build for customers? I think everybody's figuring out what their product is going to be for their customers. You have a lot of them. You have a lot of data on GPS in the private investing world. So clearly there was probably multiple things you could have created, but you had to start with, you agreed what you're doing at your own company, the most important problems. So what are you tackling first? And as a customer, what should I expect? And like you said, it's probably stuff y' all will do great. But then over time, like other companies similar to y' all might have to do the same thing. But, like, what are the biggest problems and what are y' all offering?

A (53:10)

The. The way to think about Juni AI, which is sort of the umbrella term for our entire AI platform. So we have a, you know, a big and sprawling technology platform now. We support everybody from CBRE to two guys in a garage across all these different asset classes. As you said, it's trillions of dollars, It's. It's tens of thousands, it's millions of millions of positions, millions and millions of K1s going out every year. And so it touches a lot of different areas of a GP's business, from compliance to investor onboarding, to CRM for investor relationships, to accounting to reporting. You name it. There's a wide surface area, but it all relates to running and operating a fund for a gp. Okay. And so when we looked at this, we said, okay, well, obviously, one of the really great things about what do we love about ChatGPT or, you know, whatever. Pick. Pick any model. It's not chatgpt. What do we love about LLMs? It's that you can have an interface to the computer that is your natural spoken language. One of the challenges with computers originally was, like, if you wanted to interface with them, you had to learn to program, you had to learn to code, you had to learn this, like, extremely complex language with all this really weird syntax and stuff like that. So one of the huge breakthroughs is this idea that you can communicate with the computer like I'm communicating to you right now, just using my normal words that I use to talk to any human. And I could just like, tell the computer to do stuff in those words and magically it gets done. So we said, okay, the entire interface to all of these different aspects of our application. Think just like a CFO who wants to know the cash balance in various entities stored somewhere deep down in the GL that we're managing for them. Right. Obviously, that type of question can be answered in Juniper Square today, we have these extremely powerful dashboards. You know, you can, like, get in there and, like, freaking. You know, it's like a 747 cockpit you can configure to your heart's content. But most people don't want that. You know, most people are like, I want to just type into a search bar into a chat interface, what's my ending cash balance for all the entities related to Fund 7. And then pop out comes an answer. So one thing that Juni AI is is pivoting the application to natural language interface for the user. So every facet of the application is now going to connect to a chat interface where you can interface with the computer now in a natural language. And so this extends to mobile and extends to, you know, I just had a really great meeting with blah blah, blah lp, right? And so one of the things that we're shipping in October actually is a meeting recorder connected to our CRM. So now you're in a meeting, you can have your mobile device, you could be on a zoom call or whatever. You record the meeting. The transcripts recorded, you know, the audio video is recorded into a transcript. Transcripts embedded into the CRM records automatically. That transcript is ragged. We just extract the semantic understanding from that transcript prompt. You like, hey, Chris, it looks like you learned 10 new interesting things about this LP. Here they are. Do you want to update the CRM record with that? You're on your phone. Yeah, that sounds great. Go ahead and do that. Right. So the idea is turning the interface of very powerful, complicated software into the type of relationship you have with your team member. Okay. So that's one element of this. The second element of it is agents. So our view is that pretty much all of knowledge work is going to shift to agents doing it on your behalf. You can have an investor relations agent, you're going to have a portfolio management agent, you're going to have a cash management agent. Everything that happens inside the GP is going to have an agent. And probably one of the most complicated, complex agents you'll have that we have to be careful of, but we're working on is you'll have an agent that represents you to the lp. Any question the LP wants to ask about you and your business, right? There'll be like a PhD level understanding of you and your track record. And LP can ask that at three in the morning if they want. They can, you know, so. So every facet of work inside of the GP is going to have agents that do it. So part of Junior AI is building the investor relations support agent, the portfolio support agent, the agent that helps CFOs oversee their fund administrator. Right? So stripping down all the work. The other piece of it is that we operate in a financial services highly regulated and mission critical industry where there's zero tolerance for errors, right? Like one plus one always has to equal two in the financial reports. And that is a poor match with the fundamental way that LLMs work. They're probabilistic. This next token prediction is probabilistic. So there's a big job to do of taking these foundational models which are probabilistic in nature and sort of building them into these high accuracy, zero tolerance for error use cases like financial reporting where you're leveraging the power of the model but then you're running deterministic code over the top of it to make sure that one plus one always equals two. And we do that as our most important job for the gp. Because if you just go deploy, you know, if you're just like, I'm just going to deploy Gemini across my business or whatever, it's going to start doing my financial reporting. It's going to be a disaster, absolute disaster, right? So that's not how it's going to work. You're going to need a company to build the scaffolding for you to ensure the accuracy. And then the last piece that's, that's really important that I think a lot of people miss is that for AI to do something on your behalf, it has to connect to tools that will do stuff. You just deploy chat GPT inside of fort and it's not going to move money for you unless it has access to FedNow and the ACH system and the Swift network and everything else. Well, an AI just doesn't go get that access. Those are regulated entities. You gotta be, you know, you have to have the bit code, you have to, you know, yourself be connected. So AI to do stuff useful inside of a business has to be able to connect to underlying tools that actually do the work. And so our job is to marry, to build those. There's hundreds of them. Everything from cash reconciliation to reporting to CRM note taking, you name it, hundreds and hundreds of tools that are actually deterministic code, meaning like, like traditional software code. And then there's this agentic and sort of natural language interface that runs over the top that gives the human this experience of I'm just talking to a computer and then it's magically getting stuff done. So it's, it's a really big, it's almost like a new operating system layer that you're building. It's a really big investment for us. It's rethinking the entire platform to be really agent first. It'll enable what GPs need and I think we'll see in a future. I don't think it'll be crazy to have two GPs managing $10 billion with one or two people and then these massive fleets of agents. And you get these people that are highly specialized in overseeing agents. Of course, there's a big question, what does everyone else do in this future state that we're going to have to answer as a society? It's not a, it's not a trivial thing.

B (60:20)

That is a perfect way to put a bow on this great conversation. Thank you for the time today, Alex.

A (60:28)

Yeah, my pleasure, Chris. Always great to chat.