Navigating the Wild West of AI in Finance: Policies, Pitfalls, and Opportunities

Actually surprised at the industry, how it responded. So

again, like you're mentioning back in 23 when we were the first

people to kind of try to broach in this space, my background in the

industry told me that hey, this is a very narrative industry. You know, they don't

like to take risks, they don't want to expose data. And so initially we had

our data policy of we don't pull financial data, we only push data into

CRMs. And then all of our competitors who have tech backgrounds came in and because

of their tech background, it's all, all data is good. So they started flooding everything.

And I thought that hey, the market wouldn't like that. I was very, very

wrong. The market is completely fine with AI running roughshod all over

client data and I did not expect to happen. And

so we are having to change our internal policies now to reflect the market

reality that actually that's not the case

anymore. The finance industry is pretty wild

west right now with AI. Why?

Well, hello and welcome back to Data Driven the podcast. We explore the emerging industry

of and field of artificial intelligence, data science all back in

Dubai. Data engineering, however, my

favorite is data engineer in the world is dealing with

a dental appointment and he apparently has

double the amount of front tooth teeth now that he did this morning. So that's

was the message. So I'm like oh dear God. But fortunately Candace was

ready to jump on in. You may recognize Candace from our two

other podcasts, Impact Quantum and Women in Quantum. So welcome

to the show, Candace. Thank you, thank you. Glad to be here. Awesome.

So today we have someone I know we've been trying to get on the show

for a while, Daniel Yu, who is a founder and

CEO at Finmate AI which is a custom

agentic AI for financial advisors and he

himself is a former financial advisor. So he definitely has a lot

of experience in the fintech and he also spent some time not

that far from where I live in Hopkins. And

do I see Poolesville, Maryland?

Yeah in your link? Awesome. I,

I used to, we used to live in Darnestown which is.

Oh yeah, yeah, yeah, yeah. That's cool. That's awesome. Yeah, it's always good

to see a local, local kid do well. Thank you, thank you for

those, for those don't know. Poolesville High School is kind of a big deal around

here and one of the. My wife and

I were aiming to move to Poolesville at one point. So

but we ended up going north. We're up in. We're up closer to Baltimore now,

so. But that's awesome, man.

That's cool. I was at the second

gen of the mag program there. So we were the second year of smacks.

Yeah. So by the time I was a senior it became number one in the

state. But before that it was kind of a nothing school.

Yeah, well, Poolsville. Poolsville also suffered because, well, you know that

you would know, you know that there's a ferry there. But that ferry is now

shut down. So for those who don't know, I know we kind of go off,

we go off rails, we do this. So the D.C. metro

area has grown probably way more than it was ever designed to. And they

never kept up with the bridges and infrastructure to do it. Imagine that.

So one of the big. One of the, one of the ways to get across

from Virginia to Maryland used to be the. Called White's Ferry.

And it's been around since I guess some guy was started

to ferry right after the Civil War. And basically as

a cable ferry, there's a cable on the Potomac and it's only like what, 200ft

and you just drive your car on it, a motor pulls

you across and then now you're in Northern Virginia, which

was a much more pleasant way sometimes to get to say

Ashburn or Loudoun county or Dulles Airport

from where we used to live than it is. But there is some drama.

A long term 50, 20 to 30 year legal case

had finally come to pass and the ferry shut down. And

it's kind of been a disaster actually. But I know a lot of businesses in

Poolesville benefited from all the traffic that would go through. Because if you didn't want

to spend all your time on the American Legion Bridge

with 100,000 of your closest friends, you would at least have the option of

taking the ferry. Yeah. Long winded way of saying welcome to the

show. It's cool to see somebody from Maryland do well.

So you obviously have a background in

financial finance. What made you get into AI and

kind of the startup founder role.

You've done that a while ago. Right. So

you make the move when you did? Well, I went to Cal

Berkeley and we have our incubator

at Cal and everyone that I know kind

of has one leg in tech. Right. One foot in tech. Even when I

was Doing wealth management. All of my clients were tech people, generally speaking.

All of my mentors post graduation were all startup

founders. And I actually been told for a long time to

quit finance and do startups already. And I resisted for a long

time because finance was stable. And then post TD

acquisition of Schwab. So yeah, might be a good time to try

it out. So that's when I hopped over

those, those acquisitions that you get in finance are very, can be very

unpleasant. I was, I was there. Well, this is probably before

your time. I was there at Banker's Trust when they got bought by Deutsche bank

and that was an interesting experience to say the least.

