Fund Administration: AI’s Growing Impact on Fund Services

Speaker A: Foreign.

Speaker B: Hello and welcome to Deciphered the Fintech podcast from Bain and Company. I'm, um, your host, Mike Cashman. And today we're exploring a quiet revolution in one of the corners of financial services, fund administration. It's a space traditionally defined by complexity, regulation and manual workflows. But with AI, all that might be about to change. Joining me today is Alex Robinson, CEO of Juniper Square, a leading provider of fund administration software and services. From AI powered agent teams to automating investor statements, Juniper Square is at the forefront of this shift. Alex, it's great to see you again. Would you take a moment to introduce yourself?

Speaker A: Hey, Mike, thanks for having me on. Yeah. So I'm Alex. I'm the CEO and co founder of Juniper Square. I've been at this, gosh, 11 plus years now and uh, Juniper Square is my third startup. So I'm kind of an entrepreneur by addiction or by lack of alternatives. Maybe one of the one or the other. And I live in Northern California, Marin county, kind of San Francisco Bay Area, father of three kids.

Speaker B: Welcome, Alex. Uh, we are thrilled to have you on. So today we're going to talk about transformation of this industry with AI and will it happen and when it happens, and what might be some of the intermediate outcomes for different providers in the value chain. In, in the spirit of Bain, we sometimes lead off with what we call an answer first. And so would love for you to just kick us off with your 30 second take on will AI fundamentally transform fund administration and services, or do we think this is going to be more of a cautious evolution over time?

Speaker A: Well, I think at the limit, absolutely, yes. There's no question. I mean, AI will transform all knowledge work by definition. So language, as it turns out, has a mathematical structure. And you know, we as humans utilize this when we communicate without really understanding it. We're kind of doing next token prediction when we're communicating and deciding what to say next. And, and it just turns out if you've got computers that are big enough and data sets that are big enough, then this relationship becomes apparent and it's become something that you can utilize and leverage to get work done. And so because all knowledge work is based on language or based on programming languages or math or things themselves that have structure, this, the same concept over time at the limit will roll through and affect all knowledge work. Everything that the humans do with knowledge work that isn't about some of these fundamental questions, like where does taste or judgment or intuition come from? I mean, we can get into those much deeper topics if you want I'm super into meditation and I go on meditation retreats and I. So I saw this intersection of AI and computation and consciousness and how the brain works is an area that I'm passionate about. But as it practically relates to the type of knowledge work, let's say that's done in fund administration or pretty much any job inside of a company at the limit. Absolutely. Uh, it will all be transformed and it will all be better done by these superhuman machines. That doesn't mean that there won't be a role for humans, it's just that it will evolve and it will change and the types of jobs that we do will change but it won't be what we're doing today. So I do think that that takes time. So it's not like tomorrow all of a sudden we're going to be at the beach with our feet kicked up and AI is going to be doing all the fund administration in the world, but at the limit.

Speaker B: Absolutely super evocative maybe to help ground some of our listeners and take a quick step back. When you think about fund administration and helping set the stage of where Jennifer Square is today. It's a piece of what you do, not uh, all what you do. Would you give us the background on Jennifer and then we would love to start to step through the conversation in a bit on where do we think more technology and more automation is going to creep into the value chain first?

Speaker A: So Juniper Square is an 11 year old technology and services company. We like to call ourselves a technology native fund administrator. So what I mean by uh, that we were a tech company before we became a fund administration services company and we are both. So we provide almost 2,500 GPs with some combination of software and services and help with their data and that's across all asset classes. So real estate, private credit, private equity, venture capital, crypto, farmland, real assets, you know all the food groups, all the private food groups. Um, we are managing close to 40,000 active funds on our platform north of 600,000 unique investors log in to the LP side of the portal north of a million positions, trillions of dollars of capital. So pretty good scale. All organized around a single common industry data model which we can get to why that's important when we start to talk about AI. So we've been in the business of helping private markets GPs with technology and services solutions for more than a decade now. Uh and the basic idea is our belief is that GP should focus on investing, they should focus on relationships, they should focus on their edge and what creates value in the world for, for them and their investors, doing the accounting, creating the investor statements, tracking the compliance. These are things that have to be done, but they're not where a GP has edge. So we want to help GPS with everything. That's the operational heavy lifting of raising the capital for a fund, running and operating it, managing it in whatever capacity they need, whether that's technology from us, fully outsourced fund administration services from us, data warehousing and solution support from us, or some combination of all of those things so that they could focus on what matters most, the investing and the human relationships.

