[Automated Transcript] 

Ryan: [00:00:00] So the macro facts, one, advisors are aging by every day, like the advisors as the overall industry that's not growing its numbers, meaning their advisors are aging out and dying off, and the numbers are getting smaller while the population in general of needing advice is increasing.

So there's a natural gap there, 

Deep: why can't they just hire more young, fresh grads like that? 

Ryan: They're not doing it. young people are not seeing it a pathway for a career at a rate that they're falling off on the other end.

you may have five, graduates become financial planners, but you've lost 50 on the back end from where it was before. It's just, I think it's because it's an industry where. As a 24-year-old kid, it's impossible to really successfully sit in front of some money and say, people, your parents' and say, let me manage your money.

Like, that's a tough thing to do. 

Deep: Hello there. Welcome to your AI injection. Today I'll be talking to Ryan. George Ryan's leading an AI initiative leveraging 23 years of wealth management data to build digital teammates that automate advisor onboarding, compliance checks and end-to-end account [00:01:00] processes.

Their stack uses multiple LLMs plus some computer vision for doc handling all orchestrated so that tasks like opening accounts, profiling risk and training can execute with audit trails and human override thresholds. Ryan, thanks so much for coming onto the show. 

Ryan: Thank you. Deep. I'm happy to be here.

Deep: Awesome. So. Why don't we get started tell us what does your core user do exactly, like, maybe walk us through a specific use case? 

Ryan: Yeah. 

Deep: So we have ideally with before they've got your product, and then we'll dig into how your product makes it different. 

Ryan: Sure. Before they have our product, generally a financial advisor in an office, you know, meeting with a client somewhere out in the world we'll gather information from a client and they'll say, look, we, let's, let's work together.

Um, and so they'll gather, a stack of papers that they'll have to go through and sign. All manually, all in a very sort of analog way, takes quite a bit of time and often things are missed. they'll mail, like physically actually mail, [00:02:00] the paperwork into, you know, a home office.

And then the home office will go through it. Something's wrong. These. Send it back, and you've got this sort of back and forth in what we call like a Nigo, which is not in good order. It gets rejected. that's our sort of, we have two main, I would say stakeholder camps. Those are the advisors and their assistants in the field as well as the people who are approving business and processing business in the home office.

And those two would be our power users before they use a tool like docu base. 

Deep: let's just kind of jump up even a higher level. So financial advisor, you are surrendering your account access to someone who's gonna go in and like, make trades and stuff in your brokerage or 401k accounts and all that kind of stuff.

So, makes sense that there's all this regulatory framework to make sure that they're not, siphoning the cash into their pocket or somewhere else. What's the other user? 

Ryan: So it's the home office. So think of, uh, if you were to go into, say, an Edward Jones or Raymond James office that down the street from your house that person works with a corresponding person at a centralized office.

Uh, in Raymond James [00:03:00] Case it's in Florida. In Edward Jones case is often, it's in Missouri. When they, they're the ones that actually process and approve the business. The advisor gives the advice then what we call the home office person is the person at the, okay. 

Deep: This is the whoever the financial advisor works for?

Ryan: Yes. The, the HQ. 

Deep: Got it. And so is, so that initial client setup customer relationship, that paperwork, is that your main focus or is there, are there other aspects of the, like once they're in an established relationship? 

Ryan: Sure. So once they're in an established relationship, the other piece is the maintenance.

Um, so that actually requires because of the regulatory constraints, like they have to send a mail a piece of mail every three, six months to make sure that the address is still valid. And there's all sorts of things that are required. So our platform, actually, probably 75% of the usage is maintenance.

Well, 25% is probably account opening. 'cause once the account opens, you can work with an advisor for 30, 40 years. 

Deep: Got it. Okay. So maybe well anyone, let's walk through it. A a, a new user signs up for a financial advisor [00:04:00] service. What is that nature of that initial paperwork that they're sending in? Is it just like bank account information authorizations?

Yeah, 

Ryan: that's piece of it. So each, so any, the basis of any sort of financial advisor relationship is gonna begin with sort of gathering a risk tolerance. So gathering information about, well, you know, what is your income, what is, you know, taxes you pay, your preference for investing or saving or, and also what your goals are.

That creates a risk tolerance that allows the investment advisor to, or financial advisor to figure out exactly what's the right choices for you. So what they provide options, like whether it's some insurance or whether it's some investment advisory account, or whether it's what they call brokerage account, which is more for trading.

Um, and so that's the, would be the next step. So once that's decided upon to, between the advisor and the client, um, then they basically create a. Account paperwork. Um, and so they will put all that paperwork together. It includes, like you said, driver's license, but a lot of other information that's required, like beneficiaries are a big important piece to that [00:05:00] as well.

Then the other thing that I think happens a lot is a lot of times the advisors are working with people who have money elsewhere. So like, say think of an old 401k for a company, no longer work for that can get rolled over into a single account so you can make it easier to manage. So that's a lot of the work they do that, uh, comes with a number of authorizations.

So one of them being, like you said, trading authority. So you can have twofold relationship with an advisor. One is they can trade for you without your authority or they would have to recommend a trade and you approve it before it goes through like that. That's something that would be included in that big packet of paperwork as well.

Deep: Got it. And then the risk profile, I'm assuming they don't have a sort of consolidated dashboard or do they so like clients that might be on one of many different brokerage, uh, firms. There's like a kind of a lot of different potential places that their information's coming from or is it kind of a hybrid environment where they have some kind of central platform that does some stuff and they're like slowly plugging in to like Fidelity or a [00:06:00] Morgan Stanley or whatever.

You, 

Ryan: you are wise man Deep, there is no single dashboard in this business. There's people who have been chasing and trying to build one forever, but like you said, money lies in all these different places. And also a lot of times people have assets that are nons securitized, meaning there it's a painting or some other, jewels or something that isn't traded and doesn't, so it's not necessarily have market value to really give a full picture.

And so that's, 

Deep: or like a rental house or whatever. 

Ryan: Yes. And then there's what they call alternative investments, which are private placements, again not traded. So it is almost nearly impossible to create a single dashboard. Um, at this point I think we're getting closer to that. Data is definitely much more.

Available and feeds are much more available today, but it's still, it's still in these silos. Like who's the custodian? Is it Fidelity? Okay. It's really hard to get out of Fidelity data outta Fidelity or Schwab or the others. 