But yeah, so finance is a fun field to be in until you get

acquired and then it becomes not as much fun.

So. And obviously I guess given that you're in the Bay Area, right, like everybody

and their, everybody and their dog now is into

AI. So

what exactly does, what problem does finmate solve? Like what, what,

what made you look at that, the problem space and say I can do

this better than what people are currently doing? Well, when we first

there were no AI note takers for advisors. There were the generic AI

note takers out there. But again this was in the era of GPT 2.0. So

the context windows are very, very small. The outputs

on these notetakers are not very detailed and suitable for the advisor's

requirements. And so we came up with some post processing tricks to

basically extend out the token limit to

actually get good details on the notes. And then six months later

a couple folks copying us came out and then now there's 30 of us

out there. And so we made the announcement about six months ago. Hey, we think

note taking is a bit of a commodity at this point. So we

drastically cut costs and then moved to Hntki

development as well as obviously with the speed of development

increased by AI coding, we're now starting to release a

lot more point solutions and so we'll be releasing like a scheduler

to go on top, basically taking our learnings from some of the custom

builds and commodifying a lot more things.

Interesting. So what is the big play for agentic

AI? Well, agentic AI is a very loaded word. A lot of people have

different opinions of it. So for the, for the interest of this

conversation, how would you define an agentic AI system?

Put simply, just an AI that is given access to complete

tasks. I'm currently just viewing it as like a junior

employee potentially. Right. It does a Task or two.

I think if you. Obviously there's people that just give it full

access to everything and then just let it run. And I found that that's not

really stable or consistent enough in a business context, business

operations context in particular. And so we're still

giving it hard parameters and, you know, hard inputs,

because I don't think it's quite there yet. But it can do a lot

of menial operational tasks that I think can be automated away.

So I'm doing it more as Zapier plus than anything. Okay.

So for those who don't know, Zapier is a very common

automation tool that can kind of. Kind of like N8N kind of

sort of. Yeah, yeah. It's the previous version of N8N. To be

fair, Zapier is coming up with, you know, doing their own version of AI

Pipeline. And I, and I was rude because I cut off. Candace, I'm

sorry. No, no, no, no worries. It's all good. So you talked about like, the,

the, the repetitive tasks, the

menial tasks that the agentic AI can do. So what

is. Is still too sensitive to fully automate

in this kind of. In the sector that you're

working in. Yeah. So I'll give an example from a few

months ago. It's, you know, when you're filling out paperwork,

let's say, and you're filing things. Right. Account opening forms and things like

that. Different custodians have different forms. And so one layer you can do it

is you have all of the client information and you just tell the AI, fill

out this form. Nowadays it does look a lot better, but back

even a couple months ago, you had to be a lot more specific and

field put the form

inputs into the AI so that it would recognize it

properly. It's just giving it a lot

more context for specific tasks as opposed to just

assuming you can do a whole class of tasks by itself.

Interesting. So you have to give it real guardrails and not

guardrails, but kind of direction. Correct. Yeah. And

it's getting faster and faster and better and better. And so

the breadth of which you can kind of entrust it is definitely growing month

by month. But there's still areas where we're testing and we're finding like

it's a matter of what is an acceptable fit of the rate. Right.

And you can kind of modify how detailed the

instructions are based on what you're okay with.

So you're selling this to your. Okay. To your customers.

I'm just trying to kind of think about this A little bit. And then your

customers are using it for themselves

and on their own clients. Correct,

correct. Most of it is for internal operations actually.

Yeah, sorry, say that again. Most of it is for

internal operations and so these, our clients are basically using

this to kind of automate away internal tasks. Obviously that supports

their client relations and

paperwork and things like that, but the clients generally don't directly see it.

Okay. Do you think that changes the relationship at all

between the. Between the

client and their advisor,

our tool? Not particularly, although we

are releasing basically

on our Notetaker platform. Obviously this is where we're presenting a lot of our point

solution developments as opposed to the custom agentic AI development.

We'll be releasing basically a birthday tracker, anniversary

tracker that'll let you then generate emails based

on the CRM context of it. I think AI can

help help basically have advisors

be in more points of contact with the client and. Or because it takes less

lift to be able to make these touch points. And so in

that sense I think it enhances a bit. But for the most part on the

agentic side, most of it is the background operation stuff. So maybe

they'll see turnaround times and support times be a little faster. But beyond that, I

don't think it changes the tuition differently.

So finance is obviously a very heavily regulated industry.

What does that look like? Because you're trying to push the cutting edge.

Yeah, that's got to be both exciting and

absolutely terrifying and possibly very expensive.