Speaker B: Super Hawaii, this idea of a tech forward fund administration business. I want to pull on that thread a little bit, Alex, partly because one of the knocks on the broader fund admin market has been it's historically heavily human intermediated.

Speaker A: Right?

Speaker B: This is a very, very manual business. There's a lot of nooks and crannies to this. And as you think about some of the applications of automation, what does it mean to be more tech forward? Like, how does that feel for some of your customers, for some of the market participants, and what's really different about that in the market today?

Speaker A: I think it's maybe helpful just to kind of give you some practical examples of how that's different. So we could talk about something like the investor experience where, you know, if you're on a tech platform like Juniper Square that has a administrator providing services over the top of it, your investors especially. This is becoming relevant as more and more managers think about the retail channel, how to grow their investor base through the wealth and uh, retail channels. You've got to provide a digital interface for those customers to get their reporting. They're going to expect to do things like manage their address with you and you know, add their spouse or their accountant and you know, they're going to expect something that's like Charles Schwab or Fidelity or Robinhood or you name it, from the public markets, from you as the private markets manager. That's not what the institutional LP expected from the 90s or the 2000s. So there's this entire digital infrastructure that you have to provide to the lp where the mark that they have in their mind is when I trade in the public markets, I just click buttons and things happen. Paperwork gets done, money gets transferred, accounts get settled, custody and ownership gets recorded and transferred. And everyone takes it for granted that these really complex workflows and operational processes that involve the movement of money, the movement of information, really high stakes tracking and recording of ownership, staying compliant with regulators, these are processes that in the private markets industry have really evolved to be very paper based, very manual, involve lots of people, third party law firms, et cetera. And the entire industry needs to move toward something that looks much more like the public markets experience for the investor. I click a button and I effectively make a trade and everything gets settled. Um, so that LP experience of a digital interface for investing, for reporting, for managing information with the GP is becoming table stakes. And it's something that you get with us that if you're just a traditional fund administrator that provides accounting services, you're not going to get on the GP side. The thing that we hear again and again and again from our customers and of course this is only becoming, you know, a louder point of feedback with this latest theme of this latest technology of AI entering the scene is where is my data? I want access to my data. I want, you know, like you're the fund admin, you've got all the books and records. That's great that you're doing that work for me, but I need the data. The data is mine and the data is part of what gives the GP edge. The really great GPS are uh, utilizing this data to be smarter than their competitor in the market. And so if that data is locked up inside of the fund admin or it's being delivered to you as a spreadsheet attachment to an email once a quarter, that's not good. You know, like you want a data warehousing solution, you want a BI interface over that data warehousing solution, maybe you want an API interface to that data. And ah, so there needs to be a technology platform that underpins the work of the fund admin that gives the GP access to the data so that they can utilize it to advantage in their business. And again, that's something you get with us that you don't get if you're just a traditional fund administrator. And then the last thing we like to call the like where's my pizza? Example. Okay, so you go to Doordash and you order a pizza. You got real time feedback on exactly where that is in the process. Right? Okay. The order's been placed, it's with the restaurant is getting cooked, the driver's picked it up, the driver's waiting for it, it's on its way to you, it's going to be here at 8:22pm, right? Right. Historically in the world of fund administration, the work that's happening is happening in a black box and the customer has no idea what's going on. It's not until one minute before the reporting deadline that a package gets dumped. Hopefully, you know, you meet the deadline, the package gets dumped on you. And so customers should be able to have that entire workflow be online in a structured system, be auditable. And then what that unlocks is a lot of these workflows involve the customer. Okay, so without going too deep, I'll just kind of give you an example here. One of the things you have to do as an admin is you have to keep the books and records for the fund. That means that every month you need to record all the cash transactions and do the cash reconciliation so that the bank account balance matches the books and records, you know, for the fund. Pretty basic accounting activity. Well, sometimes the admin has the knowledge to know what this transaction was and we can make a guess. But at the end of the day it's the cfo, it's the controller, it's the people inside the GP who are spending the money, who are doing the transactions, who have the information. And so if that is a spreadsheet based, email based process to go back and forth, that is misery. You know, here's 50 transactions this month, you need to go through and categorize all of them. If that workflow is happening in a structured digital system, two things can happen. One, you can train AI models to classify those transactions for you and get really smart at that. We could talk more about that. And two, anytime you do need information from the customer, you've got a digital interface for providing it. And now you've got a whole record of that interface, you've got the historical transaction, you can learn from it. So it's data, it's workflow and it's an investor experience. All of those things being online in a technology solution, that's what you get from us that you don't get from a traditional admin.

