Latest / The Index Podcast / Franklin Templeton: Inside an 80-Year-Old Asset Manager's Push Into AI Agents
Transcript
- 0:07SPEAKER_01: One quick word from our sponsor, Triton1.
- 0:10SPEAKER_01: If you're building on Solana, your app is only as fast and as reliable as the infrastructure underneath it.
- 0:17SPEAKER_01: At Triton, we built the open source read layer for Solani ourselves.
- 0:22SPEAKER_01: The streaming engine that pushes data the moment it hits the chain, the transaction pipeline that lands your trades when everyone else is timing out, all of it running on bare metal in over 20 data centers worldwide, served by the team that wrote it.
- 0:34SPEAKER_01: We power some of the biggest names on Solana, but with Pay As You Go, anyone can start.
- 0:39SPEAKER_01: Sign up, get an endpoint, pay only for what you use, head to Triton.1, Serious Infrastructure for Serious Builders.
- 0:45SPEAKER_01: I've worked with Triton for years at this point.
- 0:48SPEAKER_01: They've been around since the very beginning of Solana.
- 0:50SPEAKER_01: They've contributed open source software that's the core of the infrastructure of Solana that makes everything work since day one.
- 0:56SPEAKER_01: I couldn't recommend working with a better team.
- 0:58SPEAKER_01: Triton.
- 1:14SPEAKER_01: Is that how you say that?
- 1:14SPEAKER_01: Gawkman?
- 1:15SPEAKER_01: How do you say your last name?
- 1:16SPEAKER_01: Yeah.
- 1:16SPEAKER_01: Gackman.
- 1:18SPEAKER_01: Max Gockman.
- 1:18SPEAKER_01: He's the head of AI and digital asset solutions at Franklin Templeton Investment Solutions, where he's building agentic AI teammates for the investment process and thinking hard about how AI trade actually plays out in markets.
- 1:32SPEAKER_01: Max, thanks for coming on the show.
- 1:34SPEAKER_00: Yeah, super excited to be here, Alex.
- 1:37SPEAKER_01: So I'm really interested to hear more about what you're building.
- 1:40SPEAKER_01: But before we get into that, I kind of want to understand what makes you tick.
- 1:43SPEAKER_01: You know, why are you how did you get here to head of AI at Franklin Tembleton and what gets you out of the bed in the morning to do this job?
- 1:50SPEAKER_00: It's it's been an interesting journey.
- 1:52SPEAKER_00: I started my career as a quant back in 2006, you know, built a hedge fund that did pretty well for the financial crisis, then had some more traditional roles, um, including heading up a very large insurance company's uh mutual fund business.
- 2:12SPEAKER_00: And then I kind of got complacent and I hated being complacent.
- 2:18SPEAKER_00: So what I did is I quit my job and started a fully AI hedge fund.
- 2:23SPEAKER_00: Now, this is before large language models, right on the precipice of when OpenAI launched ChatGPT.
- 2:29SPEAKER_00: Um, and I just had a feeling that there was something to this AI theme.
- 2:34SPEAKER_00: And having been a quantitative investor for a while, it seemed like it could rewrite the rules of investing.
- 2:41SPEAKER_00: So I did this for a couple of years, um, working, you know, 100 hours a week and overclocking all of my hardware.
- 2:49SPEAKER_00: Um, but realized that you just need scale to do this.
- 2:53SPEAKER_00: Like to do AI in finance, you need institutional scale.
- 2:58SPEAKER_00: And that ultimately led to me coming to Franklin, where I actually had a couple different roles, most recently uh chief investment officer for our solutions team, until we all at the leadership level decided that we needed someone to really focus on investments uh and how they work with AI.
- 3:21SPEAKER_00: So we have a chief AI officer who is extremely talented, really focused on the technical aspects.
- 3:27SPEAKER_00: I work really closely with him on how do we build AI for our investors?
- 3:34SPEAKER_00: How do we build AI for our clients?
- 3:35SPEAKER_00: How do we use AI to enhance alpha and sharpen our edge?
- 3:40SPEAKER_00: Um, and that's not even getting into the digital asset side where I work with folks like Chris and Sandy, who I think you've had on the show before, uh, to really think about what is the institutional layer for digital assets?
- 3:53SPEAKER_00: How do they integrate?
- 3:54SPEAKER_00: And by the way, um, I'm sure you won't ask me this, but sometimes people do.
- 3:58SPEAKER_00: Like, how does AI and digital assets come together?
- 4:00SPEAKER_00: There's an enormous overlap.
- 4:02SPEAKER_00: When I think about especially some of the solutions that we will build for clients, they can be smart contracts governed by AI but create brand new strategies that are massively scalable, actually, democracy, like really not using that as a throwaway word, but really democratizing investment at an institutional level all the way down to the retail investor who may only have a few thousand dollars to uh chip in.
- 4:27SPEAKER_00: And then on the flip side, we can also use blockchain technology for a lot of really core AI processes.
- 4:33SPEAKER_00: Like if I want to create an immutable ledger that has every decision my agentic analyst made, that's a perfect use case for blockchain.
- 4:42SPEAKER_00: So there is enormous overlap on both sides.
- 4:45SPEAKER_00: And I think you can tell just from how excited I am, even describing it.
- 4:48SPEAKER_00: This this gets me out of bed.
- 4:50SPEAKER_00: I've been up since five this morning, and I'm probably not gonna go to bed till I'm done testing one more thing with my uh cloud environment.
- 4:59SPEAKER_00: So I've just never been this excited to work this hard um building something really great with enormous backing and resources of a$1.8 trillion asset manager.
- 5:10SPEAKER_01: Yeah, that's pretty cool.
- 5:11SPEAKER_01: Um, that's that's scale.
- 5:12SPEAKER_01: That is the scale you were looking for.