You know, I'm actually surprised at the industry,

how it responded. So again like you're mentioning back in 23 when we

were the first people should kind of try to broach in this space.

My background in the industry told me that hey, this is a very derivative industry.

You know, they don't like to take risks, they don't want to expose data. And

so initially we had our data policy of we don't pull financial data, we only

push data into CRMs. And then all of our competitors who had tech backgrounds came

in and because of their tech backgrounds it's all data is good. So they started

pulling everything and I thought that hey, the market

wouldn't like that. I was very, very wrong. The market is completely

playing with AI running Russia all over client data. And I did

not expect to happen. And so we are having to change our

internal policies now to reflect the market reality that actually

that's not the case anymore. Finance industry is pretty

wild west right now with AI. Really?

Yeah, I'm surprised. I guess it's all fun and games.

It's all fun and games till somebody loses money, right?

Until FINRA decides to find somebody. But until then,

I made the play. We're like, hey, we're going to limit certain features because that

requires us to pull all the data in and copy over to our system, things

like that. And yeah, no, people just don't care

anymore. There are some large corporations that are

building stuff in house because of their concerns and those companies definitely exist,

but by and large it seems like they don't actually care right now.

Wow, that boggles my mind. Yeah, it

boggled mine too. And so we were a little late for the punch and transitioning

over to this new free for all data environment. But you know,

market has spoken.

So what where do you

do? Because one of the things I've noticed in my day job is that

the financial services companies, the larger ones, prefer to

run their AI workloads for all of those data reasons

on prem in their own data centers and metal that they control.

Is that something that one, that you've seen widely and two, is that something

that you, you support or is it, you know, the

cloud? And you know, obviously not everyone's going to have

access to those types of resources. Yeah, there's three layers,

right. So there's some companies out there that were like

oh no, we have to build it all in house because we don't trust anyone

else. So there's definitely those groups out there and when

I talk to those kind of big brand name kind of bank

associated financial advisory firms,

they've interviewed not just like the industry specific note takers like ours, they've also

interviewed all the generic ones as well. And they decided hey, none of this fits

our data criteria. So we're going to build it in house. Some are okay

with cloud, we do offer on prem for the custom agentic

development side and we are in certain talks with some companies

to basically white label Arnotaker for their platforms as well.

And so it's been a mixed bag and there's

just not been any consistent reaction so far.

Interesting. Yeah. So if an AI

agent could proactively surface

opportunities or risks before

you notice them, how would that change

the way you make your

decisions? Yeah, that's also

been kind of standardized right now as well in this space, in

the advisory AI space, some people would call

it like next best action. Right. We kind of incorporated

into our morning digestion where

obviously it depends garbage in, garbage out, it depends on if the advisor is updating

their CRM properly with the proper opportunities,

opportunity sizes and things like that. But it can Check

into, hey, here's some outstanding activities, here's some outside business, or

here is a contact that you haven't talked to in X number of time

and they're considered an A tier client. The

AI can then surface these things and frankly speaking, that doesn't really require

that much AI Really. A lot of it's just programmatics. Right.

You put in the filter parameters and then set

it on a cron job timer. And then the system

just lets each advisor know, hey, here's probably what you should be looking at.

And so actually, funny thing is, as we're developing, realizing

this actually doesn't need that much AI. And so

yeah, we're programmatically building kind of the next best engine

instead. Because no actual need for AI in this case. Yeah.

So what would that mean specifically for you? So like

a well behaved AI agent, What would that look like in your world?

What boundaries would it need to respect

to feel aligned with your practice? Yeah, so

the design parameter basically is

anything that is auditable.

Anytime an AI agent touches an auditable system before it actually

makes changes to that auditable system, there needs to be given the loops there

every time, let's say for the note taker side, the AI takes the

notes, presents the notes to the advisor before it gets pushed into the

CRM, which then it is an auditable client

log. The human needs to verify that, hey, this is correct, and then

proactively pushing it. Right. Same thing for like emails. Right. Any client communication

is auditable. Right. And so before a communication actually,

or an email actually goes out to the client, the agent can save it as

a draft, let the advisor know, hey, this is in your draft folder. If you

approve it, then you can send it. Right. And so there's no full proactive end

to end in any kind of auditable system.

Okay, interesting. How did, how was the,

you know, with all the fear about white collar jobs going away?

Sure. How, how do people

respond to this tool? Right? I mean, is, do they see

that as an augment, you know, an augmenting tool. Right. You know,

I'm trying to think of a good example, Candace, but I can't think of one.

