Subscribe
Share
Share
Embed
The "manual era" of capital allocation is in its final chapter. The firms still relying on manual data extraction and analysis aren’t failing overnight, but they are falling behind one week at a time. While most of the industry continues to "white-knuckle" through 200-page documents and legacy databases, and manual Excel extraction, a new breed of Agentic Allocators is quietly rewriting the rules. They aren’t just using AI to summarize emails; they are leveraging AI-augmented workflows that intelligently automate parts of their investment and operational processes that were previously impossible to automate.
Hosted by Victoria Sienczewski, CEO and Founder of AuumAI, The Agentic Allocator is the "behind-closed-doors" look at how the world's most sophisticated Limited Partners (LPs), allocators and General Partners (GPs) are actually deploying AI, and the hard-won lessons from those building the systems.
This isn't a series about high-level theory or technical gibberish. Each conversation features industry leaders, forward-thinking LPs, GPs and experts who are rewriting the rules of capital allocation through agentic AI. Expect real-world case studies, tactical frameworks you can actually use, and moments that challenge outdated norms. You'll come away with a clearer understanding of the critical questions every allocator must ask - about data privacy, team adoption, integration, and governance - before investing in any AI solution. If you're tired of the "black box" and ready to evolve your investment office for what comes next, you're in the right place.
Welcome to The Agentic Allocator. Today, we are joined by Al Hemmingsen, CIO of a prominent US based single family office with a history going back to the mid twentieth century. The office manages a multi generational pool of capital with a quasi perpetual mandate. And he has been at the helm for the past decade.
Al:People, I think, are rightfully nervous about what AI will do. But the big thing that we're after is we're after like the 30% of people's jobs that is tedious, repetitive, beneath their education, doesn't use their gifts, doesn't make maximum benefit to the client, all of those things. And basically try to find things like that and basically liberate them from those things.
Victoria:Today, we'll discuss how Al is rethinking his investment organization's operating architecture through AI. From filling gaps in enterprise systems, the role of vibe coding, as well as navigating the cultural and security questions that arise on the journey of AI adoption.
Al:I think you have to show people what's possible. You know, it feels kind of like missionary work of, you know, you sit down with your office manager and like, okay, what are the really tedious things that you hate? We've got some new tools and okay, how could we automate this or what could we make? And you kind of do that a couple of times with a few different people and make a thing that saves them a half hour of drudgery every week or every day. And then they start to see the potential of oh, this is helpful to me.
Al:And then you can set them up with tools and a framework to kind of go forth and try to make their own lives better.
Victoria:I hope you enjoy the conversation. Well, Al, thank you so much for taking the time to speak with me today.
Al:Great to be here. Thanks for having me.
Victoria:Al, you are probably one of the only CIOs in the LT world who is actually releasing code that they write on GitHub. Would love to understand what got you interested in coding yourself and how that project started.
Al:Sure. So I guess, you know, I don't have a developer or technical background. I went to business school, and one of the prerequisites I had, it was a class that everybody had to take, it was operations and information management. And basically, they were mandating that everybody who goes through the program there basically take some form of SQL, or have some exposure to that. And then the biggest thing we used as well alongside that was Visual Basic for Applications, so really, honestly, primitive stuff.
Al:But it did open my eyes to the things that you could do if you're trying to process and deal with large data sets, was the big thing at the time. And then my first job was at a large global investment manager, and it was, you know, it's one of these organizations, you know, household name, 50,000 people, several thousand people in IT, and they they struggled with a lot the same issues that give rise, I think, to vibe coding, which is you had all these applications, you had all this data, but it didn't quite give you the reports you needed, it didn't have all the data you need, you needed to have, you know, some lookup and look across systems. And, you know, one of my first things when I came in there was, you know, VPs and the executive directors and the managing directors wanted to see certain reports, and, well, if you can query the tables, and I'm like, okay, these are, you know, these are side based tables. I can query these, and we can build some basic way to get the data out and then get a report that's in the format you wanna look at.
Al:You can make those. And that seemed that felt like enormously powerful, like outs like an outsized level of power relative to what it was. But if you had the ability to do that, you could, you know, you could be pretty impactful and make some pretty cool things. So that that kind of followed me throughout my career of you're always looking at the tools that you have, which is a pretty standard set of tools, they all have limitations, you're trying to figure out ways to do things more tailored, more custom, cheaper, etcetera. So so I just kind of kept kept doing it or trying to, you know, be a be a hobbyist coder maybe, is what it is.