Speaker B: That's super helpful and I want to try on an analogy for you what we saw with consumer banking 20 years ago with Digital Insight and all of the enablers to access the core ledger for consumers. This feels like this is the start of that for a much broader set of folks in both on the GP side and LP side and as we go down the democratization of private alts into wealth, it's going to be more and more important to get those portals right. Does that resonate?

Speaker A: It very much does. I think what we are trying to do simplistically is provide direct real time access to the underlying ledger. And if you can do that, then you can power all kinds of Workflow, you can train AI models, you can give a consumer investor real time visibility into balances, exposures, et cetera, right? And what's hard in the private markets industry is that, number one, there's a lot of complexity in this ledger, right? So you might say like, oh, this is, uh, Bain's Fund 10. Well, Fund 10 is not like one legal entity with one QuickBooks account. Come on. This is like there's going to be hundreds of different legal entities, probably thousands of bank accounts, right? Like many different ledgers and sub ledgers. And the way you guys do things in Brazil is going to be different from how you do it in Lux is going to be different from Singapore. Like massive complexity here. So you have to deal with that complexity and then you also have to deal with the fact that in the private markets industry you have these different value chains, right? So in real estate, for example, it's very common for a manager to outsource asset and property management to a third party. Well, so now the ledger for that asset actually sits with a third party, doesn't even sit with the manager. So now you have this data aggregation, probably. But the central thread that we're tugging on is the state you should get to is there should be a record of ownership that's live and current and real time. And that record of ownership should be shared by all the parties, the lp, the gp, the administrator, the regulator. The GP should be in control of that because they're the sponsor of the fund, they're responsible. But if you have that central concept in mind, then all of the workflows that you power, all these great experiences that you want to create for the investor, all the edge you want to get on decision making as a GP get easier. And that's the central thread of the problem that you tug on. And then you just deal with the complexity that I talked about. And that's what makes it a hard problem that takes decades to solve versus, like just something you whip up over a weekend.

Speaker B: Two threads I want to pull on after that. Uh, one is where do you think if I step back on all of fund services, investor services, there's a lot of repetitive tasks that happen in the background, both at the, uh, fund level as well as queries that are happening sort of on a recurring basis across this. Where do you think automation will go first? And then where do you think it will have the biggest impact? And I'm almost segregating. You know, we're in this piece in corporate America where we've seen a lot of AI Experimentation and we're just starting to see some of them actually really pay off. And so curious your take on what that will happen here.