The one I always go back to is in the Aliens movie where

RIPLEY had the, the giant exo suit, Right. Where

she could pick up like stuff that normal or fight the alien queen.

Spoiler, sorry, Movies

ship had sailed decades ago. But, but I mean,

I mean it didn't replace her per Se but because she was using that

mechanic mech suit, she was able to do more.

That's how I see AI. It's really an augmentation. But not everyone

is, dare I say, as enlightened as me or

humble, of course. And so

like how is. Because I know financial,

the financial world is filled with characters, that's for sure.

Yes. What's the general zeitgeist? Obviously you're

going to have individuals that are going to be all in on this. There's going

to be ones who are very not

into this. Like what's the gem where, where's the curve fall?

I think for advisors it's still augmentative. Right.

Because at the end of the day someone is licensed, meaning someone's taking on the

liability risk. Right, right, right. So that's one side of things.

It's like I think you'll find this in like the medical field and the legal

field as well. Right. There's a license for these things because someone needs to be

able to get sued. Right. And the AI is probably

not liable. Right. And so that's one layer. The other layer

is frankly speaking, if you wanted

a robot to do it, you could already have a robot do it. I mean

we have robo investors for a reason. Right. So that ship has already kind

of sailed as well. And so people that don't really care about not having a

human already cannot have a human. Right. So the only people with

actual human advisors are people that want a human looking after their stuff.

So I think that's not too much of concern. I think the only area of

concern potentially would, would be for the pipeline.

Right. The assistant associate to

the wiser pipeline where

because a lot of these operational stuff is just getting automated away,

some of the juniors that might in previous generations have a foot

into the industry by helping an advisor do their tasks, that

pathway is not closed. Right. And so I think there might be some concern there

with that pipeline much in

the same way as like software engineers. Right. There's no one hiring any juniors. What

happens in 10 years when those juniors are supposed to be mid

tier seniors but they don't exist because they never got hired as

juniors. Right. And so I think there's some concern around that. But for the

advisors themselves, I don't think any of them really are concerned.

Yeah, I mean that makes sense. Right. And that's why you have a lot of

mid level and senior developer types that are not that

concerned. But you're right, like are we shutting off the pipeline that

we're going to Pay for this a decade or two down the road

because as people choose to retire, et cetera,

et cetera. Like we're not going to have, we're not going to have that

backfill of talent that we've always had. Speaking of the

software industry, right. Like that is. We won't

know it's a problem until it's too late. I suspect it's probably the same

in the finance realm. I mean to be fair, when

we look at the finance industry and the client base, right. I can generally

divide into four buckets, more or less. Right. So you have

ultra high net worth, high net worth, mass affluent and then the rest, right.

The rest generally don't have financial advisors right now anywhere. And if they do,

you know, they're probably using robo advisors. Matt Affluent also, same thing,

usually are self directed or using robos. If they have

advisor, they're one of like 800 clients. The advisor servicing.

Because mass affluent, they're below a million in investable assets,

right. So the advisor is not really making any money. You get to the high

net worth, 1mil to 10mil assets invest,

that's not very complicated to deal with, to be honest. And so there's no real

skill set needed to be an advisor. Speaking as a former

advisor myself, just to be clear, I'm not

dissing people. It's not, it's not intellectually

that rigorous, right. It's like you see one case in a high net worth

individual, it's the same thing. It's not that complicated. And so

I don't think that human capital training pipeline is an

issue, at least in this industry because you

can train someone for like a year maybe kind of just shadowing how you manage

your book of business and then they should be able to pick it up, no

problem the next year. You know, we have plenty of people coming into the industry

after spending 30 years in some other industry. So I don't think that's concern. I

think the only area of real concern would be the ultra high net worth individuals.

10, 20, 30 million plus investable assets. That is very complicated

because all these dates and taxes and all the regulations, but that needs to

be continually kept up with, right? And so potentially a

concern there. But for the vast majority of industry I just don't think the human

capital is going to be a problem because the real

differentiator I think in the industry is can you generate a book of

business or not? Right. Can you get people to give you money to manage? If

yes, great, you're off and running. If no,

that's not an intellectual issue. That's a human issue. Right.

Sales issue. Yeah. Kim's

fascinated by this. Sorry. No, no, no, don't. Can we go back to the

idea of the next best action? I kind of.

I don't know, it's kind of sitting with it. And I'm

thinking about the balance between

the AI recommends. The next best action

is that guidance versus intrusion.

I'm not really certain how it would be intrusive because again, all of the system

that we're setting up, the AI is not proactively reaching out directly to clients.

