March 4, 2025 • 25 mins
Dive into the intricacies of artificial intelligence in the small business lending space in this thought-provoking episode. At the bustling Money 2020 conference in Las Vegas, we engage with Omri Yacubovich, CEO of Llama AI, as we address the pressing concerns surrounding traditional lending processes and how AI could reshape the future. With small businesses representing over 55% of the American economy, access to tailored lending solutions becomes a crucial topic of discussion.
As we explore the tension between AI-driven personalization and the potential for privacy invasion, personal anecdotes highlight the real-world implications of data misuse in the financial sector. We unpack the challenges faced by community banks as they strive to maintain personal relationships with clients while also adopting advanced technology. This episode illuminates how AI offers a more proactive approach to lending, allowing banks to anticipate the needs of small businesses while fostering trust and transparency.
Join us to listen to expert insights on how the future of finance can strike a balance between efficiency and empathy, and what this means for lenders and borrowers alike. Don't miss this opportunity to engage with transformative ideas about the intersection of technology and personalized service. Subscribe, share, and leave a review to join the conversation!
Follow us on LinkedIn
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Rachel Morrissey (00:03):
Welcome to the Money Pot.
I'm Rachel Morrissey from Money2020, and we're recording live
at our show in Las Vegas, and sofar, the day has been pretty
good.
I am here with my co-host,Sheryl Chen, who is the head of
content for Money 2020 Asia.
Sheryl has spent most of thetime on the sentient stage, so
I'm just going to ask you,Sheryl, what do you think?
Sheryl Chen (00:28):
Do you think the future is actually human?
Is the future human or is it AI?
Is it AI?
Is it AI?
I think it will be a long whiletill AI can fully reign.
Yeah, but for now, I think,humans, we will still be able to
keep our jobs, thankfully,thankfully.
Rachel Morrissey (00:45):
Very thankfully.
I was thinking AI would replaceme any day now.
Well it can, and we are herewith our guest who would know,
Omri Yacubovich, CEO of Llama AI, and so how are you feeling
about everything, Omri?
Omri Yacubovich (01:01):
Good Glad to be here.
Rachel Morrissey (01:02):
Good, so you tell me what have you seen
around the show that has struckyou, especially about AI, since
that's your yeah, so I think itrelates to your kind of previous
comments.
Omri Yacubovich (01:14):
I think early on people were thinking that AI
will replace the blue coloremployees.
Now they all get it.
It's going to replace the whitecolor.
Sheryl Chen (01:21):
I think going to replace the white collar.
Rachel Morrissey (01:23):
I think, rachel and I anybody with a data
like brain dead you know what?
Sheryl Chen (01:27):
I, my co-host is actually an ai co-host, so I
actually joked with her that sheactually stole my job.
But she was so polite.
She said that you know what?
There are enough like.
There's enough space for allthree of us under the moonlight
to share the stage.
Rachel Morrissey (01:41):
She's the true romantic that, Aianna.
Okay, we're gonna get startedbecause we are actually talking
about something besides the aiunder the moonlight as they
share the stage.
Very romantic, she's the trueromantic that, Aianna.
Okay, we're going to getstarted because we are actually
talking about something besidesAI, but it is related.
Sheryl, do you want to go ahead?
Sheryl Chen (01:53):
Yeah so maybe as a first question, as an
introductory question, can youtell us the mission of what Lama
AI does?
Omri Yacubovich (02:00):
Absolutely, so we started the company about
three years ago.
Actually, here does Absolutely.
We started the company aboutthree years ago, actually here
at Money2020.
Sheryl Chen (02:08):
Nice Money2020, baby.
Omri Yacubovich (02:09):
About two weeks before Money2020 2021, we kind
of hit the nail on the problemto our after, and that's what
today I call the small businesslending paradox, meaning
everybody talks about lending tosmall businesses, everybody
wants to lend to them, but onlya few banks can actually achieve
that.
So we bought flight tickets,came over, met with about 20
(02:33):
bankers in the course of 48hours.
It's not like you didn'torganize these networking events
like you're doing in this show,so it was much harder Cold
calling, emails, whatever youwant but we did secure these 20
meetings.
And what we learned?
That not only that the problemis real, but this is something
that bankers are looking tosolve, because small businesses
(02:54):
are kind of the backbone of oureconomy.