Speaker A: Well, the organizing product philosophy that we are pursuing now, well, there's a lot of threads we could tug on there. But as it relates to your question, is that start with a presumption that AI is. And when I say AI, I guess I'm talking kind of broadly about classifiers and so automation more broadly. But the big breakthrough here is what can these models do and how quickly they're advancing and so forth. But start with the presumption that there's 60% good and figure out where 60% good is good enough and start there. And that's a much smarter way to start than to say I'm going to take something that's 60% good and I'm going to whack away at it and try and hack it and build all these sort of like meta solutions in order to get it to 100% good requirement. So what do I mean by that practically, um, human in the loop scenarios, 60% good is fine. Right. So let me give you like a practical example that's live in our product. Today. RGPS raise capital globally through all kinds of channels. They have direct institutional teams that are going flying to the Middle east to meet with sovereign wealth funds. They have wealth focused teams that are distributing, they raise capital through wirehouses and intermediaries, they use placement agents, they use some of these aggregators, all kinds of different services. So they have subscriptions coming at them from LPs in all kinds of different ways. Hopefully they are able to use online subscription tools like ours and from others whenever they can. But practically speaking, in any fund, any large fundraiser, you're going to have hundreds of paper subscription documents where because the industry's got this antiquated history, that's your record of information that's coming at you. So you got to deal with the paper or you know, PDF of the paper. So in our product is the ability to again via chat interface, add that paper. Again. Our fund admins do this or if the customer wants to use our technology themselves, they can do this directly. Add the subscription document into the platform. We have a model that will go read that document. It will extract all of the information from the document that we need to store in the relational database. So what's the investor's name, what's their address, what's their Social Security number, text ID number and then expand there to hundreds of different data points. And then the model will make its best guess to extract the information from the document. It'll provide the telemetry of exactly where in the document we got that information. And then there's a software interface for the human who's in that loop to go review eyeball and be like, yep, actually this says Mike Cashman. Subscription agreement says Mike Cashman. We're good. So that's a case where like as long as the model is 60% accurate, that's fine. It's a huge time savings, labor savings, job satisfaction improvement for the human who's in that reviewer loop and you have a control in the system to make sure that the data is accurate. Perfect application of AI and automation. If you try to take that same use case and say, what do I have to do to get to six sigma 100% accuracy every single time with just the AI doing that work, you'd be working out that problem for the next year and a half. Right. And so that's the kind of organizing principle is where you have human in the loop is where 60% accuracy is okay. And that's where you're going to see really broad based adoption of what's happening with AI is you're just turbocharging software engineers capability to build these things faster. So now all of a sudden that subscription document extraction and review tool can be done in a Sprint, a two week engineering squad Sprint. And that might have taken before six months or something like that. So now you can build a lot more of those tools a lot faster. And so AI paradoxically at least as it relates to fund administration. The way to think about it is less that, oh, you have this like super intelligent general model that just comes in and does everything. And the way to think of it is you have a constellation of many, many different models that are all tuned for different use cases. Humans are in the loop for the next few years, but in a uh, way that 10x is their productivity hopefully. And what's really happened is you have increased the productivity of the software engineer building all that software. Maybe not in order of magnitude yet, maybe 2-3x right now, but where it's headed I think is an order of magnitude. So this is why you come back to what does it mean to be a tech company doing fund administration. It's like if you don't know how to take advantage of that, then you are going to be somebody's customer. You know, you're going to be buying solutions from someone and it's going to be tough to compete.

Speaker B: It's super helpful, Alex. You've talked a little bit in the past about the difference between probabilistic and deterministic models and what that means, especially in fund admin. You know it's easy to see for some of our customer examples of solving, uh, easier consumer facing customer service call center type applications. But when you have to be right all of the time, every time it's a little bit different. What do you think you're going to see or look for as the tipping point for when AI is going to get beyond the need to be 60% right and we're going to go past that. Is there a breakthrough or is there just going to be a sort of a gradual growth to get there?