Deep: Yeah. And you have just this very heterogeneous environment where everybody's got their own APIs, they expose what they wanna expose.

Maybe they wanna force their users to come in through their, [00:07:00] your user experience and they don't want it all API accessible. Maybe somebody does want it all API accessible. 'cause they're like a, you know, more tech front 

Ryan: company, you know this well. 'cause a not all a P APIs are the same, uh, for one or two.

Sure they'll take the information, but they want you to come to them. Right. So like they want you then to be the home base. I think another big part of that sort of initial relationship that's important that our system does is validating that it's a US citizen. Um, there's all sorts of fence in laws and other financial laws where there has to be a known customer.

Within the United States, if it's a, a United States citizen outside of the United States, you can get things done, but it has a little bit more steps to it. And so the know your customer, anti-money laundering laws are a key thing that our system helps keep people compliant with. 

Deep: That makes sense too , that the feds are like using the financial advisors as a way to keep track of what's going on in the 

Ryan: Sure.

And it'd be the same. It wouldn't be, be no different if you're trying to do a bank account. Or any, anything else. Any other, 

Deep: oh, yeah, yeah, yeah. That's true. Like all of them [00:08:00] are subject to the same body of laws. And I, since you're managing money, you have to as well. Okay. So like, and then I imagine, okay, so you've got probably.

This initial interaction probably every one of these advisor companies has their own tooling too, to like do the risk assessment, for example. Yes. Like they have a back and forth quiz and, and maybe to remind them to go and get a revised risk assessment and, you know, and like when to talk to them and all that kind of stuff.

I imagine they're trying to fill in some glue between all of these harder to integrate with pieces. 

Ryan: There is, and that's one of the challenges I think, first of all, I think a lot of people don't do, what you just said, which is ongoing. You know, life's changed. If I work with a financial advisor for say 10, 15 years, there's a lot of life that changes during that time period.

So, you know, you should go back and make sure you understand that. And and that's one of the things that I think the system can help with. we also integrate with more than 30 different technologies. There isn't like a single consolidated tech stack they have in something in the business [00:09:00] called All in One.

It's really not all in one. It's a most in one. but there isn't a technology that exists that doesn't need some level of integration with some other technology. 

Deep: Got it. So is your goal to like I'm imagining for a while now, maybe a decade or two, there's been players that build some kind of platform for financial wealth advisors that integrates, you know, albeit flawed with these external systems and tries to provide some kind of consolidated worldview.

Is your goal to like enable more efficiency in those existing platforms or to be one of those platforms? 

Ryan: So we wanna enable this. The first part, we want to enable efficiency as much as possible. We don't wanna be everything to every, we're actually coming up with developing new solutions where we are either behind the UI so somebody can bring their own.

Their own UI to like, and basically create their own workstation and they can plug into our system through APIs like you mentioned, to be able to open new accounts through, through us. Um, and that's, we want to be agnostic [00:10:00] as possible. The reason for that is we think the integrated tools are better than standalone tools.

Um, and we think it's sort of lies in the nature of the back office, which is the people that we normally work with, those home office users that I was talking about, like that they're sort of behind the scenes as well. So we think it's that's sort of what we do. And so we don't mind being behind the scenes.

We do have a nice UUI that we and user experience that we have redone over the past 18 months or so, but if that's something that doesn't fit, you know, a lot of people wanna own the workstation and so we have developed ways that we can work with them, even if they wanna own it. 

Deep: So your primary goal is to power these platforms with AI capabilities that like make that address a lot of these inefficiencies that are right in place today.

Yeah. And that in order to do that well you probably have to build a platform too, um, to be able to like sort of understand all the use cases, but your goal's not really to sell the platform itself. 

Ryan: Yeah. Think of like, we can build this and you can plug into it, or you can bring your own and we can plug into it.

Or if you wanna use Salesforce, right. So Salesforce is where a lot of [00:11:00] people want their advisor or their worker to live. So we can plug into Salesforce and that's, that's something that I think long term will, will benefit us in terms of, uh, people, because a lot of people are choosing Salesforce.

So let's figure out a way to catch that. Catch onto those. Yeah. 

Deep: So let's, let's start with maybe some, or let's take, maybe dig in a little bit deeper on some of these use cases and talk about the before and after your product. So, and remind, sorry, remind me the name of the product again. 

Ryan: Docu Pace is their company.

Deep: Docu Pace, right? Um, let's talk, maybe let's start with onboarding. So you're onboarding a new customer as a financial advisor. They've got all of this paperwork, the risk assessment results, the, um, you know, their citizenship proof tell us like maybe.

Like what happens today? Do those financial advisors like contract to some firm that's got a bunch of people in the Philippines or somewhere that are like analyzing all this stuff manually? Like what's actually happening today without your solution or pre or prepays? And then tell [00:12:00] us like what's different with Docupace?

Ryan: Yeah. So without our solution, basically somebody is going through and manually validating that information. They, and, and actually mainly filling out the forms. And one of the things that our system does is I think twofold. That's critical. One is. Each type of account, depending on who it's for, has its own set of paperwork that's required.

And each firm has their own sort of paperwork unique to that. So the, just getting the paperwork together that's needed per the account is one hurdle in itself. So think of going to an office and seeing a bunch of cubbies in the assistant going one, you know, one and a two and a three and a four. Our system automatically puts those, programmed to put those bundles together.

The second thing is validation. So if you were to, instead of having to go through and write Ryan first name on 10 different boxes, you could probably type in Ryan and it would automatically fill, um, in those boxes. And then it also will create alerts of any box isn't complete. So those validations, the, I would say the bundles and validations are probably.

The number one reason why people use docu base, they, it's the number one reason that, uh, they find value is [00:13:00] that it prevents, 'cause if those things aren't correct, the paperwork application I make automatically gets rejected, right? So regardless of what they're recommending, it's gonna automatically get rejected.

Or we also have surveillance. So for instance, if it's not a fit, right? So if the if I am very safe, secure risk minded investor and the advisor's recommending putting me in something that's very high risk, you know, high volatility, um, our surveillance system will help flag that to make sure that that's not a fit with, you know, my stated risk tolerance.

So that's, 

Deep: that's later. That's not in this onboarding 

Ryan: state. That's once, yeah. Sorry. That's right. That's once it goes through. So the onboarding, so all this stuff is validated, it's packaged together and sent to back to the customer as a single DocuSign. Um, 

Deep: so can we let, can we let, maybe let's double click on that mechanics of that a little bit.