They represent over 55%, Ithink, of businesses employing
probably around 90% of theemployees across the states.
Wow, I think of businessesemploying probably around 90% of
the employees across the states.
Our mission is to help andbridge the gap in small business
lending, helping small businessowners get access to bank-rated
(03:15):
financials and basically whatwe call fur capital.
It doesn't need to be only frombanks.
We also have some good partnersthat are non-banks.
The idea behind it is to tryand avoid and I hope I don't
step on any toes in here but toavoid the 1,000% APR on some
merchant cash advance solutions.
Rachel Morrissey (03:35):
Well, this is really interesting because we
were going to be talking aboutpersonalization, and I think
most of the time when we thinkabout personalization, or when
people talk about it, it's abouta singular consumer, like the
lone customer, right?
I remember years ago peoplewere talking to me and were like
can't you wait for the daywhere you have to?
(03:56):
You know you're flying into thenew city and at the hotel
they've already figured out yourlikes and dislikes and the
tickets to the ballgame arewaiting on the counter for you.
And there was a part of me thatwas like yeah, that sounds
great.
And there was a part of me thatsounds like no super creepy.
I have no idea why a generalhotel would know what tickets to
a ball game I want to get.
Omri Yacubovich (04:15):
And guess what Today?
They do know.
Rachel Morrissey (04:17):
They do know, I know.
But then when you're talkingabout it from the perspective of
what La Maillard is doing, sure, there's that part of it, but
how does this apply to smallbusinesses?
Omri Yacubovich (04:30):
That's a great question.
So first, I think that in thefinancial industry
personalization is even morescary than in the hospitality
industry, because the thing theycan know about you, if it's
kind of misused, could bedisastrous.
And I think there is a famousstory, I think it's about
Walmart.
Sheryl Chen (04:46):
Yes, Walmart or Target.
The 16-year-old yes, do youknow this story.
She knew that she was pregnanteven before her father knew that
she was pregnant.
Yeah.
Omri Yacubovich (04:55):
For her parents .
I think she was 16 years old orsomething.
Rachel Morrissey (04:57):
She was.
She was 16.
It was Target and it was a longtime ago, even from now, a long
time ago, 10 years Because Iwas in grad school at the time
and she had gotten informationto help her buy things that
would aid her pregnancy.
For when the baby came and herparents were like what in the
Lord's name is going on?
(05:18):
And then she was like, oh, Ithink that might be for me.
What a way to tell it is, it is, it is.
Sheryl Chen (05:26):
It's so crazy because I was just telling
Rachel yesterday that the daybefore I was telling a colleague
about LGBTQ issues and then thevery next day I was telling my
counterpart, ian Horn, aboutgoing to Phuket and just going
to a villa doing nothing andjust chilling after the Asia
show, and the very evening Ireceived an ad from Agoda
(05:48):
telling me to go to a gay villain Phuket.
Did Agoda just out me?
Omri Yacubovich (05:56):
Or your boyfriend?
I'm not sure.
Somebody somewhere is askingthose questions.
Sheryl Chen (06:05):
Yeah, so to personalization in financial
industry.
Omri Yacubovich (06:08):
Sure.
So personalization is anothertool to solve problems, right,
in my previous background, Ico-founded a company in the
e-com personalization space, sowe didn't do anything like
Target intentionally, but we didhelp figure out, kind of, if
you're entering an e-commercesite and like pink shirts or
jackets, why should you get ablack jacket as an offer?
(06:31):
So product recommendation wasactually there to solve a real
problem, which is kind of theconfusion among all the options
that you see in the store and Ican give you tens of other
examples not related to FinTech,hence I'll stop but the idea
that personalization is anothertool to solve for a real problem
again in the e-com world, it'slow conversion rates and the
inability of people to find whatthey're looking for.
(06:53):
When we're thinking about AI inthe fintech world or in the
lending world, there are a fewthings that it can do.
One of them is personalizationin a way that it helps the
borrower get to the finish linewith kind of minimum
intervention or minimumiterations with the RM.
Now, with that being said, Ithink that community banks pride
(07:16):
their relationship with theirclients, so they're not looking
to be replaced by an AI chatbotor AI agent, because then they
lose the relationship.
And that's not even the bigbottleneck in small business
underwriting.