Speaker A: I don't know. I think that like many who are on the front lines of this, I am shocked and uh, continually shocked by how fast the foundational models are improving. I mean it was just last fall that reasoning was introduced. I think it was last fall, it was like last September, so under a year ago with O1 and 01 mini from OpenAI. So we've had reasoning for less than a year. And that breakthrough has been foundational because now you can train a model on the chain of thought and the process that a human would take in any situation, right? Like, oh, let me tell you model how I would approach this problem and let me document that. And so then what matters is do you have the, because the foundational models now have that capability, the question, the value shifts to who can provide the context and the training and the chain of thought and who has the expertise to be able to take this foundational capability and uh, kind of build the scaffolding to take it up into the vertical use cases. Which is why I'm so bullish about Juniper Square's position here because it's about having the vertical domain understanding. So I would not underestimate, you know, I would just be in a posture of being ready to be shocked and surprised by the speed, velocity and the compounding effect of the improvement of the foundational models. And so therefore I wouldn't be surprised if to me the tipping point what would be the fundamental change is right now we're pursuing a product strategy that is a kind of constellation of models type strategy where the humans are the architects of all of these different use cases and training and the chain of thought and the rag and the reasoning and everything that different models for different use cases, ability to workflow and it's still humans that are at the center of that. If we get an ASI type breakthrough in the models then you could imagine a future state where you're just sort of vibe coding the AI to be like, hey, you're now running a fund administrator and go build all of the software that you need, including deterministic tooling to go execute this. So it might be worth unpacking what I mean by deterministic tooling, because I think this is not so obvious to everybody. When you're interacting with ChatGPT and you're asking it a question, it's doing next token prediction on the words that you have the tokens, you know, it breaks down the words into tokens and then uh, it, it guesses probabilistically based on the input you've provided and how it's been trained and, and so forth, what to say next, how to respond. Right. And so that, that is fundamentally probabilistic. If you go ask it again, it's going to reload, it's not going to have, it's got memory now and it's got improvements in this capability, but each time it's running that computation anew and so it can get a different answer. And so we've all experienced what it's like to have a model hallucinate. Hopefully, if you're paying attention, if you haven't, you need to learn to pay attention. Because the thing about these models is like, they are very convincing, so you have to be very astute and pay really close attention to when you're being misled. And this is like a real skill for people to develop. Otherwise it just, it seems so convincing, especially if you're not an expert, an area. So it's hard to know when they're hallucinating, but they certainly do and they certainly get it wrong. And they certainly don't do math. They're doing math on next token prediction, on pattern matching, they're not doing math on knowing that there's a real number called 2 and another one called 2 and you add them together and it must equal 4 based on the logic structure of how the logic of math works. Right. And so what we do at Juniper Square, and what customers will require companies that are providing them AI solutions to do is to say, what are the situations where we need to go call a deterministic model or deterministic code. So I'll give you a really great example. One of the core things we do as a fund administrator, and you know, this matters to GPS a lot because it's how they get paid, is calculate the waterfall. So what's the waterfall? It's the logic structure, the logic Tree for who gets paid what. And it can get very, very, very complex. But it's basically, you know, there might be a preferred return, there might be a hurdle, and then there's a split between the GP and the lp. And you do not want a probabilistic foundational model doing that kind of math. It's a terrible idea. And so instead what you want to be able to do is you want to have that math represented in a deterministic language like Python, let's say, where you can say, all right, for Bain Fund 10, again, remember the 60% use case? So let's like walk it through. So like Bain Fund 10 comes on board and we've got a limited partner agreement and an operating agreement and all of these documents that have language inside of them, legal language that describes the waterfall tree and the waterfall structure. So what's this process look like? So, 60% use case, right? The computer goes through and it makes the best guess and says, I'm pretty sure that this is how we should design the code in Python. The AI generates that to represent Bain's fun 10 waterfall. And now a human software engineer who's an expert in understanding fund waterfalls. And again, part of what we do is like build the interface for our expert accounting folks who don't know software engineering to sort of provide this bridge says, yep, this looks correct. This is how you calculate Bain's fun 10 waterfall. And then you run some like, evaluation test cases to make sure it's correct. And now you've got a deterministic bit of code running in Python called Bain 10's waterfall. Okay, so now anytime, like Bain CFO or Bain's controller or anyone has questions, maybe they would say like, hey, if I sell this asset tomorrow at uh, 2X, what's that going to do to the expected IRR of the fund? And that's a question that a user might want to ask of the AI via a, uh, chat interface. And it's just going to expect the AI to give it an answer. Well, the way that is going to happen is the AI is going to go, ah, uh, okay, this is a question about the waterfall. I know to go call the deterministic tool. I'm not going to answer that via next token prediction, I'm going to go call that deterministic tool. I'm going to run the Python code with the inputs of the what if scenario and I'm going to output it back to the user all via this conversational interface. And what you have is this blend of generative AI and the conversational interface that customers expect and the deterministic tooling that's running over infrastructure that has the data and the workflow and the foundation to get the customer the answer they need. And for the next few years, what I see, here's a great example. Um, Andrej Karpathy is this thinker who I think everyone should follow on AI. He's been very prescient and he's an amazing Communicator and his YouTube videos on AI are really, really good. He was the um, chief engineer responsible for Tesla's AI autopilot. And he has this one video where he talks about going for a waymo self driving car ride in like 2013 or 2014 in Palo Alto. And it was a flawless, perfect ride. And he's like, oh my God, self driving is a solved problem. And here we are in 2025. I just took my daughters for a Waymo ride in LA a couple weekends ago and we're seeing it roll out and it's an amazing experience but it is very limited still in its use case of course. Tesla launched their self driving solution in Austin not long ago, but that's 12 years later. So a lot of it is like you have these technology breakthroughs but then it takes time to get the technology to move through the cruft of the regulatory structure, the industry structure, the users and the retraining and the company budgets and all of that. So the future that we're building toward is one where humans are going to have to be at the center orchestrating and doing all of this model design and training and everything else where you're harnessing the power of generative AI with all of the workflow and tooling and data that you need and that's going to be kind of how this gets done. It could be the case that we just get a breakthrough, an ASI breakthrough where we go, ah, this solved the problem for us. I'm skeptical though. I think that'll be a while.