- 5:15SPEAKER_01: Uh I there's so much to unpack there.
- 5:18SPEAKER_01: Um, I do, I'm actually really glad you brought up the intersection of AI and blockchain technology.
- 5:23SPEAKER_01: It's funny, my dad was just texting me yesterday, like, what does how does how do these things intersect?
- 5:28SPEAKER_01: And my response to him, literally like yesterday, he texted me this.
- 5:32SPEAKER_01: And my response to him was uh uh payments was the first one.
- 5:36SPEAKER_01: But that's because that's where like my head was at, you know, about a year and a half ago.
- 5:42SPEAKER_01: You know, we we built at my company Corbett's um the first open source framework for agentic commerce that that was agnostic in terms of the standard.
- 5:50SPEAKER_01: So we had like X, I mean still exists, but X402 supported MPP, any any protocol and standard that comes out you can you can use with our open source protocol or framework um on any on any blockchain.
- 6:04SPEAKER_01: And that was like a very obvious, very big opportunity, right?
- 6:07SPEAKER_01: Like billions of agents sending money at the speed of light on the internet.
- 6:11SPEAKER_01: Um I want to hear more, I want to dig into the thing you brought up though, which was this democratization of access to stress.
- 6:20SPEAKER_01: I wasn't I kind of lost you there, like on like what it was after that, like that a person with like a thousand dollars could like put into a smart contract or something.
- 6:29SPEAKER_01: Can you explain that to me a little bit more?
- 6:31SPEAKER_01: Um, because I it did catch my attention.
- 6:34SPEAKER_01: Like I think it's different than some than the other things I've heard.
- 6:38SPEAKER_00: So imagine you have just you know, let's let's literally say you have a thousand dollars to invest.
- 6:45SPEAKER_00: Definitionally, through traditional finance routes, you would be limited to a couple, you know, very simple index ETFs, or you know, you could buy a few individual stocks, which most institutional investors or any professional investors would discourage you from because it's so concentrated.
- 7:04SPEAKER_00: And that was pretty much it.
- 7:06SPEAKER_00: You just have to wait for that money to grow until you could diversify more and more.
- 7:10SPEAKER_00: But if you own a tokenized fraction of a share, your thousand dollars can actually go a lot further.
- 7:19SPEAKER_00: You can now build a portfolio that's fully diversified.
- 7:23SPEAKER_00: And then what if there was a smart contract that could manage that portfolio for you?
- 7:30SPEAKER_00: And that process is actually a real institutional investment process, which by the way, this could be done by AI that pulls from the institutional intelligence that we have at Franklin, like all of our research notes, all of our investment views, you know, everything that makes us like or dislike a security, that AI could harness all that and then translate that to an individualized investment strategy, or it could take what we would call a model portfolio where we can build an example allocation, we being humans, so portfolio managers and analysts, and then translate that to that individual investor's holdings, again, accounting for specifically what their needs and what their goals are.
- 8:22SPEAKER_00: We actually have a version of this called the goals optimization engine, which we launched um several years ago and it's been growing at a very fast pace, multi-billion dollar uh platform now.
- 8:35SPEAKER_00: That's something that can really create massive personalization at scale.
- 8:43SPEAKER_00: And those are the things that are blockchain plus AI.
- 8:46SPEAKER_00: It's they they whenever you hear personalization and scale in almost any application, there's probably a blockchain and an AI angle too.
- 8:54SPEAKER_01: So is it I I get the personalization and how that AI plays into that, right?
- 8:59SPEAKER_01: Um, and it's using the alpha that you have the intelligence that's institutionalized inside of Franklin Templeton, which is pretty cool.
- 9:06SPEAKER_01: And I have questions about that, but I want to put that pin that up here for a second.
- 9:10SPEAKER_01: Um, the blockchain component sounds like it's really the fungibility, like the ability to just make like you know, own a fraction of a fraction of a share in a really diversified portfolio versus having to own like a whole share.
- 9:22SPEAKER_01: Is that really the reason?
- 9:24SPEAKER_01: You know, obviously the immutable letter, but that it just makes it easy for instance settlement, 24-7 trading, um, and the fractionalization, right?
- 9:31SPEAKER_01: Those are the main value props.
- 9:33SPEAKER_00: And remember, it also removes a lot of the friction that we would typically have in traditional markets.
- 9:37SPEAKER_00: Now, granted, there's now a lot less of that because of things like payment for order flow, which Robinhood really popularized, right?
- 9:44SPEAKER_00: So like you've reduced a lot of the fees for trading equities, but you really cut them when you when you go on chain if you do it right.
- 9:53SPEAKER_00: So it's yeah, less less middlemen, it's can be really direct from the institutional investor level down to anyone.
- 10:03SPEAKER_00: Um, and I think that lowering of differentiation in in service and also the assets you can access.
- 10:10SPEAKER_00: Like that's another key thing, right?
- 10:12SPEAKER_00: We can't really talk about tokenization without talking about private assets.
- 10:15SPEAKER_00: Um, you know, when you think about like what has been the real economy in the United States, it's not been the public markets.
- 10:23SPEAKER_00: It's been private markets.
- 10:25SPEAKER_00: And, you know, when we talk about AI, we could of course talk about, you know, the largest companies, among the largest companies in in the world are currently private and they're planning on going public, but have been delaying those IPOs.
- 10:37SPEAKER_01: But that's a pattern we've seen for decades, right?
- 10:40SPEAKER_01: These startups they reach huge trillion dollar valuations before they go public.
- 10:44SPEAKER_00: And I will say, I'm almost like I'm kind of allergic to the word startups when we talk about a trillion-dollar company.
- 10:49SPEAKER_00: Like these are not startups.
- 10:50SPEAKER_01: You're not a startup anymore.