Yet the ability to personalizethe experience helps the RMs do
(07:36):
a better job.
So they don't need to draft nowtens of emails a day saying hey
, mr X, you forgot to fill inyour tax returns, can you please
do that?
The system can generate itautomatically on their behalf.
But then all they do is a clickof a button and that email is
sent out.
So they keep full control asthey like.
But they don't need to do allthe writing in order to make a
(07:58):
personalized email based on thecontext of the client, based on
the step that they're in andalso any nuances that are also
related to regulation, as if youneed to unfortunately decline
opportunity, you have to send anadverse action notice and guess
what?
That that one should also bepersonalized, meaning providing
the reasons why do you declineit.
So again, it could all be donemanually, and that's how it has
(08:19):
been done for years, but theability to personalize the
message and use Gen AI togenerate those responses are a
crucial part on thepersonalization side.
Another example ofpersonalization and that's kind
of the more advanced banks, ifyou think about the loan journey
, you typically start bychoosing the product that you
need meaning SBA, loan, autoloan, whatever that is and then
(08:42):
it will take you forward with aa closed set of questions and
documents that you need toaddress In our world.
You should first say who youare.
If it's a business, what's yourbusiness name, what's your
address, how much money you needand what do you need it for and
then the system canautomatically recommend what's
the perfect product for you.
(09:03):
So instead of getting declinedon an SBA because of eligibility
criteria that I don't know,they do minimum 100,000, you
need only 90,000.
That's a stupid reason to bedeclined, but guess what People
are being declined for thattoday?
The system can automaticallysay, hey, this is the right
product for $90,000 for workingcapital.
So that's just another flavorof the ability to personalize
(09:24):
based on data, both declared oneand then data enrichment for
third-party sources,governmental databases, open web
, et cetera.
Sheryl Chen (09:32):
So the thing with AI, as with everything else, is
that we need to draw a very fineline right, a fine balance.
So how can financialinstitutions draw a line between
creepy and being helpful, likemaybe oh, I heard that you're
planning to propose to yourgirlfriend soon.
Like through the iPhone, I'vebeen listening to all your
(09:52):
conversations.
You're planning to propose toher.
So here's a home loan, becauseI think you guys are going to be
moving in soon.
Omri Yacubovich (09:59):
So I think there is data usage and data
abuse.
You don't want to abuse thedata.
Rachel Morrissey (10:09):
Yeah.
Omri Yacubovich (10:10):
That means that you shouldn't collect
everything that you can collect.
So if the bank collects onlythe data that they need in order
to assess the risk for acertain business or a certain
borrower, that won't be misusedin the fine line you're talking
about, but it will be used tomake sure that it's a good
(10:32):
borrower and, furthermore, Ithink the more data that they
have over time, the wider rangeof borrower they can serve and
basically serve more underservedcommunities that, if you just
tie your models to FICO score,as a lot of them do on the
consumer side, you're limitingyour ability to lend to
potentially good borrowers.
Rachel Morrissey (10:53):
Right, especially with some of these
less served communities.
Omri Yacubovich (11:01):
Underserved.
Sheryl Chen (11:02):
Underserved communities.
Rachel Morrissey (11:04):
When I'm thinking a little bit about this
idea of small businesspersonalization, what do you
think that small businesseswould look to to the banks?
What are they expecting whenthey say, when you say a
personal experience for smallbusiness, because the personal
experience for you or me is notnecessarily exactly the same
(11:24):
thing as a personal experiencefor a small business that you or
I would run right.
So what would a small businessthink of as a personalized I
mean loan?
you know, being told nicely andlegally that you didn't qualify
or you did qualify, or that youmight qualify for something that
was smaller or whatever, is onething, but what about some of
(11:44):
the others Like are they lookingfor stabilization?
Others Like are they lookingfor stabilization, are they
looking for more informationabout cash flows and treasury
management?
What is the kind ofpersonalization that you're
really talking about for that?
Omri Yacubovich (11:59):
So, first, behind any small business, you
have you and I standing behindit.
That's where they're runningthe show and the ones that are
applying.
So the personalization is kindof tuned to how we feel about it
and our one that are applying.
So the personalization is kindof tuned to how we feel about it
and our experience in thejourney.
That being said, the ability topersonalize the offering is
tied to the data that bank has.