Speaker B: Super insightful, honest. One of the things that's going on in my mind right now is I think about the different providers of fund admin. You're all in kind of very different states of how tech forward you are and I put Juniper at the forefront of this. But there's a whole lot of folks who are really a, ah, body shop out there. Sometimes when we look at industries like this, the industrial logic here is if you get a breakthrough that is a tech level breakthrough, you start to see really different industry structure. Right? Uh, you have one provider or A couple providers with a substantially different cost position essentially to serve the market. Just playing through that a little bit. Like how do you see the industry evolving for the next couple years? I mean there's probably maybe limited ish near term, call it next one to two year impacts from a broader structural perspective. But if you look out three, five plus years, what do you think this market starts to look like?

Speaker A: Well, I do think that fund administration from an industry structure perspective is very well suited to this kind of. It's not a winner take all dynamic, but it is a winner take most dynamic. Right. Because almost everyone in the industry is running the same play. They have the same mix of onshore offshore resources, they have the same technology that they buy principally from third parties. It's a commodity offering.

Speaker B: Right.

Speaker A: Every fund needs fund administration and it varies, but it's a commodity. And then Everybody prices within plus or minus 10% of each other. And you do have some market participants that are pursuing really low cost strategies, but they have very low market share and no credible GP goes with them. And so by and large, everyone is clustered around the same playbook, the same structure, the same pricing, the same margin. Almost all these fund admins are owned by private equity firms that are pursuing the same EBITDA and sort of three to five year hold and investing strategy. So structurally, everyone's locked into the same play. And if you're a private equity owner of a fund admin business and you're going to hold this thing for the next three years, right. Are you really going to go plow $50 million into R&D for the next three years so that you can develop some breakthrough transformation for the next guy who's going to own it? No, right. You're like, how do I put another 10 points on EBITDA margin? How do I grow this thing at 15, 20%? How do I, you know, torque pricing a little bit? I'll torque offshore. I'll uh, you know, make some tweaks here and there. I'll buy a few businesses. That's your move, that's what you're doing. So there's very few in the industry. We are investing this way and we just raised a Series D and we've got great backers who.

Speaker B: Congrats.