Victoria:That's honestly fascinating, Al, in terms of how you've how you dove in early on in your career and how it's followed you, and how you're also integrating it into the culture of your organization right now. Before we dive into what you're doing on a day to day basis, we'd love to also just ask you about vibe coding. So almost, I would say, very few of your peers have done it. I'm curious, where do you start? What does it look like on a day to day basis for you?
Al:Well, felt very natural for me because, know, when I start, I'm gonna betray how old I am. When I started my career, my first job, I had these It was like the sequel bible, the Visual Studio Bible, like the Excel and Access Bible, these big, like, 700 page things. They just sat on my desk because I would thumb through them, and then that kind of became like Stack Overflow. You'd look up how to do stuff, you know, basically Google searching, how to write things. And then GPT came along in the other chatbots, and you just start kind of You'd spawn the web browser and ask, hey, how do I do this again?
Al:What what does this need to look like? Or I'm more And then the questions got broader, right? Like, okay, and now it's like, you know, I have this workflow. This is tedious and repetitive. How should I think about it?
Al:What are some options? And it would start to guide you along. And so so kinda like the first version of that was just like, okay, you know, let's let's try this in Python chatbot, and it would draft a script, and I'd download it and run-in like, error message here. Take a look again. You know, we're we're working through it.
Al:And then, you know, ClaudeCode came out a few months ago, we tried a few things on that, just kind of giving it a little more agency, letting it do stuff, and it's pretty profound, pretty powerful. We saw the ROI on it instantly, and it's huge.
Victoria:Can you dive into that ROI a little bit more in terms of maybe giving an example of a specific workflow or a specific element that you've automated and how you measure that ROI internally?
Al:Yeah. So the biggest thing is there are things that you can do now that were just either either they were too expensive years ago, or they're just not feasible. So we had a project where, you know, we have data that goes back fifty plus years, like transaction data, returns, prices, tax stuff, all that. You can imagine, you know, back in the old days, they were running this stuff on on things that predate Windows, like DOS and prior systems. Right?
Al:And the date, you know, they're not they're not really machine readable today, like, it's not like a plain CSV file, or it's not it's not relational database structure, like, it's not in a modern way that you can you can query it or really interact with it. And so years ago, maybe eight years ago at this point, we we had a consultant come in and say, alright, look at all this old stuff we have. We wanna keep this history. We want it to be readable going forward. You know, standing up these kind of emulated systems isn't the best answer.
Al:And the quote came back, yeah, we can do that. It'll cost you, you know, it was a 6 figure quote. It was some insane price. And we just kind of said, well, not now. Let's just defer it.
Al:So put it into kind of a a hosted emulated system that can run kind of like an emulated old version of Windows, the old version of the software, keep it secure, we don't have to worry about it. But it, you know, those businesses are kind of like melting ice cubes. They keep ratcheting the subscription fee. And it was actually maybe a couple months ago. Quad Code had been out for a couple months.
Al:I'd heard of it, hadn't used it, and I got the latest invoice for the hosting service, and it had crossed $10,000. And I just kind of I just got annoyed one afternoon. It was on a Monday, and I said, you know what? Let's try this. And so I fired up fired up Claude, and we just started talking to it, like, okay, here's what we have.
Al:Can you read this data? Where are the transactions? How do the reports work, you know? And sort of six hours later, and I wrote it exactly because I was so shocked. Was $59.32 of API credits later.
Al:We had all the data out, modern text delimited formats. We had a report writer on top. All the data was there. We confirmed it all. It all moved over.
Al:All the reports you would write on the old system and the new system matched, so it passed audit and credibility. It was writing it correctly. We could replicate the entire system, and basically, in an afternoon, you save $10,000 in perpetuity.
Victoria:Incredible.
Al:Yeah. That was the thing that kind of flipped it. And then and then we just kinda went crazy looking for these other things in a in a very old organization with legacy data and legacy systems, and there's there's a lot to be done there. There's there there are many of those types of projects where it's maybe a $100 in credits and a couple days of time, and the the cost savings is enormous.
Victoria:Incredible use case and incredible ROI, Al. Unbelievable. And would love to dive into the system that you've designed internally. So you have some enterprise grade services and tools, databases that you that you still maintain. And then in many cases, you also do it yourself with some scripts that patch things and execute different workflows internally.