(12:19):
So, instead of waiting forsomeone to come and say, hey, I
needed cash yesterday which is ascenario that banks actually
would not like to give you thatcash necessarily being proactive
about it and saying that thisbusiness maybe the hospitality
business that you discussed theyhave some cycles.
So, before the next cycle,saying, hey, it seems like a
great business, you're probablygoing to hit the low season soon
(12:41):
.
You need some cash to bridgecash flow issues.
So that type of personalizationin a way allows the banks to be
more proactive and kind of morehelpful for their clients.
Furthermore, you see some toolsthat embedding tax returns into
the ecosystem, right.
So basically, it provides moretools to the business owners to
(13:01):
run their business better.
Back to the more simplistic wayof thinking about it business
owners are busy running theirbusiness during daytime and
whenever they're free to dealwith their financing, their
bankers are probably alreadyasleep.
Typically, they live ondifferent time zones, now how do
(13:23):
you bridge that?
Because I think every bankwants to be useful and be
accessible to their clients, andhaving the ways to do it in
parallel, asynchronously and,even without an RM, provide some
information or help the ownerget through the finish line,
(13:44):
that's another huge value thatthey can get.
Rachel Morrissey (13:48):
So it's a way for them to actually personalize
it to the time of the actualowner's need, as opposed to
worrying about traditionalbanker's hours.
Omri Yacubovich (13:56):
Call you.
9 am your first meeting.
Rachel Morrissey (13:59):
The call center that's like, we're open
from 9 am to 5 pm on.
Midwest time.
Yeah, I can see that that wouldbe very interesting.
Omri Yacubovich (14:10):
And you're touching on another important
point, those automated systemsthat are sending emails and
texts.
They typically have kind of arule base.
Like you said, 9 am to 5 pm.
Having the ability tounderstand when someone is
opening their emails or when dothey engage with their platform,
when do they click the clicks,allows the system to become
smarter and know when to try andengage.
So, whenever they're back athome Now we're not tracking
(14:33):
their geography, but the systemcan track the time of the day
that they're interactive, so thesystem can better approach them
during these hours and say hey,you forgot to fill out your tax
returns.
This is how you do that forgotto fill out your tax returns.
Rachel Morrissey (14:49):
This is how you do that.
What do you think aboutsomething like an AI almost
being, or the banks using thisinformation to personalize?
But, for example, you said thatthere's businesses that are
cyclical.
They can be cyclical yearly,they can be cyclically quarterly
(15:10):
, but they know that they'retalking to their client when
times are in the fat, they'regood.
And what do you think about abank that says, hey, you're not
always going to be in this spot,this is a good time to put away
.
This is like what is the, whatwould be the kind of limits
(15:31):
around that, so that you are notkind of overextending yourself
there?
Omri Yacubovich (15:36):
I think that's conflict of interest.
Question right.
Rachel Morrissey (15:39):
It does feel because we were talking about
Amazon.
Right, that was part of theintroductory to the, to the in
the inside the app for this, andit pretty.
I mean, yes, what Amazon doeswith data is quite amazing, but
it's pretty easy, like she said,to have some input about.
You know, I'm interested inthis and I'm interested in this
(16:01):
and they put two and twotogether and they put that you
know shiny jewel in front of youand then you're like you know,
I'm going to click and impulsebuy, but that's the mission of a
.
Does that fit the mission of abank?
Not at all, right.
So how does the personalizationfit the mission of a bank?
Omri Yacubovich (16:18):
So I think the limit that some bankers see in
terms of personalization theythink about the relationship
managers that are interactingwith clients and that's what
they interpret as relationshipor personalized relationship.
My claim is that that's not apersonalization.
It could be a person deliveringthe message but using the data
(16:39):
like you talked aboutseasonality and other stuff to
have the rm call you at a timelymanner saying, hey, I, I see
you, I understand your business,this is what you need and this
is why you need it is like athousand times better than just
calling you out of nowheresaying, hey, we have a new loan
to offer you.
It's like oh, my gosh yeah.
Rachel Morrissey (17:01):
And why you would need it at any time.
It has nothing to do with theproduct design.
Omri Yacubovich (17:06):
The compelling event is something that differs
from one client to another,especially in business.
It has a lot of data pointsthat could indicate that that
could be someone's looking toexpand their business or just
bridge any cash flow issues, ormaybe the entire industry is on
its way down.