Speaker A: Thank you. We announced $130 million Series D where we're very clear we're investing over a multi decade time horizon. And the play that we're making is it will take a few years. All the stuff I'm explaining, I just explained the waterfall example to you. Like there's a hundred of those and they all vary in great complexity. And so there's a lot of work to do. But the company that figures this out and gets it right, if everybody else is operating at a 50% gross and a 30% net margin, which is how most of them are operating in the industry, somewhere between 40 and 50% and somewhere between 20 and 30% on EBITDA, and you can operate at a 70% gross margin or 75% gross margin or more like, again, with the breakthrough potential of AI, that's a big problem for everybody else, right? And not only that, but like your product and service is becoming differentiated from the commodity offering that everyone else can offer, because only through you can you get the pizza delivery tracker and the data access and all these things. Right. So there's one more variable I really like, okay. Which is, and I think that this is not appreciate. I have a lot of friends that are venture investors. I'm, um, an LP in venture funds. And in some ways we have been working on the labor arbitrage automation problem at Juniper Square for more than a decade. Before AI was popular and before everyone's like, oh my God, I'm going after the labor budget in law firms or whatever, right? So we have some experience here, and AI is this breakthrough technology, but it's just the latest technology to sort of hit the scene. And a lot of the same constraints still apply if people forget. One of the big constraints is if you're going after the labor budget inside of the customer, then you are dependent on the customer's ability to transform their own workforce in order to realize the benefits of the labor savings. This is a super important point because it is very hard to do, right? It is like very hard to get your Comcast fleet of customer services representatives or whatever to drop the old way of doing things and learn some totally new way of doing things. And by the way, do that in a model where every time some new technology is adopted, there's a riff and 30% of the employees leave or whatever. Right. Like, good luck with that. And so if you look at our situation and the reason I like this so much, we provide a service for a price to the customer using our labor. Customer just buys the outcome from us. And it's a factory. They don't care how the factory is made. They want it to be secure. They want to have access to data, they want to have visibility, transparency, A, uh, robust trust in the system, of course. But like, at the end of the day, there's a factory over there, inside the fund administrator, you as the customer, don't really care. You want to buy accurate, on time, high quality fund administration outcomes. And so if you're a fund administrator that's growing at like 60 to 80% as we are, then a bunch of things happen. All of a sudden now you can be realizing productivity gains from these technologies that don't require you to have a dynamic in your workforce where you're constantly riffing, right? Because if you're doubling your headcount every year or more, then you just say, okay, great, a fund accountant's now going to go from being able to manage eight funds to 10 funds, so we'll just hire fewer fund accountants and we'll keep the really great ones, right? So all of a sudden you have a sort of positive flywheel, a positive feedback loop in the workforce dynamic. And then the other thing is, it's a real skill to figure out how to take new technology off of the shelf and retool your workflow and your labor force while meeting the constraints of accuracy, quality, et cetera. Most companies, this is not what they do. Most companies have been doing the same thing for like decades. So what so many venture investors miss is they just presume that this is a solved problem. The model shows that this thing gets it right eight out of ten times or whatever. Or the imputed labor savings could be X inside of the customer. And then instantly the venture investor just goes to, wow, labor savings of X times Y customers, My God, the TAM is huge. And what they miss is this huge step of what's it going to take for the customer to realize that benefit? Because the economic value has to get created somewhere. And is the customer really going to go riff half their team? You know, just every year, half the team gets riffed. So I don't think people are really thinking this. True. So I think that one of the great things about fund administration is you're on the other side of that problem. You control the labor inputs, they're inside your shop, you're delivering the outcome to the customer. You're in control of that problem, and the customer's happy to just buy. The service that gets more and more and more automated every year makes a ton of sense.

Speaker B: I think turkeys never vote for Thanksgiving. And it's just really hard to get people to the point where you, if you need to believe that in order to make the math work, it's actually just really hard a lot of times to really play that through. One other thing I wanted to touch on, um, is the Idea of me. This is the exact right phrase of passporting KYC identities and taking it across fund to plan. It's another place where you could see it could start small and then it could wind up having pretty significant impacts across, uh, funds and even in marketing and how people think about fundraising and, and reaching out to a broader audience. Can you talk to us a little bit about where Juniper is here and then where you see this potentially going on a much longer horizon?