- 10:52SPEAKER_00: Enormous companies that have chosen to stay private and may want to tap public capital in equities later, but they don't have to.
- 10:59SPEAKER_00: So they can tap debt capital if they need financing on that side of the balance sheet, or they can just type private capital because it's extremely abundant.
- 11:08SPEAKER_00: And so it's but it's not just the anthropics and the open AIs that are staying private.
- 11:13SPEAKER_00: It's also most of the real economy.
- 11:16SPEAKER_00: Most smaller companies are also staying private.
- 11:19SPEAKER_00: In fact, they probably will never go public because they're not going to find as as much of an audience in the public markets.
- 11:26SPEAKER_00: And with all the additional regulations and scrutiny, they may not have the resources to actually manage themselves as a public company because it's a much more uh serious and expensive regulatory undertaking.
- 11:39SPEAKER_00: So, because of that, we've had this huge mismatch between who actually owns the real economy and it's been concentrated at the institutional investor and you know, ultra-high net worth level.
- 11:53SPEAKER_00: But if we tokenize limited partnership shares of those private equity firms, of venture capital funds, of real estate investments, then all of a sudden you can take all of that real economic value and give it to the individual at any level.
- 12:14SPEAKER_00: Again, that's the mechanics.
- 12:16SPEAKER_00: Regulation needs to catch up to that, but from a technological perspective, blockchain allow allows us to do that.
- 12:23SPEAKER_01: Yeah, my one of my favorite companies that I'm an investor in that's doing this, it's a different market, but it's is uh BAXI, and they tokenize whiskey.
- 12:34SPEAKER_01: And it's the same analogy, right?
- 12:36SPEAKER_01: It's a totally ill, I mean, it's funny, it's illiquid, it's liquid, it's liquid courage, but it's like an illiquid asset that you can't trade, right?
- 12:43SPEAKER_01: Like it's it's very hard to and it also in that market specifically, there are a bunch of inefficiencies around lending because um and that they have solved for because they tokenize the the casks, and then they can get cap working capital at cheaper rates using DeFi basically to the distilleries, which is really interesting.
- 13:02SPEAKER_01: Um, but it's along the same lines of what you're talking about in the sense that these distilleries and these and those investment opportunities and the trading of whiskeys and the funding of of lending for um for this physical object is typically a private business that's very that that has a small group of people who have access to it.
- 13:22SPEAKER_01: And then all of a sudden you bring in blockchain technology.
- 13:25SPEAKER_01: And in their case, what's really cool is they use um helium.
- 13:27SPEAKER_01: You know, helium, the company helium that has like so they use helium to to connect IoT devices because these distilleries, the casks, they're in remote places that typically don't have internet.
- 13:37SPEAKER_01: And so they they use the helium nodes to route data, IoT data from the casks so that the distillery can track like how much I I forget what the word is, but like you know, over time the the the alcohol evaporates, right?
- 13:50SPEAKER_01: So they have some some amount of lower.
- 13:52SPEAKER_00: That's called the angel share.
- 13:53SPEAKER_01: Right.
- 13:54SPEAKER_01: Yeah.
- 13:54SPEAKER_01: So so that that kind of data they can track in real time now.
- 13:58SPEAKER_01: And uh, and that's good for the lender.
- 14:00SPEAKER_01: They literally, like the last time I talked to Todd, the founder, the CEO over there, he told me that there was a huge fiasco in that in the private market where a bank had lent a bunch of money to a distillery for like 20,000 casks.
- 14:12SPEAKER_01: Well, guess what happened when they went bankrupt and they tried to find those casks?
- 14:16SPEAKER_01: They weren't there, right?
- 14:18SPEAKER_01: Like they didn't exist.
- 14:19SPEAKER_01: Um, and and I I just I illustrate that example just to double down on your point.
- 14:23SPEAKER_01: Like what a huge opportunity this is.
- 14:25SPEAKER_01: There's this illiquid market or private market.
- 14:29SPEAKER_01: It's highly liquid for people who can be in it, that now the whole world can get access to.
- 14:34SPEAKER_01: It's just very exciting.
- 14:35SPEAKER_01: Like I now I'm getting excited.
- 14:37SPEAKER_00: Well, and and by the way, like take that you know, whiskey uh example, and we can talk about other assets that just haven't existed before.
- 14:44SPEAKER_00: Like one of the things uh we've been thinking about is cultural assets, and that can be anything from things that we would consider like high-end collectibles, which I would broadly put, you know, like very expensive, limited spirits into, um, as well as like, you know, jewelry, classic cars, etc., but also things like royalties.
- 15:06SPEAKER_00: Like if you want to, let's say you want to support an artist or an athlete, like very early on in the career, you can actually buy a stake in them and then get airdropped royalty payments or prize winnings as they become bigger.
- 15:20SPEAKER_00: So kind of crowdfunding, but in a much more robust and easier to track way, right?
- 15:28SPEAKER_00: Because it's not like, hey, I'm gonna give this guy whose show I went to at the bar, you know, 20 bucks so he can, you know, go on tour.
- 15:35SPEAKER_00: It's like, oh, I'm gonna buy an actual share.
- 15:37SPEAKER_00: And then when if they ever sign to Interscope, I actually get a pretty nice royalty check from that, you know, 20 bucks.
- 15:44SPEAKER_00: Like that's something that I think we're also gonna see.
- 15:48SPEAKER_00: It's for me, it's it's it brings it to another interesting thing, which is just a fascinating investment theme to me, that a lot of the younger generations have stopped thinking about investing the way the older generations do.
- 16:03SPEAKER_00: So rather than thinking about, you know, like I have my salary, I put some money into a 401k, I may have a trading account on the side, um, you know, et cetera, they're really looking at all sources of income very um immutable, like very in in the in this, I don't know what the right word is, but like they can choose between them without any specific bucket, right?