Think about COVID andrestaurants.
That's a very simple example.
(17:29):
But what if banks could reachahead of time and say hey guys,
you need to cut off your costs.
That's a trajectory, notnecessarily that a single
business owner can see thefuture that far, but I think
that's a great tool for bankersto really be trusted advisors
and not just salespeople.
Sheryl Chen (17:48):
Yeah, so this is also circling back to data usage
and what do we call it?
Misusage.
Omri Yacubovich (17:55):
No, no Data usage and this is abuse Abuse.
We coined a new term CorrectAbuse Abuse yes.
Sheryl Chen (18:02):
So I also wanted to , when we talk about data usage
abuse and also, like we alsowant to like take a peek on what
people have been they have beensweeping under the rug, right
so, when it comes to workingwith financial institutions,
banks what are the actualreasons for data silos and lack
(18:24):
of personalization that'shappening right now?
Omri Yacubovich (18:27):
I do think that they started personalizing
things.
So the data silos exist becauseit needs to exist, because of
regulatory, of course.
Even if you think about usingGen AI in the context of lending
, it has to be siloed.
You don't want one client'sdata to leak to another client,
(18:48):
so we're looking at it always inthe scope of specific
application or applicant datasilo that allows to create those
insights, not just forcommunication purposes or for
promotions, but also to actuallyassess the risk and help the
underwriters.
And that's where the big valueis helping the underwriting team
to be more efficient, smartabout it and really understand
(19:10):
the business.
Because when it comes tobusiness lending and I think not
a lot of people are aware ofthat that's not as easy as
underwriting a consumer loan,because for a consumer, you take
their FICA score and a few moredata points and that's it.
And it's not as lucrative asunderwriting $100 million deals
that you do whatever you do andthen you give them a check at
whatever rate and hope for thebetter doors of more established
(19:33):
businesses.
The small business sector isbuilt out of thousands of
different business types, sothink about makes codes.
You've got I don't know howmany, but thousands of them
right, and that reflectsdifferent business
characteristics.
Now, if you're an underwriterat a bank, nobody can expect you
to be an expert at 20,000different industries, and I
(19:55):
think that's another kind ofimplication for, yeah, a way to
use AI that allows you tobasically teach you and make you
more knowledgeable aboutindustry's pitfall or
benchmarking or anycharacteristics that otherwise
have no way to assess.
Rachel Morrissey (20:13):
Yeah, it'd be a great way to get extra insight
into any industry where thebanker is supposed to.
You know, the ideal is, ofcourse, the banker is an advisor
and can really help anybody andunderstand the industry.
But nobody can be an expertabout every kind of small
business or every kind ofindustry.
So this idea of this kind ofdata insight helping feed and
(20:37):
teach the underwriters and makethem become experts in more
kinds of businesses is kind offascinating.
If you think about it.
It really changes that becauseit would eliminate or it could
not necessarily, but couldeliminate a lot of biases.
Based on my experience with oneindustry and what I would think
of that, it might have verylittle to do with the realities
of another industry and createnew biases, though.
Omri Yacubovich (21:00):
Create new biases.
Rachel Morrissey (21:00):
So there we go .
That leads to the next question.
So talk to us about that.
How does it create new biases?
Omri Yacubovich (21:06):
So, before creating new biases, I think
it's not about educating theunderwriters but putting the
spotlight on anything theyshould be aware of before they
approve or decline kind ofopportunities, because it's
endless Extending theirknowledge.
Rachel Morrissey (21:21):
I'm not trying to indicate that they're not
doing their job.
Omri Yacubovich (21:24):
When it comes to biases, I guess, yes, people
are biased, so if they had badexperience with a loan that went
south in a certain industry,they're probably going to look
at it in a certain angle that isless favorable.
So when you automate some ofthese decisions or
recommendations, you definitelycan get rid of these biases.
(21:44):
And yet I think from theregulator perspective, you still
want underwriters to sign theapproval and say you can do it,
or at least the parameters thatinfluence the decision.
Even if it's fully automated,the new biases could be the data
sets that the AI was trained on.
(22:05):
So if it's trained on a certaindata set, that could create
some biases, and obviously thereare ways to avoid that.
Rachel Morrissey (22:14):
There's ways you should avoid that, but it's
an interesting problem solutionthat leads to another problem
that needs a solution.