Speaker A: Yeah. This is another example of we have been building toward a vision of a passporting system for LPs. What do we mean by this? If you're an LP who invests with many managers in private markets, as most LPs who invest in private markets do, you know, like, if you're of any kind of, uh, substance in the checks that you're writing, you've got venture, you've got private equity, you've got private credit, you've got all the asset classes, you have many managers within those asset classes, and then each of those managers have many funds, many SPVs, et cetera. So what that means is that it's this weird dynamic where you are in a repeat game in the relationship between LP and gp. It's the same lp, it's the same gp. Right. But just the way the regulatory process works and the way the systems have been set up, it's like every transaction is approached anew, uh, as though that repeat game weren't happening. And so it is a huge headache. It's like, wait, I invested in Bain Fund 9, I gotta go through KYC again for Bain Fund 10. Like, what the heck? And last time it triggered this review and I was in KYC hell for six months while they tried to figure out what this trust is or whatever. And so where we should get to as an industry and what we're able to enable on Juniper Square is every fund and every customer and every LP account and every position. All of it ties back to a single common industry data model. And the LP has a single login to all managers and all funds on Juniper Square, we're down to the ledger. That vision we were talking about, about ledger down to the record of ownership, all the way up to the portfolio of the LP is a single unbroken chain. So the scenarios you want to enable here are, uh, you want to prevent an LP from having to go through KYC multiple times. You want them to be able to authenticate once. And initially it starts with, I'm going to authenticate once with the manager and then I can reuse that with the manager and where it bridges to is actually as long as I've done this once with the platform Juniper Square, any manager on Juniper Square can reuse that. And then where you want to get to as an industry, because there's Juniper Square, there's many other scaled reasonable scale, but there's many other fund admins, there's many other systems, this is only really going to work truly for the user where you get this sort of seamless experience. If we can all, uh, those of us that have the capability to do it sort of agree on a, you know, like in the AI industry, thanks to Anthropic, we've sort of settled on this model context protocol, this MCP protocol for how an AI can share its model context. So like if you use ChatGPT, let's say, and you've been using it for two years and you've built all this history and context with ChatGPT and ChatGPT knows your kids names and where you lay, it just knows everything about you. And you want to go take that knowledge that ChatGPT has and use some other AI tool that doesn't have that knowledge and you want to in a controlled way let that other model access the knowledge that it has about you. You can do that via mcp. So we need to get to the same thing from a KYC perspective where the major providers like us, and we are open for business on this. The issue is just no one else of any kind of scale is pursuing the same issue. So our initial focus is make passporting, make universal subscription, make the experience really great for the LP inside of Juniper Square. So once the manager's on there, it's like this amazing seamless experience. But then eventually where you get to is all right, well you should be able to, if you've gone through that KYC process with Juniper Square, you should be able to authenticate with another provider using that credential, using kind of a, uh, protocol. And so I predict that at some point in the next few years the industry leaders will settle on some sort of protocol. I see it going that way versus there's a lot of companies that try to build a network around this and I just don't see that playing out. It's not, you don't go through KYC enough that you're going to go have a dedicated service and it's too hard to get the flywheel going. So I see it more as a thing where it's like when you have 10ish type players, all of some middling scale what you really need is a, uh, protocol and I see it going in that way for the industry. I don't see it as a winner take all type thing at all.

Speaker B: Super helpful Alex. Really appreciate your perspective here and hearing both the industry and about Juniper. Appreciate you coming on. Where could listeners learn more about what you're building at Juniper Square? From you.

Speaker A: So junipersquare.com is our website. I post on LinkedIn regularly about AI and what we're building and what we're learning. So Alex Robinson from Juniper Square on LinkedIn. You can follow me there. And those uh, would be the two primary places that I'd point people to.

Speaker B: Terrific. Thanks so much. And for everyone listening, please subscribe to Deciphered on Spotify and Apple and leave us a review if you enjoyed this conversation. We'll be back soon with more deep dives in fintech topics that matter most.

Speaker A: Thanks Alex. Thank you Mike.