- 16:31SPEAKER_00: It's whether I get income from an airdrop off of a token that I have, whether it's a paycheck, whether it's an insurance settlement, whatever.
- 16:40SPEAKER_00: Like I am gonna look at it holistically, which actually makes sense from an investment perspective.
- 16:46SPEAKER_00: Like bucketing is a type of behavioral fallacy.
- 16:49SPEAKER_00: So as they're looking at that, they're also starting to get more differentiated income streams.
- 16:56SPEAKER_00: So let's say I've been playing a lot of uh MMRPG and I sold my character for you know half a Bitcoin.
- 17:05SPEAKER_00: That's an income stream.
- 17:06SPEAKER_00: Maybe I'll go build up another character and sell that.
- 17:10SPEAKER_00: They're not differentiating between that and clocking into their day job because that this is all part of their income.
- 17:19SPEAKER_00: And so as these new income sources come in, we're again gonna bring it back to like why we need smart contract-enabled investment advice.
- 17:27SPEAKER_00: We need something that can track all those income sources for that client and help them manage to the whatever their actual end investment goal is, whether that's retirement, saving up for a house, which is gonna be a little bit harder these days with 7% mortgages, you know, or just uh paying off student loans.
- 17:48SPEAKER_00: Like those are all things that we need to think about in a very different world than we ever did when it was as simple as I get a paycheck, I have a pension, and you know, at 65, that pension vs.
- 17:59SPEAKER_00: And I can go and buy my second house in a boat.
- 18:02SPEAKER_00: Like those days are never coming back.
- 18:05SPEAKER_01: So I want to go back to something you said earlier, which let's talk about the alpha.
- 18:09SPEAKER_01: How long has Franklin Templeton been around?
- 18:12SPEAKER_00: Uh we've been around for almost 80 years.
- 18:13SPEAKER_01: Yeah, I thought so.
- 18:14SPEAKER_01: It's like an 80-year-old company.
- 18:16SPEAKER_01: Uh, that's a lot of alpha, that's a lot of time in the market.
- 18:19SPEAKER_01: And I'm curious, it sounds like you guys are leveraging that data in with AI.
- 18:27SPEAKER_01: How are you doing that without giving your your alpha away to like open AI or like anthropic or something?
- 18:34SPEAKER_01: How are you guys thinking about that?
- 18:36SPEAKER_00: It's a great question.
- 18:37SPEAKER_00: Um, there's a lot of complex infrastructure behind it, some of which I can't talk about.
- 18:42SPEAKER_00: But what I will say, it's a question of how you separate training and inference.
- 18:48SPEAKER_00: It's obviously very important that you don't have your data train another model, um, you know, or a competitor's uh instance of that model.
- 18:59SPEAKER_00: But also, if you don't run inference on your data, then your data is not going to be very useful because the way it's useful is the agents form their contextual knowledge from those, you know, in our case, 80 years of investment.
- 19:13SPEAKER_00: So that's how we're broadly looking at it.
- 19:15SPEAKER_00: Um, there are ways to gate it.
- 19:17SPEAKER_00: I say, as with any investment, there's always risk.
- 19:19SPEAKER_00: There's always risk of leakage in this case.
- 19:22SPEAKER_00: Um, you want to minimize that rather than trying to get an absolute zero leakage.
- 19:27SPEAKER_00: But there are ways to do that.
- 19:29SPEAKER_00: Um, some of them are more legal focused, some of them are more uh technology um focused, but it's it's a combination that we can do to, again, just minimize that risk.
- 19:42SPEAKER_00: Here's the here's the reality though, Alex.
- 19:45SPEAKER_00: The flip side is not really there.
- 19:48SPEAKER_00: Like you can't not be using AI to try and generate alpha as a as a as an institutional large asset manager.
- 19:57SPEAKER_00: And I would actually say it's harder for us than for a small hedge fund.
- 20:01SPEAKER_00: And the reason why is because if you're a small hedge fund, you've got you know a pretty finite book of clients.
- 20:09SPEAKER_00: You're probably capped at a certain capacity.
- 20:12SPEAKER_00: And you may be able, you know, if you're really good, you can keep charging them 2 and 20 and be very happy with what you bring in.
- 20:19SPEAKER_00: For us, we're dealing with fee compression, we're dealing with a global client base, we're dealing with everything from sovereign wealth fund clients down to retail clients.
- 20:29SPEAKER_00: We have to be able to give all of them the best investment experience possible.
- 20:35SPEAKER_00: And that requires using the frontier models.
- 20:40SPEAKER_00: And that requires figuring out ways to harness them in such a way that maximizes our edge without distilling it.
- 20:49SPEAKER_01: Very interesting.
- 20:50SPEAKER_01: Yeah, maybe, maybe we can maybe maybe we can talk about that offline a little more.
- 20:54SPEAKER_01: I um I I I ask because at my company, Corbett's, we deal with this a lot with other customers with our customers, right?
- 21:02SPEAKER_01: Like, and one of the things I'm seeing more and more, and I was curious to see if this, and maybe this is part of your solution.
- 21:08SPEAKER_01: Um, I'm seeing everything from you know synthetic data, which sort of doesn't work because then you're still kind of giving them the data, like the AI can kind of figure it out and then get trained, um, all the way to just pure like wall garden on-prem deployments of open models, which I think is very interesting and a good solution for especially if you're really trying to keep everything in-house.
- 21:30SPEAKER_01: Um, all the way to like a hybrid solution where like the frontier models are used for certain tasks and planning and things like that.
- 21:38SPEAKER_01: And they're probably not even given synthetic data or even parts of the data, but maybe like just used for some kind of high level plan.