Like on your AI injection, we'd like to cover kind of three things. Yep. What it is you do, how it is you do it, and then the ethics of it, like should you do what you do. Um, right now we're kind of digging a little bit into the how so on. So you [00:14:00] mentioned that these different financial advisory offices have different forms that need to get filled.

So maybe walk us through the configuration setup on your tooling. Like do you, like, what do you have to do to configure the tool to be able to handle a set of documents and how, what does that, how does that work? Is, do you have like a smart browser that's like digging into a browser and filling out stuff?

Or do they all have APIs that you have access to, or do you like dump stuff into like a standard J format and then they like, wire it into their product somehow? Like how do you handle that? 

Ryan: Sure. So we, we integrate with a, a company called Quick which basically is a forms library for the industry.

So instead of us having to make sure that we are keeping, every provider's forms most up to date, we plug into a system that does that. Right? So that's part of the partnership. The validations are done through RBA team currently, so RBA team will, when during site configuration, um, will go through and actually map the different fields to the different [00:15:00] boxes.

That's something that with the new platform that we are have, are, have released that we, like I said, we worked on the last 18 months is to turns that from a BA responsibility and push that to the field so that the field you, the field admin can do so through a wizard that we've created. So a front frontend wizard.

Does that answer your question? 

Deep: Part of the question? It answers part of the question. So there's a third party company. That maintains like a universal repository of forms that any of these advisory services use. That was kind of what I understood. Yes. Do they give you an API to actually fill out those forms then as well?

Ryan: Yeah, so we actually, yeah, we pull the forms into our system and it's through API. 

Deep: Okay. So yes. So then that answers the question there. So you have API access to the forms because everybody's on some normalized platform of forms. 

Ryan: Yes. 

Deep: And then, um, and then let's talk a little bit about the compliance checks.

So then your system is like accessing like an array of APIs to check for specific stuff, like somebody offers [00:16:00] you an API for citizenry compliant, like if somebody's a US citizen or not. Yes. Um, or are you doing some compliance checks on your own and actually like reading the documents and trying to assess certain things?

Ryan: So we are not doing the second part, we are, do we have access to the OFAC fincen database, which basically will flag things for us. Um, it's just a, it's a data feed, I think. Think it's, it's, it is like basically like a toll charge. Like it'd go through and I think it's 75 cents or 90 cents or whatever for each time it hits.

Um, but that's something that makes it, and it will flag the account for, not necessarily reject it, but it'll flag it for review by a human in order to to, they could push it through if they, you know, it goes through manual review, but it's sort of a first check 

Deep: and the human reviews are happening by your clients.

Like the financial advisor's house has somebody that there's like an API into your platform that tells them to go review and then it Yes. Facilitates the, 

Ryan: there's, there's the whole alerting system. We, we call the, we call it each thing called a work item. It's a lame name, but it [00:17:00] makes sense. Uh, and there'll be a work item flagged that is for review.

Deep: Got it. So in the onboarding stage, like what's, what is the, or if any, is the ai, has the, has the machine learning or AI system addressed at this point? Is it like the raw information extraction? Phase to like pull stuff from content or, um, in order to like map to those forms that you're filling out?

Ryan: Yeah. From an AI standpoint, the account opening widget that we are creating, basically called basically the account opening agent. Yeah. Um, that will basically do all those things automatically, basically pull the information from the CRM and automatically populate the, instead of having to somebody manually go in and enter the first name, last name, all the rest, the information in the CRM, it will it will pull it also, if it's available in the marketplace, again, you'll find it through like the USPS mailing address database so that it could pull that information if it's, there's something missing or other pieces of data that it may be missing from.

Deep: Right. So that's like on a per document basis, you do the information extraction and then you do the schematization to like map [00:18:00] it to your form schema. And I think you said quick form. Yeah. And then there's , a determination of whether or not you maybe need human human checks or something.

Ryan: Yes. And the reason that. The system is set up the way it is, is that it, different names are in and pulling it into structuring in a way that can get pushed into different systems is the, too often the names are the names mean the same thing, but say something different. And so that's where the mapping, and where I said the BA configuration comes in is you'll have one con custodian that will have, um, I'm trying to think of an example of a name where it basically means the same thing, but they call it two different things at, you know, custodian X they'll call it something and custodian y they'll call it something else.

Yeah. And so by pulling the, and creating sort of the data layer where the data's been extracted and, and basically in that, like the schema you said, then it makes it easy where it doesn't get rejected, it could be sent. 'cause what we, I would say the core of what we offer is allowed for some multi custodian accounts.

Say if you wanna have one process, can 

Deep: you say, just a quick question, when you say custodian, you [00:19:00] mean a financial advisory office or something? 

Ryan: That would be the, so the cus. I'll go start from the client and I can work in, so there's the client that works with the advisor. The advisor is either, um, an RA themselves or they're affiliated with the broker dealer, which is a lot of our clients are.

So the broker dealer is often not the name on the, on the door, but they are the ones that like are on the disclosure that you see at the bottom of every email communication or website. They do not, aren't the custodian in most cases. In most cases, they are. They send that. The where the money actually rests is at the custodian, which is generally Fidelity, Schwab, or Pershing are the three are the three.

But you don't actually work with Pershing. Pershing becomes just the custodian that holds the asset in Secur and secure. Okay. So 

Deep: that's a general, so yeah, so the custodian is like, is like the financial house that's actually executing trades and all that stuff? 

Ryan: It is the house. That's, that's, 

Deep: and do they usually, like, does, do, does Fidelity and all these guys, do they usually have specific interfaces for financial advisors so that they're [00:20:00] like a tailored interface?

Or are they just logging in with customer accounts? 

Ryan: They do. So they are actually, oh, they're a competitor and a partner of ours. So if you are have all, if you're an advisor and you have an say an RIA and you have all of your assets with Charles Schwab, it's, there's a good chance that Charles Schwab covers everything that you need to, to do from them.

So like, use their interface. They have, oh, 

Deep: Uhhuh 

Ryan: Advisor workstation. Where we come in is often, if there's multiple custodians involved, we can we become that single process and that single interface that can be done, where the business can be done in multiple places. 

Deep: Okay. That makes sense. Okay, so back to the AI stuff.

So that, so the machine learning is like helping with the information extraction. And is this machine learning piece also like orchestrating the compliance checking as well? 