Victoria:How do you think about the kind of transition between those two different contexts and how you manage that internally?
Al:Yeah. Like, think I was reading, like, Anthropic deployed Workday to manage all their HR type stuff. So, you know, if Anthropic is not confident enough to use their tools to stand up something enterprise grade like, you know, an amateur like me or my small team has no hope of it. Right? So I think the enterprise system of record, we have a couple of them that store, you know, all of our transaction data, returns, balances, all that stuff.
Al:They're secure. They connect to dozens of financial institutions. There's verification processes. There's user management and permissioning rights and all this stuff. We're not gonna make that.
Al:It's There's no reason why we should. But like a lot of, you know, basically every system that people use, there's limitations. It doesn't do You know, we have custom things in our in our lives that don't occur with other users, therefore, software company is gonna spawn and build that stuff. So so what we've tried to set up is like, kind of in between the seams of what the enterprise system can do and what another system can do. There's stuff that just can't get done.
Al:Basically, can you make things that that fill those needs? And it's generally like, you know, we are we're reading from the systems, we're not writing to them. So it's the the system has an API, we can query it, so you can go on the web browser and you can run reports, or you can query the data out for other purposes. So we're reading from the systems, not writing. They're kind of limited scope, so if they're writing something, it's writing to, like, an Excel spreadsheet or a special kind of report or something like that.
Al:It's not writing to like enterprise systems. So at least for the moment, they're they're very much like limited scope things, but it feels like, you know, it's it's kind of like what you used to do on your own, you know, with, you know, the Visual Studio Bible or Stack Overflow trying to make a solution that's automatic, but now you're now you have much more powerful tools to do it.
Victoria:Incredible, Al. And would love to dive into obviously, you're not the only one who is writing these scripts and architecting the system design internally. How do you think about enabling your team, whether it's a controller or someone on your accounting team, to actually start utilizing these tools in a day to day way?
Al:Yeah. So there's a couple pieces to it that I've thought about, like, people I think are rightfully nervous about what AI will do. You know, they have people that may be used to work in manufacturing in their families, and, you know, those jobs got offshored. Or, you know, if you have some sort of technical work, maybe some of that's being done off shore. Know, people are rightfully worried about what happens to them in the future.
Al:So I think the first thing that I've tried to do is, and it's easier in a small organization, but take that fear off the table of, you know, if somebody's worried about their very survival, they're not worried about, like, you know, leveraging their gifts to the benefit of the company in the best way. Right? So you have to take that stuff off the table, where the message that we've I've tried to push through consistently through the firm is we're not looking to outsource, reduction in force, you know, radically, you know, basically create any sort of existential threat to your place, and we have a team that I think you can you can do that with. But the big thing that we're after is we're after, like, the 30% of people's jobs that is tedious, repetitive, beneath their education, doesn't use their gifts, doesn't make maximum benefit to the client, all of those things, and basically try to find things like that and and basically liberate them from those things. So that's kind of like the psychological part.
Al:And then the technical part is, I think you have to show people what's possible. Like, you know, people that have like 50,000 Twitter followers or, you know, go on podcasts, like, you know, that that ecosystem of people is aware of what Claude Code is, and has seen examples, and it's kind of obvious, but like for a lot of the population, it's not. Right? So you have to show them what's possible. You know, it feels kind of like missionary work of, you know, you're gonna sit down with your you're with your office manager and like, okay, what are the what are the really tedious things that you hate?
Al:And we've got some new tools, and okay, how could we automate this, or what could we make? And you kinda do that a couple of times with a few different people and and make a thing that saves them a half hour of of drudgery every week or every day, and then they start to see the potential of, oh, this is this is helpful to me. And then you can set them up with with tools and a framework to kinda go forth and and try to make their own lives better.
Victoria:What is that framework? How do you think about that internally?
Al:Yeah. I mean, so it's it's early. But, you know, you can see where this is going. You're gonna have, at least in our case, what we've tried to do is we we wanna have like a common library of apps. So if you've made something, you know, you're testing on an isolated environment or some local environment, you've made something that's cool, It should have mileage beyond just you to the firm.