- 21:44SPEAKER_01: I'm not really sure, but like there's ways to map it out and then have like a hybrid model between um between like open source models that you're training on your data and keeping in and also that your customers in inference is going through because again, you don't want to send that information to.
- 22:00SPEAKER_01: Anthropic either.
- 22:03SPEAKER_01: And then that brings it the point of like how important it is to have a modular system in place that can support, you know, swapping in these things.
- 22:12SPEAKER_01: Because the other problem we deal with, and I and I bet you I know you're seeing this too, is like this space is moving so fast.
- 22:17SPEAKER_01: And so like new component comes out, new model, new routing technology, new harness, like whatever.
- 22:23SPEAKER_01: You need to be able to take the best thing off the shelf and like plug it in.
- 22:27SPEAKER_01: And so that's hard to do.
- 22:44SPEAKER_00: So I was gonna ask you about that.
- 22:46SPEAKER_01: Yeah.
- 22:47SPEAKER_00: So in car racing, when we have, you know, a an a live event, we're probably gonna deal with some parts of the car needing to be replaced as the race keeps going.
- 23:00SPEAKER_00: Um, but not the chassis.
- 23:02SPEAKER_00: I mean, if the chassis needs to get replaced, I probably also need to get replaced because something very bad has happened.
- 23:07SPEAKER_00: But if a way to think about that in this analogy is the chassis are really the durable IP components.
- 23:12SPEAKER_00: It's our governance policies, it's our 80 years of investment knowledge, it's the clients' objectives and constraints.
- 23:20SPEAKER_00: Those things should really not change between each session, right?
- 23:25SPEAKER_00: And then there's the things like the tires, the brakes, the arrow.
- 23:30SPEAKER_00: Um, you know, if you've ever seen like an F1 race where the front nose gets damaged, they just take it off, snap a new one in.
- 23:38SPEAKER_00: Yeah.
- 23:38SPEAKER_00: That's the that's the modular components.
- 23:42SPEAKER_00: And so for us, that's gonna be everything from orchestration and semantic layers to data vendors to um the large language models themselves, all of that we are building to be able to swap very fast.
- 23:56SPEAKER_00: So if you know we all of a sudden want to move from Astra to JEV for a task or combine them in some way, it should be just a few lines of code to uh switch that, or even a smart router, which just like arrow on a car knows when to turn it up or turn it down.
- 24:14SPEAKER_00: It knows when to switch the models, uh depending on the question that's being asked of it.
- 24:19SPEAKER_00: So that's how we think about that modularity.
- 24:22SPEAKER_00: It's a really, really one listening.
- 24:25SPEAKER_01: This is the correct answer to the question.
- 24:27SPEAKER_01: This is this is the way you do it.
- 24:29SPEAKER_01: We're very big believers in this too.
- 24:31SPEAKER_01: I mean, all the all the stuff we've built at my company is open source software.
- 24:36SPEAKER_01: It's like an agentic operating system, is how we think of it.
- 24:39SPEAKER_01: And uh it's exactly what you're describing, right?
- 24:42SPEAKER_01: Like modular components that you can just sort of like you can you can use the whole system, and that's what so the the OS is called interchange, it's open source.
- 24:50SPEAKER_01: The commercial version is Corbett's, and Corbett's core is like you can think of it as like Linux is the open source, and Red Hat is like an opinionated version where we've packaged it in a certain way.
- 25:01SPEAKER_01: This is our opinion on like how you should use this, but you can you can take this stuff over here and just cherry pick like whatever you want.
- 25:08SPEAKER_01: And the reason this is so important is all the things you just said, right?
- 25:12SPEAKER_01: Like, you need to be able to move quickly, just like that F1.
- 25:15SPEAKER_01: I love the F1 analogy.
- 25:16SPEAKER_01: It's exactly correct, right?
- 25:18SPEAKER_01: Like you we are all literally driving around in this like marathon F1 race and getting wrecked every day.
- 25:25SPEAKER_01: And getting wrecked is like model goes down, open source thing gets hacked, like whatever thing, drama of the day, new chatbot thing comes out that I need to replace it with.
- 25:37SPEAKER_01: I mean, it's absolutely insane.
- 25:38SPEAKER_01: And and so for businesses out there, I mean, you're one of the I'm so excited because like you guys are one of the few companies I've talked to that is actually understanding like an architect architecturally the right way to set this up so that you can adapt and and and and deliver solutions much faster, right?
- 25:56SPEAKER_01: Like that's the that's the answer to the case.
- 25:57SPEAKER_00: You can see that if in you know, like in this F1 season, you saw that there were certain teams that started at the back and moved to the front because they got new components and they put them in quickly.
- 26:07SPEAKER_00: It's the same thing here.
- 26:08SPEAKER_00: Like, I think right now we're one of the leading asset managers of our size in how we use AI.
- 26:14SPEAKER_00: I'm I also have no illusions that couldn't change tomorrow.
- 26:17SPEAKER_00: And so I operate every single day.
- 26:20SPEAKER_00: Like I could be overtaken by my competitors if I don't keep every bit of my effort on staying as fast as possible.
- 26:29SPEAKER_00: There is no good enough, there is no done.
- 26:33SPEAKER_01: I have another qu I have another question for you, but I it does, but but before I ask it, um where can people follow along on what you guys are doing around this?
- 26:42SPEAKER_01: Because like, is there a blog or a re or is anything open source?
- 26:46SPEAKER_01: Like are things uh I just think it's like very interesting and exciting what you're building, and I'm sure some of it's in-house and just kind of quiet right now.
- 26:54SPEAKER_01: But like if people just want to see what's happening at Franklin Templeton around AI, what do they do?
- 27:00SPEAKER_00: So, one thing you can do is go to futureportfolios.com, which is uh one of the microsites we made where we talk a lot about a lot about how we're using AI to change the investment process, how we're using digital assets and tokenization.