Omri Yacubovich (22:23):
That's why I think that AI will not take our
jobs.
Rachel Morrissey (22:26):
That's why.
Omri Yacubovich (22:31):
For now?
For now, I think people don'tlike the equation of the
industrial revolution to the AIrevolution, but I think there
are some similarities andoverall we need to be optimistic
and I believe we might havemore free time.
That's a great thing.
Rachel Morrissey (22:48):
I remember when I was studying economics,
they've talked about Keynesiantheories, is it?
Invisible Hand no that's AdamSmith, but in Keynesian theories
, part of his thing was thatcapitalism was going to be so
efficient that we would all havea lot more time on our hands.
(23:09):
We'd only work like a 15 hourweek and then all of a sudden,
because the production would betoo high, and then we would only
have to work a 15 hour week andwe'd have a lot more leisure
time to educate ourselves orread or do things that we love,
and we wouldn't need to work somany hours.
I'm not giving him a star forthat.
(23:30):
I think he underestimatedcertain things when he came up
with that, but I still thinkthat there's a little bit of an
ideal there that we could reachfor, which is, if we are going
to have tools like this, maybewe should be thinking about how
we really want to apply it toproductivity and what that will
actually mean for human beings.
Omri Yacubovich (23:49):
But I think that part of capitalism people
are trying to get the most outof everything.
Rachel Morrissey (23:54):
That's why I thought it was a kind of an
overestimation.
Omri Yacubovich (23:57):
But even in today's world, if it can be more
efficient, and I can tell youthat this startup, lama, is way
more efficient than my previouscompany.
Not because people are workinghard.
They are working hard, but theyhave way more tools to achieve.
Sheryl Chen (24:10):
They're working a lot smarter.
Omri Yacubovich (24:13):
Exactly hard, but they have way more tools to
achieve.
They're working a lot smarterExactly, but they're working as
hard but smarter, so they canproduce way more.
So, in a capitalist world,we'll just try to get more.
So I don't think we'll freemore time.
Rachel Morrissey (24:25):
I don't think we're going to end up with a
15-hour work.
Sheryl Chen (24:26):
Yeah, you're going to spend all the other time
coming up with like 15 otherside hustles.
Rachel Morrissey (24:32):
I spent all the other time coming up with
like 15 other side hustles.
That's probably true.
Okay, well, we are at time.
That was really fast, is thereanything?
You want to tell us before wego.
Omri Yacubovich (24:48):
Thank you again for having me.
Rachel Morrissey (24:49):
Thank you so much for joining us.
Omri Yacubovich (24:51):
It's a great show.
Rachel Morrissey (24:51):
Oh, thank you.
Yeah, I hope you so much it's agreat show.
Oh, thank you.
Yeah, I hope you all enjoyingyour time, okay, so, uh, thank
you everybody.
Thank you, Sheryl, for being myco-host for having me.
Sheryl Chen (25:03):
I wish we had more time.
I wish we did too and thank you.
Rachel Morrissey (25:07):
Uh, we want to thank our live audience.
We want to thank our podcastaudience.
We want to thank our podcastaudience.
If you guys think you have agreat idea for a podcast episode
for the Money Pot, please goahead and email us podcast at
money2020.com, and we would loveto hear from you.
And to all of our listeners, goahead and tell everyone to
subscribe.
Have a great day.
Welcome to the Money Pot.
I'm Rachel Morrissey from Money2020, and we're recording live
at our show in Las Vegas, and sofar, the day has been pretty
good.
I am here with my co-host,Sheryl Chen, who is the head of
content for Money 2020 Asia.
Sheryl has spent most of thetime on the sentient stage, so
I'm just going to ask you,Sheryl, what do you think?
Sheryl Chen (00:28):
Do you think the future is actually human?
Is the future human or is it AI?
Is it AI?
Is it AI?
I think it will be a long whiletill AI can fully reign.
Yeah, but for now, I think,humans, we will still be able to
keep our jobs, thankfully,thankfully.
Rachel Morrissey (00:45):
Very thankfully.
I was thinking AI would replaceme any day now.
Well it can, and we are herewith our guest who would know,
Omri Yacubovich, CEO of Llama AI, and so how are you feeling
about everything, Omri?
Omri Yacubovich (01:01):
Good Glad to be here.