Ryan: Yes. 

Deep: Um, external APIs, 

Ryan: yes. 'cause it's not only it's not only tied to the, whether the investment or the recommendation matches the client's need, it also is tied to whether all the forms, like we have a, a permission, like a form CRS is the new regulation that comes out.

So make [00:21:00] sure that that has been supplied to the client and signed by the client. So the client, you know, can attest that they receive this information or the account won't go through. The other one would be is the license checking for the advisor. So is the advisor licensed not only to license, to sell that investment license, to sell that investment in that state, um, is another thing that's important.

Deep: Got it. So maybe can you like, describe some of the challenges you have there? 'cause that's an awful lot of stuff for the AI system to deal with. I imagine some of the documents have like, you know, maybe messy handwriting and stuff too. You probably have, like then there's the challenges of mapping the schema.

There's the challenges of figuring out how to get compliance and there's probably like no way around having a set of rules to determine exactly what the agent has to go figure out. Sure. And so what are some of the challenges that you're running into? 

Ryan: So there's a couple challenges.

One is there's the easy one is the regulatory rules, right? So like the regulatory workflow, because tons of [00:22:00] documentation. The most of the, like the SEC and FINRA and other regulators provide access to that information. So like that's part is easy. The other, the more difficult part is every firm, so there's the sort of national regulatory bodies that require certain things to happen.

Well, every firm also has their own processes and that's unique. Right. And so the challenge we come into is, comes into customization. So they wanna make the platform fit their specific, what they call like WSPs, which are written supervisory policies, the nerdy name, but it is what it is. And so that while we can really easily take care of like FINRA, SAC ones because they're universal across the board, when we get into like individual firm level requirements, it becomes more challenging because that's where they push us from a system requirement standpoint.

Deep: And what about like, I mean this is a lot of stuff to kind of test and verify, like how do you test and verify that the systems are actually working? 

Ryan: So that is a challenge. Testing has been has [00:23:00] throughout do occupation 23 year history testing and QA has always been a challenge.

'cause moving from like a coding environment to a production environment has always has its challenges as well. So that's part of, we test and then we send it to the client to test, and then we deploy it to the field, right? So we have multiple layers of testing done, both that, you know, the software side as well as the user side.

AI has helped been helpful in being able to do multiple use cases, uh, like test, a thousand different use cases and single click of a button. That's something that's definitely has expedited from a testing QA standpoint. But we still have fine tuning that I don't think we. We're human in the loop for most of those things, we still haven't really let allowed the AI to do, uh, to run on its own.

Deep: Yeah. So like I imagine with all that information extraction and field and form population, you're presenting the results of whatever the machine got to the human and the human is like going in and like verifying each and every item. Yes. Yeah. 

Ryan: Yes. Testing schema. 

Deep: [00:24:00] Yeah. And maybe like, walk us through what that feedback loop looks like.

So now they've corrected some stuff. Are you gathering that and then using that for extended ground truth so that you can like vet the next version of the models better? 

Ryan: Yes. We'll go back and, and train with the model. That's something that we're still or really, really early stage on. I think that's, that we actually don't even have that in product in the field yet.

We're still trying to figure out exactly the right process for doing so. 

Deep: Got it. And then the. You have clear, like you in theory, you know, you've, you've got a nice loop where you've got humans telling you what's right and what's wrong based on their corrections. You can store all that information.

You can kind of grow your validation and testing suite so that the next time your model builders like torque the model around, you know, you, you can sort of see how well it performed in a statistically meaningful way. 

Ryan: Yes, and that's the, what I think is unique for us, which is we have 23 [00:25:00] years of data that shows when how often the certain account paperwork is rejected and for what, for the reasons why.

And so that can help bring us up to speed faster. Yeah. But that's rearward looking, not necessarily where the industry's going. Right? So like, it's good, it's something unique to us that can, hey, maybe train our models faster and get them more tuned in faster, but we can't stop there. We're gonna need to keep pushing forward with our development.

Deep: Yeah. And on the, on the like sort of form processing side, like what models have you guys seen to be there's like sort of an array of options there, right? There's like the more modern multimodal models that are trained on imagery and text, together like the, like the Gemini series models like two five Flash and Pro.

Then there's these like, older school models that have been trained specifically for document forms processing that have, just this, they kind of predate all of the LLM work. So, but they're so, you know, and all of the big players have [00:26:00] one of these models, like I think in, in Google, it's like the document AI stuff.

It's called something similar in Azure. And that stuff has a lot of like, historic, really meticulous, detailed training and where they like compartmentalize documents. They tend to be pretty good at all that and then, you know, then there's, you know, just like models that are, then you've got size permutation differences.

Like you might go through and make a light pass, um, then you've got your information extraction problem. Then you have to like, kind of assimilate it all and schematize it. And then maybe you've got a few auditors like auditing this stuff. Like maybe walk us through, like if you can, like how do you think about all these different models with different model power capabilities and intel and reasoning abilities?

And then you have like, it you're probably dealing with a lot of images too, like to what extent are you you know, like compartmentalizing with image. Maybe walk us through some of the challenges there. 

Ryan: Sure. I don't want, I don't wanna get too far out of my knowledge set, but I, I can talk, speak to what, [00:27:00] what the product team specifically is doing.

And that's, we're working with Palantir in order to help develop the multilingual model. The other thing that I think is important is when we comes to our clients working with a lot of the large enterprises, they're doing their own AI work. They have their own AI models and so we've created the system to allow their models to plug into ours so that they're not forced to use our system.

They can use their data 'cause that is their data. Um, and they can use their models. And so that's, um, we use philanthropic has been helpful with our coding. So that's what we, a lot of the stuff with testing and other things we talked about. But Palantir has been a new a new venture for us. Um, and we just on the early stages of that, but it's something that, I think it's probably the, what I've seen our product team most excited about since I've been here for five years is, is the ability to, um, work with Palantir.

Deep: And how do you work with them? Is it just like they have a, a SaaS platform that your people use, or do they, or do you actually have like consultants on their side that help you out? 

Ryan: There's consultants on their side. So we actually had people go to, like an academy, [00:28:00] uh, that showed how to work with their system.

And they, after they completed that, then they'll do a consulting services on top of that. 

Deep: Got it. Okay, cool. So I think we covered like the onboarding pretty well. Um, there was another area I was kind of just prepping for this a little bit beforehand that sort of struck my fancy, which is now you're like, you know, you're actually like in the maintenance mode that you were describing and you know, you might have an advisor that.