It's surfacing up opportunities for the advisor. This is a human in a loop step

that I think people are starting to talk about a lot more in kind of

AI agent design. And so

the advisor ultimately picks. Yes, no. Yes, no. Right. Like,

here's a client anniversary coming up here is the client birthday coming up here is.

Oh, you know, in your previous meeting with your client, he

mentioned that his daughter had a software ball game that you can

maybe write an email and ask about. Right. And they can generate these

emails for the advisor. Right. Like, hey, happy birthday, you

know, appreciate your part of us. Right. And then answer something that they mentioned

that's part of their note log. Right. I mean, could.

Could the advisor do that himself? Yeah, probably take a lot more

time because he has to read through and. Right. And so I think it's just

kind of shortens that gap. But then ultimately, if the advisor doesn't want to send

the email, then the advisor just doesn't send that email. Okay. So like, because

you have the human in the loop and let's. You're given a

recommendation, then the human in the loop can choose to ignore it.

Yeah, okay. Yeah, yeah. The other thing that, you know, some firms were talking

to us about was they have. If they do like holistic

planning, right. There's different financial areas. You have insurance, you have, you

know, estate planning, you have all of these other areas, like a

checklist, per se. Right. And then they wanted the AI to kind of go through

the client log and see, did the advisor talk to the client, each client

about all of these, let's say eight topics. And if they didn't,

maybe the advisor before, or maybe the agent before the advisor's next meeting

with that specific client would say, hey, you didn't talk about XYZ topics.

Consider bringing it up. And then if the advisor tells the AI, oh,

they're actually not interested, then it's like, okay, great, check it off. And then

don't need to worry about that in the future. Okay. Yeah,

that's very. Yeah, it's an assistant. Yeah, yeah. I mean, I mean,

but I mean you're right. In, in, in old days

of old, like you would have a secretary or like an assistant that would do

this right. For you and be like kind of do this. But no one,

unless you're very. Unless you're in the corner office, no one gets an admin of

their own anymore. Right, Correct. Yeah. So this is kind of like I,

I was talking to somebody about this the other day. It's kind of like the,

I don't want to say the Uber vacation because that's not really the right word.

But the whole idea of like with Uber, Lyft, etc, you, you

get a private driver. Sure. Right.

Days gone by, only the very wealthy had private drivers.

Right? Yeah, right, right. Like a

commodification, like so, so now anyone really with an Uber account

and you know, can have a private driver. Right? Yeah.

You know, when I go on business trips, 9 times out of 10 I'm not

going to rent a car anymore because rent a car is a pain when I

can just, you know, call up. Well, actually, when I was last in the Bay

Area, I rode Waymo a couple of times. That's great. Yeah, that

was an interesting experience

I got. It was. The scary part was when it couldn't make a

left turn. So I kept going. I went around the Moscone center

like three times because of traffic. And at some point I'm like, I got

to hit the help button and you know, eject.

But they're testing now. Yeah, yeah.

So I mean, but it's one of those things where I think AI is going

to give the masses kind of an experience or

concierge level services that ordinarily would have been

reserved for, you know, the very well off. Oh, absolutely.

Yeah. Whether it's your own private secretary. I'm sorry,

no, it's not even private person. Right. But even

for. It's like commodifying skill sets. Right. This is the concern of a lot of

like engineers is software engineering used to be a very

lucrative position even for entry level. But now that's been

commodified way and we're seeing it internally as well. We're

using a lot more AI coding than we did a year ago and that's allowed

us to with a very small team. We haven't done any fundraising. We're now competing

with people that have fundraised $100 million plus

it's allowing us to compete with them. Because

we are an AI first company. We do all our development through AI now. And

so that's letting us develop all these software tools that would have previously taken years

for people to build. No, I mean, it's true.

Like, I'll do a little commercial for what Candace and I are working on,

right? When you have three podcasts, you start realizing that there's

things that can be automated away, right? And you know,

on my whiteboard behind me, I would always have like, you know, notes or

back of the napkin or whatever on software to build to make things easier.

And it just, nine times out of 10, those would not be built.

Right? But now with AI, it's just a matter of like, right now I have

an audio conversion tool being vibe

coded in the background. You know, it's been

working on it all day. And it's funny how quickly we adjust to it. First

it went from, wow, look at what this could do to oh, my God, it's

taken like 40 minutes. What's going on? We're so impatient.

It's ridiculous. But I mean, in the past, this is something that, like, you know,

I'd be like, well, you know, I have this idea, then I'll sit down, then

the kids will need to do something, and then, you know, like, just life gets

in the way. But you're. I mean, to your point, like, I.