Al:Right? So what we do is we have everybody uplift their final product to a Microsoft environment, where it's gonna be visible and usable by everybody, but then also, you know, other people can come in and monitor it for security, for maintenance, upgrades, all of that stuff. So it's not Nothing becomes individually owned, it all gets corporate owned. That's probably the framework for We That that will lend itself, I think, longer term to kind of systematic, repetitive security testing and monitoring, just usage monitoring. You know, one of the other principles that we've tried to just at least talk about internally is it's kind of like analogous how, like, you know, driver assisted cars or planes work.
Al:Like, the plane has autopilot, but who's responsible for what the plane does? It's always the pilot. It doesn't matter, like, what machine or what person was interacting with the plane, like, pilot is responsible for what happens. Know, you're the creator, you're responsible for what the thing is doing. So we're very, you know, we're very perhaps slower, but it's, you know, we're reviewing what it wants to do.
Al:You know, some of these these tools will make plans for you or proposals of what they're going to do. You can look at the actual script, at least at a at a basic level, what it's going to do, approve it. So you're you're ultimately accountable for what you create is the framework that we have.
Victoria:Fantastic. And how do you spread that kind of information internally? I know every app script that is created is uploaded to a Microsoft environment. But culturally, do you have weekly stand ups that you share kind of best practices or what folks have created? How do you make sure that everyone is benefiting from that knowledge as well?
Al:Yeah. We do. I mean, it's it's kind of easier in this kind of a construct because it's so new, like, there's not really, like, a central source of authority. Right? You know, maybe in different domains that are more established, like, I don't know, maybe accounting or finance.
Al:Like, are, you know, there's kind of a prescribed control environment that if you came into an organization, like, okay, this is how you separate checks and balances, and, you know, there's, you know, there's less common dialogue around it because it's kind of set. AI is so new, like, there there are relatively few centers of authority, so it's Yeah. It is. It's part of our weekly We do a weekly meeting on Mondays. AI systems is a discrete part of that.
Al:It's a defined part and section, And it's what have people been using, what have people been finding, here's what's been created, here are things that are new. We do something that's particularly interesting or relevant for a large group. We'll have somebody who made it do the demo, how it works, and, you know, basically basically make people aware of what's what's become available that they might be able to use or get inspiration from.
Victoria:Absolutely fascinating. I love the importance of sharing information and best practices and learning across the organization. So making sure that if one person leaves, that that institutional kind of collective memory is still there and and able to be utilized by the rest of the team in the organization. And Al, you mentioned you want to get rid of the 30% of people's jobs on your team that are the drudgery, the kind of repetitive manual work analysis, moving data around. When we successfully eliminate that 30% in the future, maybe it's even gonna be 50%, where do you think we will be as an industry?
Victoria:What will people's day to day work look like within an institutional family office?
Al:Yeah. I don't I know. Maybe I'm too much of an optimist, but I take the overview on humanity. Like, I think human, like, you know, human ambition and human greed are kind of roughly infinite. So I think, you know, if you free up if you free up people's time, like, think you'll just find more and higher and better things to do.
Al:And I think that's been the trajectory of humanity. Like, you know, people would talk about, like, oh, we're gonna have washing machines and dishwashers, and we'll all work twenty hours a week. And guess what? Nobody works twenty hours a week. And we've automated, like, you know, I'm old enough.
Al:Like, I took an accounting class where you were literally writing in a ledger. Like, again, like the 15 hundreds. Right? Double entry, writing it twice, adding it up. Like, I am old enough that I did that once.
Al:We've automated all of that, and we're still short of accountants. So I I I take the optimistic view on some of this. I guess, like, for my own, like, my foundational training and skill is investing. And so, like, you know, some of these tools, like, what do they do for, like, my investing work? You know, I think the the investment proposition is maximized.
Al:Like, basically, the more stuff that you can see and the more stuff that you can efficiently filter and be highly selective on, that's that basically maximizes, I think, your your investing outcomes. The more that you can see and the more that you can filter. So basically, the the widest top of funnel feasible and the narrowest bottom of funnel. And I think that, you know, I've had a few instances where like, you know, I'm going to a conference, there are 300 GPs in attendance, I have 15 meeting slots, I need to, as efficiently as possible, process all of that stuff to to create the highest quality meetings that I can and therefore the highest quality bottom of funnel that I can. You know, the tools are just better.
Al:So I think it's just you'll see see and evaluate more stuff, filter through more stuff, but probably be augmented as you do it.