It has to like execute actual trades. You have to make sure, presumably that they're executing trades that still are within the risk profile. 

Ryan: Yes. 

Deep: And you probably have options there in terms of like even what to do. Like how does that stuff work? Does 

Ryan: so, 

Deep: yeah. 

Ryan: Yeah.

There's two ways. So there's post trade surveillance, which is what we have today, which is uhhuh. Trade is submitted. Trade goes through in the, clearing process. If it triggers that it should not be valid or isn't clear, then it'll get kicked back. But , as you can imagine, there's challenges there, right? So the trade's already gone through, then you have to go [00:29:00] and unwind it. It's a very big, it's a headache. What we're the holy grail we're searching for, which is the pre-trade surveillance, so that the system can alert, can review all that information before the trade goes through, or while the trade is being processed and can flag it and reject it, then it's 'cause we have 47 or so alerts.

And there's all sorts of things that are unique that a lot of people may not think of. One of them is advisor can't make a trade in their own account before they make a trade for their client account. So like the system can help review and track for that. It's called front running. Um, they also have advisor.

Deep: Why would they have their own account? 

Ryan: So advisors have their own money that they could trade. They like if they're trading a portfolio they shouldn't, they're not allowed to trade their own personal money in front of the client's money because it can affect the value of the trade. 

Deep: Oh. Like meaning their personal accounts, which have nothing to do with their day job.

But they don't want, they don't want them to like I imagine there is potential conflicts there, right? Sure. Like a, you don't want them Guinea pig with, uh, a client to [00:30:00] like check stuff out. And so are they allowed to trail investments? 

Ryan: Um, there is, but there's generally, depending on the advisor agreement, there's generally a trailing period.

So they have to wait. They have to wait for a certain period of time. It depends on the relationship, but I think 

Deep: it is. Well, I imagine there could be like innocent mistakes made there too, where they just forgot that they did a trade. 

Ryan: Yeah. And that's why the system, a surveillance or even like a regulatory or a compliance system isn't necess often is not necessarily trying to become judicial.

It's more about trying to help people from, preventing them from getting in trouble.

Deep: Yeah, sure. And what are some other examples of, I mean, we can all think of like. It seems like some of the mistakes, most of the mistakes are not like really malicious or with any kind of mal-intent.

Ryan: No. 

Deep: They're just like a list of regulatory stuff that they're allowed to do and not allowed to do, but are like, is it possible that they just like, I don't know, miss like space to ticker and like put in the wrong ticker and that kind of thing? 

Ryan: That's possible those checks are, again, part of the trade surveillance system.

Th those are sometimes harder to find [00:31:00] unless the documentation shows that like it's written at the different fund name or different stock name was intended. 'cause there's, you know, there's stocks out there that could be RIX and then RIXX. And so how you'd have to make sure that you know, you that. That you're trading the right security.

And that goes back to the trade desk. So in these businesses, the actual, one of the biggest points of, I would say risk is at the trade desk, because once a trade becomes done, if the trader makes a mistake, which would be the, in this case, then they often are on the hook for unwinding it. So like the risk is that they can generally find a market for it.

So if they bought it a dollar, they maybe have to sell it at 97 cents. So they're, they're responsible for that 3 cent delta. But if you're doing 20 million shares, then that's, the 3 cents is a lot. Right? 

Deep: Yeah. So my guess is if I was running one of these financial advisory services, I would be really leery of letting all these financial advisors kind of pick from the [00:32:00] entire pool of potential tickers due to my risk exposure as an entity.

So do they just, do most of them have like really prescribed and constrained tickers that their advisors can choose from that fit? A priori determined risk profiles. 

Ryan: They do, and they also have, um, there's a industry regulation called the lowest cost available. So meaning that you have to whatever share class that you're purchasing has to be the, for the client or trading for the client.

It has to be the lowest has to have the lowest operational fee to it. So the product tables that, this is not what we do, but it's, we, uh, integrate with the system that does that. Each firm has its own set of product tables that's basically a giant database that allows for what is what's allowable and what's not.

Yeah. Um, and that will get that if it's not in the product table, it's not, and it's never gonna get through on the system. It won't even be allowed to be submitted, I believe. 

Deep: So does that basically mean that there, it's mostly index funds and ETFs? 

Ryan: No, I mean, there's all sorts of 

Deep: managed funds. 

Ryan: There's all sorts of things that, that's actually pretty advanced.

And that skills where the, so [00:33:00] there's the custodian that basically creates the products. There's the home office provider that decides between those products, which ones they allow the advisors to sell. And then there's the advisor who decides what recommendation to make. So e each one of those times it's the filters getting smaller and smaller.

So it isn't as big as you may think. But the original like product master that's at the custodial level is gonna be rather large. 

Deep: But generally they're not like choosing individual companies. 

Ryan: Not that most advisors do not do stock picking. That's more of a trader's mindset. But I guess they could if they needed to.

Um, a lot of the assets from an advisor to client standpoint, the clients most shares are from the companies they work for, from stock grants. They're not necessarily from going and trying to buy the next big stock that comes down. 

Deep: So let's talk a little bit about the risk assessment there.

'cause I imagine like somewhere in a client's portfolio allocation is. You know, a set of percentages of like, you know, X percent's gonna be in, I don't know, for sake of simplicity like bonds and X [00:34:00] percent in cash and X percent in equities. And then it there and then we even within there, like there's, you know, maybe an s and p 500, maybe there's a momentum version of that's a little more aggressive.

So like your system might take those, like I can imagine them doing flubs there, right? Like, Hey, I accidentally put, you know, a hundred percent of the s and p 500 index fund into a momentum version. I probably shouldn't have done that. Does, like, how, how does that checking work, because that's a little bit more sophisticated reasoning required there, like you have to A, know the risk profile B, know the tickers, C like guess that they've deviated.

And then you have to define those rules and Sure. Like, you know, a more sophisticated model. You know, with a lot of reasoning time, like a G PT five with a lot of reasoning time can do really well at that kind of thing, but you don't really know what these models exactly, how long they thought for necessarily, and there's a lot of variables [00:35:00] there.

So I'm curious how you guys walk that. 

Ryan: Sure. So we're moving out of what I would call the operations function into the supervisory principle function, which is like, which is the next step. So going back to the scenario we talked about where opening an account, so an advisor submits a load of paperwork for the client, the client signs it, it comes into the home office, the operations person reviews it.