Every one of us has access to a junior

level developer, right? And if you use other tools, whether you

string together agents, whether it's OpenClaw or,

you know, Squad or whatever, the flavor

of the day is, you potentially have an entire team

of junior level people, right, that can kind of do stuff.

And then it's incumbent on you as the manager, engineering manager, to make sure

that it's being built correctly. Speaking of which,

one of the things I was surprised by actually is as we started offering

these custom agent AI developments, we started running

the competition from dev teams from overseas.

So as much as we think that, oh, hey, like, prices have gone down here,

apps have exponentially gone down kind of in overseas, you know, code

shops, software development shops. And so, yeah,

it's been looking more and more like a race at the bottom

in terms of development cost. Because even. Because the thing to

do previously, right, if American engineers were too expensive, you'd go

overseas, Ukraine, India, wherever, Pakistan, and try to hire

engineers. But know, we're like, okay, well, with AI development,

we can get a lot more productivity out of American engineers. But then the same

thing's happening with overseas talent as well. So I don't know, the quality

still. Right. I think potentially we're, you know, stacking

spaghetti code on spaghetti code. All right. Because I think

overseas engineering has been known to not be architected

properly for scalability. And then the concern then is if

the folks that are not really focused on architecture are also vibe

coding on top of it. Right, right, right. There are scalability

concerns down the road, but you know, for some small projects, maybe that's not a

concern. Right, so. Exactly right. Like, and I think that the

smart, the smart play is to, to take all the low risk stuff.

Yeah. And bump it out to AI, even AI

with. And even, even if the engineering folks overseas

are, you know, on power, on parity with, you know, the

North American developers. Right. There's just certain cultural assumptions that

you have in being here. Like I remember when I worked at this

one company and they, they were all excited because they were,

they were basically. India had become too expensive, so they out

offshored their stuff to China. But it was all about car insurance.

And the whole notion of car insurance and legal liability

and how litigious we are here in the States just did not compare

compute to them. Right. It had nothing to do with like their capabilities

as engineers. It's just they did not, they really lacked the cultural

context. I mean, this is, this is going back a ways, I think, I think

like so like the whole notion of, you know, even though people had private

property, but like the whole notion of it of like, wait a minute, you can

get sued. What does that mean? Like, well, you know,

there's just so much context that if

you, if your butt is in a seat in North America, you

do have a lot of understanding of that. Whereas whether you

were born overseas or, or, or born here, like, you just have that understanding. Whereas,

like if someone has never left their home country, they may not

know. Right. You know? Yeah,

So I always thought that was interesting. That always fascinated me.

And then here we are with AI, Right. Talking tech. Right. Like we're talking about

not just prompt engineering, but context engineering as a

real interesting different way to look at it. Right. Because an

AI is going to also may not have all the context, particularly

in a specialized space like finance. Right. And you mentioned

it too. Like you, you did a lot of early on in

2023, which is like stone age here. Right?

Frankly, yes. Yes, it was pretty bad. Yeah, well, yeah.

And you know, it's funny because like, you look back and you're like, you know,

November. I remember it because my first experience with ChatGPT

was I. I was at the Vegas

airport leaving Reinvention. And I had actually,

this was the year I left Microsoft. Right. So like, for me to go to

reinvent at all was like heresy. And the fact that I was presenting,

working for another company and using a MacBook to do it was just

completely the inversion of my world, you know, 10 months prior.

And then I'm at the airport and I'm like, all right, I keep hearing about

this, but let's see what it is. And I can actually have not a

coherent conversation, but a pretty decent conversation. Way

better than the old techniques of natural language processing.

But again, you look back at it now, it's like, oh, how quaint. Right? And

that was only, I mean, four years ago,

three and a half years ago. It's really

amazing how quickly our expectations adapt. Yeah. Kick can

go into college freshman year, it's like, I'm going to major in computer science. And

by the time he meets, it's like, oh, wait, there's nothing available.

There's nothing there for me to do. Right. Yeah. It's kind of nuts. Yeah.

I actually have a kid who is, he's a

sophomore in high school and. Oh, wow. He,

you know, he was, he was an AP Computer sciences, things like that.

And. And then he opted to not do like the second

course for computer science. Right. And I was like, why'd

you not do that? I was like one, I was kind of upset he didn't

talk to me. But Candace tells me that's very common for teenage boys not to

just talk at all, which is true. And

so I asked him, I was like, why didn't you not go with.