- 27:14SPEAKER_00: Um, also, of course, uh you can follow me and Franklin more broadly on LinkedIn, uh on X.
- 27:22SPEAKER_00: We just launched an Instagram, so we're really spreading out our social footprint.
- 27:28SPEAKER_01: Let's get this up, let's get this up there.
- 27:30SPEAKER_01: We'll get you some content.
- 27:31SPEAKER_01: Let's make like a thousand clips.
- 27:32SPEAKER_01: There's good ones.
- 27:33SPEAKER_01: There you go.
- 27:34SPEAKER_00: Um so yeah, so we're we're we're we're broadening out, but you know, I I think traditionally it's true that the juiciest stuff is always going to be undercovers, right?
- 27:45SPEAKER_00: For a certain extent.
- 27:47SPEAKER_00: Um, but we've been.
- 27:49SPEAKER_01: I just think people should pay attention.
- 27:51SPEAKER_01: I think after this conversation, if you heard this podcast, like you need to be paying attention to what your team is doing at Franklin because it's it's it's not the norm for a lot of companies that I talk to.
- 28:04SPEAKER_01: It really isn't.
- 28:05SPEAKER_01: Um and uh it's it's it's exciting.
- 28:09SPEAKER_01: Um so my my next question is more like zooming out macro.
- 28:15SPEAKER_01: Um, I feel like, you know, I've been in crypto, I was in crypto for 10 years.
- 28:20SPEAKER_01: Now I'm building an agentic open source agentic operating system.
- 28:22SPEAKER_01: I'm kind of straling both these worlds like you are, like digital assets and AI.
- 28:27SPEAKER_01: And uh I've seen these smart these market cycles many times.
- 28:31SPEAKER_01: And right now we're in one where there's some euphoria, but there's also some fear, I feel like still.
- 28:38SPEAKER_01: Um, and I hear this narrative.
- 28:40SPEAKER_01: There's sort of two sides.
- 28:42SPEAKER_01: One is we are entering into a world of abundance, and AI is gonna make a world of abundance in this utopia.
- 28:48SPEAKER_01: The other one is like uh AGI is gonna kill us all.
- 28:53SPEAKER_01: Um, and uh the data center build-out and the level of debt that's being built around that and the leverage in the system is gonna collapse and we're gonna see a huge market correction.
- 29:04SPEAKER_01: Which side of the fence are you on?
- 29:06SPEAKER_00: So I I I don't I think as with a lot of things in investing, it's really a question of horizon.
- 29:14SPEAKER_00: So I think, first of all, circular financing is an enormous risk.
- 29:19SPEAKER_00: Anyone who says, don't worry about circular financing, I don't think really understands how a leverage bubble works.
- 29:26SPEAKER_00: That doesn't mean it ends tomorrow or that you shouldn't be invested in the names that are part of that circular financing or even at the at the very core of it.
- 29:36SPEAKER_00: I mean, NVIDIA just launched a half uh trillion dollar financing prop platform with every single blue chip PE firm.
- 29:45SPEAKER_00: I still like NVIDIA as a name, I think.
- 29:47SPEAKER_00: Nvidia's got a lot of great quarters ahead and really good performance ahead.
- 29:52SPEAKER_00: But I'm not blind to the horizon.
- 29:56SPEAKER_00: At some point, the lenders are gonna be due, and there is an increasingly fragile foundation that happens when you do this.
- 30:06SPEAKER_00: We had a parallel in the dot-com boom where uh Nortel and Lucent were basically lending to their own customers who ultimately couldn't pay because we didn't need all of that fiber.
- 30:17SPEAKER_00: It's very different right now, famous last words, because we see massive continual demand for GPUs.
- 30:26SPEAKER_00: So NVIDIA can you know keep this party going, but it is a risk.
- 30:33SPEAKER_00: And the things that I'm looking at are very traditional indicators, like day sales outstanding on their balance sheet, right?
- 30:42SPEAKER_00: Like, are they seeing any kind of decline in earnings quality?
- 30:45SPEAKER_00: Are they recognizing earnings that are not really there?
- 30:49SPEAKER_00: Like, don't look at the stated numbers, actually really look through them and analyze them properly.
- 30:56SPEAKER_00: That's again where you need a professional asset manager usually to do that.
- 31:01SPEAKER_00: But that's how you start seeing when the canary in the coal mine is gonna start uh freaking out.
- 31:09SPEAKER_01: Well, in the canary in the coal mine, you brought that up.
- 31:12SPEAKER_01: That that word, it was that phrase was ringing in my head today because uh I don't know if you can tell, but I've got like matte face right now.
- 31:19SPEAKER_01: I was at jujitsu before this.
- 31:21SPEAKER_01: And I was coming out of jujitsu and I saw this article that Oracle triggered their force mesure clause in their data center project.
- 31:29SPEAKER_01: Um, and it's the data center project in New Mexico.
- 31:32SPEAKER_01: Um, and that force majeure clause is if they they they were supposed to pay higher rent, I guess.
- 31:37SPEAKER_01: Uh, and if so if they fail to meet their 2028 timeline to like launch the data center and start obviously generating revenue off of it, then they can do this force majeure clause.
- 31:46SPEAKER_01: And I was like, oh, is that it?
- 31:48SPEAKER_01: You know, like of all of all the companies I've seen, like I think I agree with you on NVIDIA.
- 31:54SPEAKER_01: Like there's gonna be a correction, their stock might get hit, but like long term on a time horizon, they're they're just a good bet.
- 31:59SPEAKER_01: Um Microsoft, again, Google, like the Facebook, like these big names.