If it's all in good order, then it goes to the next step, which is the regulatory supervision. There in. Essentially there, what you would have AI do is do a, a particular set of rules, going back to what I said about the, you know, written supervisor procedures other sort of general concepts of what the, what risk tolerance may be stated on the file versus does it match what the product is.

That's what AI can do is just basically elevate issues for the human to review. I don't think we're a long way from, from, I think people in the business being comfortable of letting AI do it all. So what they're using now is basically trying to help expedite the process and we call it remove, [00:36:00] like chasing down false airs.

Yeah. Um, that's something that is a lot of the time suck and so we're trying to help help narrow that down. 

Deep: That makes sense. So maybe we shift gears a little bit. 'cause based on what you're saying you know, I think we've, we like to cover the three things that I mentioned, like what you do. I feel like we've covered that pretty well.

Um, how you do it, I feel like we've covered that fairly well. Then there's the harder question that we like to ask, which is like, should you do what you're doing? And this isn't to say that what you're doing is overtly bad, but I think, probably you would agree that there's this second order effects of companies that can be like unforeseen, that lead to problems that, you know, the original founders of the company maybe never envisioned.

And like the example I always use is. You know, nobody really envisioned that. I'm just assuming nobody at Facebook really envisioned that getting everybody addicted to Instagram would lead to an increased risk of suicidal ideation and execution within 14-year-old girls. But that happened very much so.

So, 

Ryan: still happening 

Deep: It's still happening. It's a massive problem. [00:37:00] I don't think anyone, you know, in these social media companies really anticipated that they would be threatening democracies in the entire Western world, but they are. And so I'm curious, like what do you see as a, the like kind of straight in front of us, ethical issues that you face and think through and what do you think are the second order effects?

Assuming everything you guys are up to works out great. What do you think some of those might be? 

Ryan: Sure. So I think the first thing is by removing paper from the process, we become less wasteful as a business. That's like going back to the old days where there's literally. The files are printed, they're mailed, they're also what we have record keeping requirements so that they have to be kept around for four seven years, which means Iron Mountain is a company that basically owns warehouses all over the country that just is full of file boxes, just sitting there doing nothing.

You know, having that in the digital format is helpful. In terms of, we have long-term storage where I think not exactly the timeframe, but it's less data and energy intensive 'cause it's in [00:38:00] long-term storage that just basically sits there at rest. Yeah. Once a certain time passes. So I think that's, from an ethical standpoint, actually think we're doing good.

The biggest thing I think what we achieve for is, so we wanna democratize the financial planning process. And by that I mean people often think financial advisor, oh, they must work with rich people. And the reason that the financial advisor only tends to work with rich people is because. The time it takes to work with a client is intensive and that you gotta have a certain level of assets and a certain amount of money in order to make it worth the financial advisor's time.

Yeah. However, technology can lower that bar. And I think that's not only lower the bar in terms of who the advisor can work with, but lower the bar in terms of how many people they can work with. And I think that's something that is our focus and is in the end are seeing our long term goal, which is allow more people and more Americans, 'cause we're just American based to have access to financial advice.

If the data shows, it's very clear that a person who has advice makes better decisions. They don't sell out when the market [00:39:00] and lose their money when the they don't have the market drops. They tend to, over the long term, make much more sound, rational financial decisions. 

Deep: So let's play this out a little bit though.

Sure. Like, let's say you're successful, your machinery works, it gets really good at. Extracting information from the documents that are prepped. It gets really good at flagging things that are outside of the risk tolerances. Do you think there is a problem, I call it the Homer Simpson problem, like you've got, you know, a complete idiot ification of the operator, the human operator, because the machines are doing so well, if we fast forward five or 10 years that, this is basically the last generation of people who even knows what the heck's actually going on.

And then you just get there and they just check a box and there's a human there for the sake of the regulators to keep them happy, but they're just kind of dumber and dumber, and now the machines starts screwing up and we don't even really know. And, and individuals get affected and impacted, like, do you think there's any kind of risk there?[00:40:00] 

Ryan: So you're actually touching on what, I think it's one of the sort of third rail topics in the industry now, which is, is AI and technology gonna displace the advisor itself? Right. So like, if you look at how people are using LLMs is you basically as an advisor, right? So as an assistant who sits off the side, they can ask questions to.

I think what's important, at least in my experience in working with advisors is they're much more counselor than portfolio manager in terms of understanding the client, understanding what anxieties the client may have. And I think that to me is still and the inter pro relations interpersonal relationships with both a, you know, a spouse, spouses as well as their children, right?

There's some definitely human aspects to it. So while it is certainly possible that the advisor may be less efficient in like deep calculus math that's needed in order to do some level of, Monte Carlo analysis. I do think the EQ level will actually match that and probably even surpass that in terms of the need to pro that to rise in order to match, uh, the moment.

Deep: Okay. So [00:41:00] this is an interesting, so I'm one of those people who never uses a financial advisor and has never seen a reason to, but like, I think I'm an outlier 'cause with ai I'm like running basically like my own little mini hedge funds, so.

Sure. And I found it just in like, I mean, a long time ago I was doing individual tick or trades just for entertainment and generally losing. And of late I am basically running really kind of sophisticated trading strategies that are so easy to do now.

Like I just hang out with GPT for a while. Um, I define a thesis. Define sleeves within that thesis. Within each sleeve, I'll, come up with you know, like. Like five sub topics. So for example, a sleeve, like a higher level thesis is like, you know, like everybody else, AI is gonna change everything.

A lower level sleeve might be, I think hey energy's gonna be a huge deal. So it's my AI and energy sleeve. And then underneath that I'll have like, individual like usually I, I just pick five sub areas, so I might have one you know, be generation, [00:42:00] like power generation. Another one would be infrastructure for power, that kind of thing.

And then underneath there, you know, I'll put like X percent into like an ETF that covers this area, X percent into like individual tickers and bets. And then I'll just have the machine like recommend that percentages and then you update like, I don't know, every few weeks or whatever. Like that's,

Ryan: it's pretty amazing, isn't it? 

Deep: It's wild. I'm already, you know, I can beat spy by like, right now I'm up like 37 points over SPY. It's crazy. But yeah, like I can't imagine somebody subject to all this regulatory environment. Is even gonna be allowed to execute trading strategies like that because they're so, they'd be so hard to regulate.