Yeah, because the teacher called us like, you know, he's one of our best students.

I get the dad. I get the dad, Max. Right? And boast.

But. And, and then I asked him, I was like, what's going on here? And

then he goes, I want to take physics instead or so AP physics instead. And

I'm like, I can't argue with that. You know, like,

it's not like you just didn't want to do anything, it's just.

But I think that, no, you're right. Like within a four year time span,

it's very risky. You go into college and you expect to

make. I mean, people don't look at colleges, that

people should look at college as a financial transaction.

Right. They ought to. Will they? I don't know. But like

if you're gonna drop, I don't know, let's say

$200,000. Right. On a college education, which is

very easy to do. Yeah. You should at least

be making at least a hundred thousand dollars a

year obviously, because you're not going to keep all that money. But you, you know,

you can see the end of the road for when you're going to pay that

off. Right? And I just

can't imagine. No, that's not happening. Except

for a very few and it's not when they come out of college

because you're not going to get that at college. You've got to go get the

masters, go to B school, you know, it

depends whatever it is. Right, like. Yeah, like I said

though. Yeah. No, I'm sorry. Okay, but like you said,

go ahead, we do this all the time. No, but

it's not uncommon for if you're at a really good computer science

program and you get into a really good big tech, you could get that. But

you're right, it's not a guarantee.

Sorry Candace. No, I should say. Did you hear, I think it was yesterday that

Brown decided that they're going to be a 100 grand nugget now

starting next year. What's that, like a year? Uh

huh. 100 grand a year. Huh. So

400 if you do the whole four year thing. My

God, that's, that's a nugget to choke on.

I mean, yes. Oh my God, I can't imagine

that. Like, yeah, that's a bad investment.

I mean overall, yeah. I mean, no knock on Brown. But like think

about the money you have to make getting out of that. Right. Like, and you're

right, you're right, Candace, you're not going to walk out the door with a six

figure salary. It can happen, but it's not common. But at

least if you go, I mean, wow,

I mean, maybe they're gonna teach you really good

discipline on entrepreneurship because that's the only

way. Or it's just gonna be trust fund kids. I mean, that's really, I mean,

right? Realistically, yes, because entrepreneurship. Right. I mean, you know

startups, 99% of startups fail. Right. So that's not really

a viable stable option for most people.

Like man, I can't imagine brand name

schools like computer science. Like I will say if you're

good at coding, you're still going to get a good job. Right. Because. Right.

It's, it's very evident that AI still needs

a lot of importing. And if you're a talented engineer, you're a talented engineer.

But because of the prevailing wisdom of the past three years,

has been, past six years, let's say, has been just go get a CS degree.

I think a lot of people that are just kind of phoning it in, hoping

for an easy cash out are the ones that are struggling right now. Yeah,

no, I think you're right. I think the whole Learn to code movement,

was, you know, Learn to code will raise you and your

family up from poverty into, you know, the, the good life.

Yeah, I should have paid attention to that back when I was in middle school.

You know,

I mean I, when I, when I, I'm old enough that, you know, my

first computer was a Commodore 64. And you

know, when I got it, I originally wanted it to play games because it was

a, it was, it was, it was a good gaming

rig at the time. Yeah. And I remember after, you know,

we grew up kind of poor and I remember after buying the computer I would

ask my parents like, hey, I want to get this game. They're like, how much

is it? And they're like, my mom laughed at me and she goes, are you

kidding? And she's like, well, why don't you write your own games?

And much to their credit actually for other reasons. I have it

on my Desk. Every Commodore 64 shipped with one of these,

which is basically. And kids, this is before

Google, before even Yahoo was a thing. Before aol,

before aol. Aol I think existed. But you, it was not

everywhere. But yeah, just like military Internet. Internet.

Yeah, like it was not. Yeah, and, but like I

would read it and like in, in one of these chapters there was a whole

thing on how to do animation. So like I was very much self

taught, I went to college, I had to convince my parents that computer

science was a valid career path because they were like, you know,

because in those days it was doctor, lawyer, engineer. Right. And

then if I wanted to get. Yeah, right, right, right. And my, my

dad had a saying like, you know, don't get a BS degree. Like as in.

Right, right, right, right. And like if you do that

then, you know, because we're not, we're not going to help you pay for it.

We're not going to co. Sign loans. Like they're like, if you want to do

that, like, you know, those are my options. So I had to convince my folks

that, you know, not only was it a viable career path,

but like computer Science was a viable engineering discipline.