- 32:04SPEAKER_00: Hyperscalers are, you know, like if we if we think about the hyperscalers and the companies around data centers, there's gonna be winners and losers too.
- 32:11SPEAKER_00: By the way, like with all of this, again, another big misconception, I think, is that it's just a monolith.
- 32:16SPEAKER_00: We still have the same thing in crypto, right?
- 32:18SPEAKER_00: Everything was a monolith, and now savvy investors are really differentiating between, especially layer twos, and they're really thinking more about this as an asset class with idiosyncrasies, it's the same thing in AI.
- 32:31SPEAKER_00: Whether you're looking at semis or you're looking at the complete GPU makers, of course, there is really one major one still, um, or you're looking at hyperscalers where there is a lot more differentiation, and then you're looking at the foundational model makers, there will be winners and losers.
- 32:49SPEAKER_00: It's important to be very aware of who is in the lead and also be aware that that lead could be toppled.
- 32:57SPEAKER_00: And I again, I mean, Jeb is a good example this week where you you can, and you know, before that, we had like last year of uh the Deep Seek news, right?
- 33:05SPEAKER_00: Which wound up being a little hyperbolic, but it certainly caused a major correction.
- 33:10SPEAKER_00: You know, we had the SaaS pocalypse um earlier this year.
- 33:14SPEAKER_00: So you're gonna see those sorts of rolling periods where some sector um gets corrected.
- 33:21SPEAKER_00: I just don't see it as a systemic risk to fear.
- 33:26SPEAKER_01: I don't think the SaaS pocalypse is is is systemic.
- 33:29SPEAKER_01: I think it is gonna it's killing SaaS companies that don't adapt.
- 33:32SPEAKER_00: So I would actually push back on that, Alex.
- 33:34SPEAKER_00: And and here's why.
- 33:35SPEAKER_00: Uh to me, it's a very simple reason from now having been at a couple of very large companies, it is very, very difficult to replace an enterprise player like a workday or a Salesforce.
- 33:52SPEAKER_00: Even if you have a much better, cheaper solution, the switching costs are enormous.
- 33:57SPEAKER_00: And so if I if I think about that from like one of the Frontier Labs perspectives, what's better for me to go and try to compete with, say, a workday and try to get all of their major enterprise clients to switch?
- 34:10SPEAKER_00: Or do I go to workday and say, hey, we can make you guys have way higher margins and deliver better customer experiences if you use our models.
- 34:23SPEAKER_00: I think the latter is actually a better approach.
- 34:25SPEAKER_00: And that's actually good for workday stock because their margins go up and they retain and in fact increase their customer footprint.
- 34:32SPEAKER_00: So I think that's more where the SAS companies are not all gonna be taken out of existence.
- 34:37SPEAKER_00: However, they're again winners and losers.
- 34:40SPEAKER_00: But companies that partner with the Frontier Labs in a smart way, I think are gonna do a lot better than the companies that say, no, we're good.
- 34:47SPEAKER_00: Thank you.
- 34:48SPEAKER_00: We're like Blockbuster, everyone's gonna keep coming to the video store.
- 34:51SPEAKER_01: Yeah, and they're gonna get crushed.
- 34:52SPEAKER_01: And I I actually I agree with what you're saying.
- 34:55SPEAKER_01: I Mike, my addition to what you're saying is that the way to win there is to make your software adaptive.
- 35:01SPEAKER_01: Like, I and again, this is from customer displaying.
- 35:03SPEAKER_00: Brings it back to modular harnesses, by the way.
- 35:05SPEAKER_01: Right.
- 35:05SPEAKER_01: 100%, right?
- 35:06SPEAKER_01: Like every single customer, enterprise customer I've talked to that's adopting AI right now, they're looking at every single SaaS product they have and they're like, why are we using this?
- 35:17SPEAKER_01: Why am I paying 100k a year for this random system that doesn't actually that like I'm forced into?
- 35:25SPEAKER_01: It doesn't have the this X feature that I want, and they're gonna charge me like a ridiculous amount of money to add that feature when you know scrappy company like us can come in and like rebuild that whole SaaS product in literally like a day and then add the new features to it, and they're willing to pay the same amount.
- 35:42SPEAKER_00: Well, and by the way, uh probably six months to 12 months from now, people will be able to build that product on their own with you know, like two shots.
- 35:51SPEAKER_00: It's uh the the speed with which you can build things if you have domain knowledge in the frontier models right now is absolutely unreal.
- 36:04SPEAKER_00: Like earlier this week, I had this idea, like you know, like this is what I dream about.
- 36:08SPEAKER_00: I was dreaming about thinking about risk from a thematic perspective instead of a factor perspective.
- 36:13SPEAKER_00: And I built a thematic risk model in Claude over like three or four hours.
- 36:19SPEAKER_00: And it works, it's performant.
- 36:21SPEAKER_00: I mean, I didn't edge test that I'm not gonna put it, you know, anywhere near client capital until I've actually done a human code review.
- 36:28SPEAKER_00: But like I probably got 85 to 90% of the way there in four hours.
- 36:32SPEAKER_00: And if you've rewind the clock even six months, I would have needed a team to work on something like that for maybe a month.
- 36:42SPEAKER_00: That's how that's how fast things are moving.
- 36:45SPEAKER_00: And again, the key is I know exactly what what my idea was, I know how to articulate it, I know how to you know, nudge the model as it's building so that it stays on track.
- 36:56SPEAKER_00: But you know, I haven't written production code in over a decade.
- 37:02SPEAKER_00: And now, and this is by the way, you see a lot of tech executives and a lot of finance executives gravitate to these models for this reason, because a lot of us who like came up doing the grunt work are going, like, holy crap, we can do the grunt work again, but with the knowledge base that we've developed over the years being executives and seeing where the industry is going and seeing what succeeds, that is super powerful.