Ryan: Yeah. 

Deep: You know, and you basically have to fall underneath all of the regulations of, you know, an ETF almost just to do that. , And the fact that I, I mean, to be frank, I don't really know anything about financial anything. You know, I took an economic econ class in undergrad and that was about it.

Other than that, I mean, from a professional standpoint, um, I've done a little bit of advising, like at, Morgan Stanley, [00:43:00] places like that. But usually just like, specific machine learning stuff. It's mostly just entertainment for me. It's just like fun. It's like, you know, going to Vegas seems idiotic 'cause I know the probabilities, but like here, I, you know, like I can prove 

Ryan: much more charged variables here.

Yeah, 

Deep: yeah. I mean, I can back test against, so it seems to me like if I can do that then like a regular person can figure out to stick money in an index fund and, you know, maybe balance it with like, can set up a 60, 30, 10 bonds, split. But you're right, the hard part is, is like a thinking to ask those questions b knowing to do it.

And I think most importantly, not doing something idiotic when the market crashes every time, you know, the president says something stupid. So like,

Ryan: yeah. I mean you're, I would say you're the outlier. Although that camp I think is growing in terms of people wanting to have their hands in the money, right?

Yeah. And be able to to itch. Scratch that itch of like really being able to a, be the market or, you know, be the general thing. I think that those people always exist. They've existed since [00:44:00] we, people were taking trades over the phone. Yeah, it's just gotten way, way more sophisticated and likely getting way, way better results.

I think from an advice standpoint, it often comes down to broader, like sitting across the table from somebody when they lose their spouse and advising them what to do. That could be pretty cold. If you're asking Chat, my wife just died, what do I need to do? Like, well, it may give you a list of things.

It's not gonna have the same compassion. And I think that's, that, that's taking it, taking the financial advisor out of the context of just about investments in trading and putting them into sort of whole life management. And you're already seeing this as terms of really before AI started growing, I think you are seeing the advisor becoming more.

Deep: More therapist, 

Ryan: you know, more therapists, more like, life coach in terms of, um, you lose your job, you know, what, what are the five things you need to do if you lose your job? Like that sort of thing. And I think it's not like the replacing that therapist, but I think it's more about understanding you, understanding your finances and being able to say be that sort of angel on your shoulder when something [00:45:00] may happen.

I would say an advisor makes all of their money, at least all of their value, delivers all their value and they're like five days out of five years. Right. It's just those, those small instances where a mistake could be a catastrophic mistake could be made. 

Deep: Yeah. I mean, I can see there being an incredible amount of value simply in putting a human conversation in between you and like a significant change to your portfolio.

Ryan: Sure. 

Deep: Whereas if you're like direct trading like me, I don't know, tomorrow I have a stroke or something and then I just do something idiotic. And I think having like a barrier there. Makes sense, right? Like or you get super depressed or super anxious or I don't know, you're like bipolar or whatever.

Like you can imagine just simply having to have a conversation with a human that can, like, I mean, I can, yes. I mean I I I, I never meant to imply that I don't see a value in financial advisors, actually. 

Ryan: No, no. But I do. 

Deep: Yeah. 

Ryan: I do think there's value in discu, like in pointing out the fact that, similar to a doctor, when I go into my doctor's office because my foot [00:46:00] hurts, there's a 100% chance I have already Googled and researched why does my foot hurt?

Yeah. So I think the idea of chat and other AI is, gonna require advisors to up their game in terms of the level of complexity, uh, that they explain things. And also every advisor's gonna have to understand that whatever advice they give is gonna be checked against. I mean, it should be checked against some, level of a, a agent.

Deep: Yeah. And, and, yeah, that makes sense. So 

Ryan: going back to your ethics thing, that could mean, it could mean there's less fraud, there's less bad advice. You know, who knows? Hopefully that's the case. But that, like you said, the law of unintended consequences is undefeated.

Deep: Yeah. So I had an a guest on say something interesting the other day.

I like, I always, this has been a fascinating conversation by the way, and thanks so much for digging in, but I, I wanna take it out a little bit and ask you project out five or 10 years and like I imagine your wildest dreams of the platform you're building are realized, [00:47:00] and we've already sort of started along the lines of this conversation what do you think's different?

And I want both the like, positive vision's fine. And I also want the potential like. Dystopian rendition, like what are the things that can go really wrong as well in that tenure out plan, assuming, you know, we're in a world where, you know, these financial advisors are sitting on all kinds of machinery, it sounds like, you know, part of your answer is like, Hey, my hope is that they get much more empathic and, they're showing a lot more empathy.

There's a lot more kind of in that therapy, like, you know, human emotion interaction that feels like it's in contradiction with the normal, business constraints of trying to like, let them cover a lot more people a lot more efficiently and just the sort of nature of companies and capitalist societies of trying to shrink their workforce.

Yeah, give us that vision. 

Ryan: Yeah. So I think this, it's really interesting and really this has been a fascinating conversation. It's going into like all sorts of [00:48:00] fun things that I sit alone and think about. So it's like great to talk with another human who thinks about them as well. So the macro facts, one, advisors are aging by every day, like the advisors as the overall industry that's not growing its numbers, meaning their advisors are aging out and dying off, and the numbers are getting smaller while the population in general of needing advice is increasing.

So there's a natural gap there, I think and positively. 

Deep: But why, like, why can't they just hire more young, fresh grads like that? 

Ryan: They're not, there's they're not doing it. They're not, young people are not seeing it as, as a pathway for a career in a, at a rate that they're falling off on the other end.

So like, you may have five, graduates become financial planners, but you've lost 50 on the back end from where it was before. It's just, I think it's because it's an industry where. As a 24-year-old kid, it's impossible to really successfully sit in front of some money and say, people, your parents' and say, let me manage your money.

Like, that's not of a, that's a tough, that's a tough thing to do. 

Deep: Well, where do they, where have they traditionally recruited from these firms? [00:49:00] 

Ryan: They've, there's a couple. One they've traditionally recruited out of bank channels and out of fund companies and out of, like, think of like reps on the phone. And also the large places like an Edward Jones or somewhere have recruited where they become like an assistant in the office and then eventually mature up to being an advisor, which is not a bad path by any stretch.

So there's, I think that that's the one is this age gap challenge. I think AI tools and other technology advancements can help solve that so that one advisor can now do the work of what 10 used to do on the downside. It could unlock the laziness that like we've never seen before. And I, when I, what I mean by that is.