I know there's some. There's debate about that, you know,

but. But, you know, but I remember, and this is way back before

there was dice.com kids or LinkedIn jobs. So I had

to, like, bring. When they visited me at school, I had to show them, like,

the Sunday New York Times, which had the job postings, and it was like this

thick of, like, the paper was on Sundays was about this thick.

Candace grew up in Newark. She knows what I'm talking about. And it was the

best bargain because it was only like, a dollar. So you had, like, reading for

the entire week. But the job section on Sundays was, like, this thick. And

about half of it at this point in the early 90s was

computer jobs on Wall Street. Right, right.

Typing. Yeah. I mean, if we're going to reminisce again. So I

went to Clemente as well. Okay, I know Clemente.

Yeah. I was second year for this program at Clemente as well.

And so, you know, we're doing True Basic. You know, we're doing WebDef or

Notepad. Right. Our animation was like

marquee tagging. Right. And yeah, this was

like, when we were, like, learning about different search engines. Google was not the major

player at the time. It was like altavista, Ask Jeeves. And,

you know, I hated debugging because the IDEs were not very good

back in those days. And I'm like, I hate to buy this code

so very much regretting that decision. But, you know, here we are.

For those not in the dmv, we call it here. District, Maryland, Virginia,

Roberto Clemente. I think he's talking about the middle school, which was one of the

original Magnus schools. I think it's still Roberto Clemente in Springville or

Springbrook or something like that. It's in German Town. Or

Gaithersburg. Gaithersburg. Clemente is in Gaithersburg, and

there's another one in Potomac. I think the other

one was the og. I think they overflowed, which is

why they made Clemente. And so I was a second year to go to that

place, Tacoma Park, I think. Right, that's it. It's Tacoma Park.

Yeah, Yeah, I remember. Yeah. And then it was Blair into Poolsville. Yeah,

yeah. So. So for people who are like, what the hell are they talking about?

So if you think of, like, if you look at a DC map, the little

top of, like, the little square that's like a diamond, that's where Tacoma

park is, more or less. And then Poolesville's Way out to the west,

it was really the sticks. Poolsville is interesting because it was like an island. It

was a landlocked island because there was like literally,

Literally you could drive for like 30 minutes and it's just farms and then all

of a sudden. Yeah, I would drive by

cornfields and cow farms on my way to school.

That's pretty wild. But no, I

mean, I mean you're right. So this is interesting. I want to be respectful of

your time and things like that. So I want to make sure we land the

plane. So where can folks find out more about finmate

and what you're up to and you personally?

Yeah, so easy way to find us is our website finmate AI

f I n m a t e AI. You can follow us on LinkedIn as

well. Again, what we're up to is one custom

agent AI. If you want some builds that basically fit your

existing tech stack. We're exploring a little bit

beyond just traditional wealth management. So we're talking to insurance folks as well. Now

then, in terms of our platform product itself, we're

starting to commoditize a lot of different products so hopefully

we'll save everyone a lot of money down the road. That's cool.

I think it's very exciting. I thought the way you explained it was

fantastic and I'm very interested in

finding out more. That's great. Perfect. We're trying to

do some mastermind classes as well if you just want to chat about AI

and how to think through Agentix. So be able to look after that on our

website as well. Oh, very cool, very cool. I do see you have a consulting

tab on your website too. I presume that's probably falling now. I think that's cool.

I think there's a lot of people that could benefit from learning AI, particularly in

the finance space. Right. Yeah, we'll see how

that shakes out. Because I'm surprised that they've been so open minded.

Now I know that's kind of like stereotyping, but in

my experience fintech financial firms do

tend to be very much aware of the cutting edge but they're very, also

very risk adverse too. So it's always been this weird kind of

mix of. I was always lucky because when I was at, when I was at

Merrill Lynch I was on the special projects team and those, those guys got

all the cool stuff. Right. So like

everything but access to where the quants hung out. Like they, they had their

own, they had their own floor and stuff like that. Yes, yes.

But like folks like Merrill though are big enough that they can do

development in house. And so, yeah, yeah. Merrill is one of those exceptions where

they were big enough to kind of do their own thing. And I know now

they're owned by B of A, which I can't

imagine that was. That was a. That was cold

water in the face when that happened. But. But, yeah. So cool. I'm

excited about this. I always have a start. I started my career in.

In finance. I worked for a startup that was in Germany that was. We would

call it a fintech company today. Sure. So there's always a soft

spot in my heart for this. And anytime you want to come back in the

show and you want to show something off, just. Just let me know. Sounds great.

Cool. Awesome. And we'll let the outro music play.