Victoria:It's such an important point, the the leverage and the opportunity that AI will create to actually screen and filter more investment opportunities. You think about the universe of alternative managers, you know this. Right? There's around 56,000 globally, and even the most sophisticated organization today will screen maybe 300, maybe 400 a year. And so obviously there is some level of a myopic view which, you know, we can discuss whether or it's right or wrong, but I am fully aligned with you that the more opportunities you're able to screen and also collect data on, which then you can leverage internally to better improve your screening and your sourcing, will generate better investment outcomes.
Victoria:But it's still to be seen, right, and still to be proven how AI directly impacts that, but it's an extremely exciting opportunity. Yeah. And Al, why don't you again just dive into your kind of future view of the industry? If you were to advise a peer organization, so a CIO at a institutional family office on how to get started, what would you say to them?
Al:Yeah. It's it's kinda funny, like So I joined the family office about ten years ago, and I was previously at a global asset manager, then I was at a consultant, where a lot of the stuff you would make or think about, kind of the default view was everything is proprietary because you're competing with people. That's much less so in a in a family office or, like, in an endowment or foundation. Like, you benefit, you leverage yourself by collaborating with as many people as possible. So I've I've tried to do that as much as I can, and for the most part, like, again, if you're Again, most of the time, for most sized institutions, I think that's true, that you benefit much more by collaborating and sharing.
Al:So there are a lot of people that are that are doing kind of what I'm doing, or even kind of more elaborate things than what I'm doing. There's a lot of sharing to be had. I threw up at GitHub years ago, and I'll see where that goes now that like, Claude is basically reading all the GitHubs and presenting it to you. But, like, you know, if you make something that's interesting and you're in kind of a seat like mine, you're not competing with people, so share freely, you know? I think the other imperative, though, is there there are aspects of the job that are a little bit competitive.
Al:So I mentioned, like, this this one specific conference I'm going to that does feel competitive in that, like, the highest quality GPs or the highest quality opportunities. It's no coincidence those calendars fill quickly. So if you're if you're AI augmented and you can process all that stuff faster than other people and you can get into contact with people faster than other people, you do have an advantage. So I guess it's it's maybe a little bit maybe there's two ideas there that seemingly are in conflict, but it's like you you kinda have to share a little bit to leverage yourself and to have a view of what's possible. But then I think you have to have a pragmatic view that, like, this is a competitive business, and if you don't leverage this stuff in a good way, other people will and they'll just outcompete you.
Victoria:Absolutely. For example, one of the use cases we hear a lot is, okay, I have an investment professional fantastic. When they leave to start their own fund, I wanna be the first one in the door. So can you scrape LinkedIn for me? Or can you, you know, try to track them in some way?
Victoria:And AI will give you that opportunity to be able to actually keep track of these individuals, for example, in a systematic way and reach out to them to be first. Or in terms of market dislocations, if you need to switch something around and act quickly, you'll also be at an advantage. It's a fascinating point about the two areas, about sharing information, and then also kind of accelerating your own competitive position.
Al:Yeah. And there are few The other thing that I I think about, like, are a few firms, like, it used to be like the benefits of scale were really obvious. Like, are a few firms that have kind of done versions of this where they they're very large, have large staffs, they've collected all this data over years and years and years, private data that's hard to get. And then basically hand mapping it, which is a torturous job, I'm glad that AI has come so I'll never have to do it. But like mapping, okay, here are the people inside these firms that led these deals, and here are the deals, and here's how they perform, and here are the operating improvements and all this stuff.
Al:And that stuff is a gold mine once you have it, but it takes, you know, I don't know, hundreds of thousands of hours of human labor to create, and now there are a few systems, like we're subscribers to one, which is kind of doing that. You upload all of your private data to it, and it starts to make that mapping for you. So you can peer group, you can look for, you know, who are the likely people spinning out and spawning new firms, what have the best deals been, all of that stuff. And it's it's pretty wild that now you can do that with a machine.
Victoria:It's fascinating. Al, thank you again for taking the time to join me. It was such a fascinating conversation and what you've built and how you've approached AI integration internally. So thank you again so much for your time.
Al:Happy to. Thanks for having me.
Victoria:That's a wrap for this episode of The Agentic Allocator. If today's conversation gave you a clearer vision of where the industry is headed or helped you pinpoint exactly where your own process is stuck, go ahead and follow or subscribe wherever you get your podcasts. And if you're curious what Agentic AI might actually look like inside your investment office and how to get there without compromising on security or control, visit auumai.com for resources and demos on AI native LP and Allocator workflows. Until next time.