If a financial advisor's work is done is more done by these tools what do they do with that time? Is it to serve more clients? Is it to serve them better or is it to work less? And I think if that third one, that third leg of the stool is somewhat dangerous in terms of there's nobody at the wheel right when things may happen.

Deep: Yeah. I don't know if you saw, but [00:50:00] like just, I think it was yesterday or the day before, there was a new article released by, um, in HBR with this term that I absolutely love. It's called Work Slop. Uh, I've 

Ryan: read it this morning. 

Deep: Yeah. Yeah. I mean, that's a reality. I think there's like, you know what? I think this, it's sort of goes along with the article from a few weeks ago that 95% of, you know, of AI projects fail.

But like work swap is this, just for the listeners here, uh, benefit, it's this idea of, you know, somebody has to like write a paper, write some code, produce something, aI's good enough to produce a nice piece of crap, and they do that, and then they maybe like polish it a little bit and send it off to you.

But it's not really like a well thought out piece. And apparently this workshop is like, proliferating in across, white collar America and causing all kinds of inefficiencies. 

Ryan: I'm seeing all the, I'm seeing it all the time within interpersonal communication. So like, if you're communicating with somebody and they respond with the AI [00:51:00] response and it's like way more detailed and thorough than just if they were just speaking with each other.

And so Yeah. And that's, granted it's gonna be a super user, right? Somebody who's really sort of AI forward, but that, that, I do think that, that, that lop the don't tell me in 5,000 words, tell me in 50, uh, like there's some there, there's, this data was clear in that article. Like it's a real problem.

Deep: Yeah. And it's not even, like, yeah, there's, there's the excessive verbosity problem where it's so obvious it's ai, but there's. There's like, I mean, I get all kinds of weird stuff. I mean, I, like every programmer has gotten it something from their manager or a client where they just, shoved everything into an AI system and it generates this gigantic list of stuff that's suddenly urgent.

And you're like, yeah. I mean, like, hello. Like, let's not insult people who actually know what the hell they're doing. 

Ryan: Make work. Yeah. 

Deep: Yeah. And, and every, every role has an analogous kind of a scenario happening. That's the developer rendition of it. But I'm sure there's writers and editors being handed [00:52:00] garbage left in, you know, 

Ryan: I mean, I'm a marketer by trade.

Like, AI can create all sorts of blog content that's terrible and not needed. I mean, like, there's the, as a marketer, like we have been given the keys to these tools that have made us bad at our job in terms of like being able to really be specific and have an opinion and don't be boring, right? Like those are things that AI sort of can suck out of the.

Process. 

Deep: So going back to the question about like the five, 10 year out vision, it sounds like that's one of your concerns is that we drown in work slop, if you will, from the financial advisory front, but like, I don't know, what do you, what do you guys, what do you think is the antidote to that?

You know, like how do you keep somebody engaged when the machine is doing the vast majority of what their work do you redefine work to be more along the empathy and, therapy roots or you know, like how do you do that? 

Ryan: I think that's part of it also. I think there's other things that can happen.

Like how can we push the industry forward, um, in terms of [00:53:00] product proliferation or innovation service innovation. Yes, I do think that the people who just simply process information or process applications, I do think there's long term career risk there, but I think if you can find areas to where.

Analysis and human interaction is, is required. I think it's gonna be more positive. I don't, I, I think there's gloom like between, I would say my five year forecast would be way more gloomier than my 10 year forecast. 'cause I think we're gonna go through it with significant period of disruption. Just like any other industry before we really figure out, okay, we should use it this way.

We should. I think an example would be, yeah, like my 7-year-old daughter's in first grade, right before school started, the school passed a law that. Kids couldn't have phones at school. And so that, that's like past reversing like a 25 year trend where, you know, every kid walks to school 

Deep: 17 years late, but it's, 

Ryan: but, but you know what I mean?

Like, we went through that tough period and I think now we're maybe back on the 

Deep: upside finally. Well, you know, kudos to your school, [00:54:00] but like, Seattle district hasn't figured this out yet. I mean, like, I don't, God knows when they'll figure it out, but like entire states have figured it out. Yeah.

Entire countries have figured this out, but, so it'll take time. But yeah, the data, that's the challenge here, right? Like the challenge is that we're really good at inventing very disruptive technologies as a society, but the lag. To build up societal immunity to the problems that get generated.

And that disruption is pretty significant. It's the time it takes to run a bunch of double blind studies and analyze them. It's the time it takes for society to debate all this stuff and all of the, you know, interest groups to get, like, eventually drowned out by the data. And that's, you know, it's like a 15, 20, 25, 30 year cycle.

Meanwhile, we're disrupting faster than the stuff we can keep up with. But, you know, eventually I, I, I, I like, I, why don't we leave it on this note? 'cause I feel like it's a a positive note, although I don't know that we have to leave it on a positive note, which is somebody in a, in a prior episode said something that I [00:55:00] thought was really interesting.

They said, my hope is that with all the ai, that it pushes us back into the physical meaning we've had this kind of 50 year. Run of people paying the bills and making careers for themselves by staring in a glass box all day. And maybe that was the anomaly throughout history and that we go back to building physical stuff.

I feel like maybe if I have to read between the lines on what you're saying, you're saying like, Hey, maybe it's been an anomaly that those 50 years have been about like, manipulating numbers and data and like kind of non interpersonal things. And maybe, the analogous to the physical is that we're gonna spend a lot more time on the, the EQ side of the fence, like talking and, and looking people in the eye and like maybe going back to like the 19 hundreds or something where, you know, you really sat down and really had conversations, 

Ryan: yeah. I mean, I think there's things like. What if there's never famine in the world again, because you know, the exact same, [00:56:00] the exact right crop to plant in the right type of soil for right amount of time and given based on weather patterns that have been analyzed over there's so many things that can help solve these, like giant challenges.

Like we could come up with a vaccine in, in two weeks versus two years, or, you know, all these examples. The downside of that is, are we gonna be ready to be more human when that time comes? Right? So like, yes, it pushes us all together, but are we gonna be love each other enough or to, or actually come back together?

I don't know. I think that remains to be seen. 

Deep: Yeah. Awesome. Well, thanks so much for coming on the show. This was a really fun conversation. 

Ryan: Thank you. Deep. I appreciate you having me.