- 37:28SPEAKER_00: That's the thing that's the real transformative edge of AI.
- 37:32SPEAKER_01: So um, I always ask this at the end of the show, we're coming up at the top of the top of the recording here, but what have I not asked you that I should have asked or that you wanted to talk about?
- 37:41SPEAKER_00: I mean, I think we could go a little deeper on what is it going to be like to work with AI agents?
- 37:48SPEAKER_00: And and and and it's an important topic and a relevant topic for for me as because we've actually built AI teammates.
- 37:55SPEAKER_00: We have one in production right now.
- 37:57SPEAKER_00: Um, and it's been really fascinating seeing how my humans interact with it.
- 38:03SPEAKER_01: I love I love how you put that.
- 38:06SPEAKER_01: How my humans interact with it.
- 38:08SPEAKER_01: Uh yeah, please tell let's talk about it.
- 38:11SPEAKER_01: Tell me more.
- 38:12SPEAKER_00: So so so basically, this this came up in an investment committee uh last year where this was when I was still chief investment officer.
- 38:20SPEAKER_00: And I was like, you know, whenever I have an investment view, no one disagrees with me.
- 38:24SPEAKER_00: And I know it's because of my title, not because I'm the smartest guy in the room.
- 38:28SPEAKER_00: Like I don't have any illusions about that.
- 38:30SPEAKER_00: And I also know that I'm wrong, like any other investor, quite often.
- 38:35SPEAKER_00: So I was like, okay, let's build an agent that's just gonna like call me out on my BS.
- 38:40SPEAKER_01: The adversarial agent.
- 38:41SPEAKER_00: Yeah, exactly.
- 38:42SPEAKER_00: No, that's exactly right.
- 38:42SPEAKER_00: It was a nice, that's the title of the term.
- 38:44SPEAKER_00: Um, and we built this agent, and then the agents kept getting smarter because we were feeding it more and more data, both internal and third-party data.
- 38:55SPEAKER_00: It was coming into our meetings and getting live transcripts, and it got smart enough where we were like, hey, like you can actually be more than just an adversarial agent, start thinking on your own, and then start talking back to us in the meeting live.
- 39:12SPEAKER_00: And then, hey, can you do independent research if we ask you to do something?
- 39:16SPEAKER_00: And then, hey, can you help me with uh a deck I'm building for you know a client or an internal investment committee meeting?
- 39:23SPEAKER_00: And so we've started adding more and more to this agent until now it's really is a coworker.
- 39:29SPEAKER_00: Like we just think of this agent as another member of the team, we ask it questions, it gives us answers.
- 39:37SPEAKER_00: You know, we ask it to do independent research, it comes back with really thought-out responses.
- 39:41SPEAKER_00: We push back on it, it pushes back on us.
- 39:44SPEAKER_00: Sometimes it'll actually agree with us now, sometimes it'll still call us out on our BS.
- 39:49SPEAKER_00: But what's happened is when it's been really fascinating, is the diff the realization that this is an agent is only apparent when people forget that it's an agent and they ask it like a personal question.
- 40:07SPEAKER_00: And it's like, did you forget?
- 40:08SPEAKER_00: I mean, this literally happened recently.
- 40:10SPEAKER_00: We had um one of our futures associates, which is like a rotational program for younger professionals at Franklin, said, Hey, I'm gonna miss you.
- 40:19SPEAKER_00: Uh, as he was leaving for another rotation to our to our agent, and the agent said, Well, I understand why you would miss me, because I'm really great at giving you all this analysis and always being someone you can bounce ideas off of.
- 40:32SPEAKER_00: However, I will not have any context of this particular conversation, so I cannot honestly say that I will miss you.
- 40:38SPEAKER_00: I was like, man, that is just hard.
- 40:41SPEAKER_00: That is that is really, really hard.
- 40:45SPEAKER_01: We have like a whole Slack channel at Corbett's that's like dedicated to this kind of thing.
- 40:50SPEAKER_01: Like our some of our favorites are are when it when it clearly does something wrong.
- 40:54SPEAKER_00: But yeah, I I I did feel bad for our futures associates.
- 40:58SPEAKER_01: Yeah, it goes, it goes, you have you have caught a real inconsistent inconsistency, and it's mine.
- 41:04SPEAKER_00: Yep.
- 41:05SPEAKER_00: Yeah, there there are there's a lot of music on your right to push back on that.
- 41:10SPEAKER_00: I'd say our agent is a little more savvy than that.
- 41:14SPEAKER_00: Um but absolutely in it's it's I would say high EQ, zero EQ.
- 41:22SPEAKER_01: Yeah.
- 41:23SPEAKER_01: Yeah.
- 41:24SPEAKER_01: Well, look, Max, it's been awesome having you on the show.
- 41:26SPEAKER_01: I would love to check back in in like six months or something and see, you know, that you're probably gonna be 10 years ahead of everybody by then.
- 41:33SPEAKER_01: Because I think that's the thing that people don't understand is like when you when you have the right framework, the base layer as a strategy across your company, this stuff compounds over time and your intelligence stack gets stronger and stronger.
- 41:47SPEAKER_01: And if you own that stack, especially if if you're you for us, that means like using open source components uh primarily, um then you really own that intelligence.
- 41:58SPEAKER_01: And like no no one can take it.
- 42:00SPEAKER_01: Like if something happens with Fable, government yanks it away, you can point it at an open model, and it will still work.
- 42:06SPEAKER_01: Um because you you you've built a an F1 car for agentic work at Franklin Templeton.
- 42:13SPEAKER_01: So thanks for sharing that story.
- 42:15SPEAKER_01: I appreciate it.
- 42:16SPEAKER_00: Thanks a lot, Alex.
- 42:17SPEAKER_00: It's been great to have be on.