Rachel Morrissey (01:02):
Good, so you tell me what have you seen
around the show that has struckyou, especially about AI, since
that's your yeah, so I think itrelates to your kind of previous
comments.
Omri Yacubovich (01:14):
I think early on people were thinking that AI
will replace the blue coloremployees.
Now they all get it.
It's going to replace the whitecolor.
Sheryl Chen (01:21):
I think going to replace the white collar.
Rachel Morrissey (01:23):
I think, rachel and I anybody with a data
like brain dead you know what?
Sheryl Chen (01:27):
I, my co-host is actually an ai co-host, so I
actually joked with her that sheactually stole my job.
But she was so polite.
She said that you know what?
There are enough like.
There's enough space for allthree of us under the moonlight
to share the stage.
Rachel Morrissey (01:41):
She's the true romantic that, Aianna.
Okay, we're gonna get startedbecause we are actually talking
about something besides the aiunder the moonlight as they
share the stage.
Very romantic, she's the trueromantic that, Aianna.
Okay, we're going to getstarted because we are actually
talking about something besidesAI, but it is related.
Sheryl, do you want to go ahead?
Sheryl Chen (01:53):
Yeah so maybe as a first question, as an
introductory question, can youtell us the mission of what Lama
AI does?
Omri Yacubovich (02:00):
Absolutely, so we started the company about
three years ago.
Actually, here does Absolutely.
We started the company aboutthree years ago, actually here
at Money2020.
Sheryl Chen (02:08):
Nice Money2020, baby.
Omri Yacubovich (02:09):
About two weeks before Money2020 2021, we kind
of hit the nail on the problemto our after, and that's what
today I call the small businesslending paradox, meaning
everybody talks about lending tosmall businesses, everybody
wants to lend to them, but onlya few banks can actually achieve
that.
So we bought flight tickets,came over, met with about 20
(02:33):
bankers in the course of 48hours.
It's not like you didn'torganize these networking events
like you're doing in this show,so it was much harder Cold
calling, emails, whatever youwant but we did secure these 20
meetings.
And what we learned?
That not only that the problemis real, but this is something
that bankers are looking tosolve, because small businesses
(02:54):
are kind of the backbone of oureconomy.
They represent over 55%, Ithink, of businesses employing
probably around 90% of theemployees across the states.
Wow, I think of businessesemploying probably around 90% of
the employees across the states.
Our mission is to help andbridge the gap in small business
lending, helping small businessowners get access to bank-rated
(03:15):
financials and basically whatwe call fur capital.
It doesn't need to be only frombanks.
We also have some good partnersthat are non-banks.
The idea behind it is to tryand avoid and I hope I don't
step on any toes in here but toavoid the 1,000% APR on some
merchant cash advance solutions.
Rachel Morrissey (03:35):
Well, this is really interesting because we
were going to be talking aboutpersonalization, and I think
most of the time when we thinkabout personalization, or when
people talk about it, it's abouta singular consumer, like the
lone customer, right?
I remember years ago peoplewere talking to me and were like
can't you wait for the daywhere you have to?
(03:56):
You know you're flying into thenew city and at the hotel
they've already figured out yourlikes and dislikes and the
tickets to the ballgame arewaiting on the counter for you.
And there was a part of me thatwas like yeah, that sounds
great.
And there was a part of me thatsounds like no super creepy.
I have no idea why a generalhotel would know what tickets to
a ball game I want to get.
Omri Yacubovich (04:15):
And guess what Today?
They do know.
Rachel Morrissey (04:17):
They do know, I know.
But then when you're talkingabout it from the perspective of
what La Maillard is doing, sure, there's that part of it, but
how does this apply to smallbusinesses?
Omri Yacubovich (04:30):
That's a great question.
So first, I think that in thefinancial industry
personalization is even morescary than in the hospitality
industry, because the thing theycan know about you, if it's
kind of misused, could bedisastrous.
And I think there is a famousstory, I think it's about
Walmart.
Sheryl Chen (04:46):
Yes, Walmart or Target.
The 16-year-old yes, do youknow this story.
She knew that she was pregnanteven before her father knew that
she was pregnant.
Yeah.
Omri Yacubovich (04:55):
For her parents .
I think she was 16 years old orsomething.
Rachel Morrissey (04:57):
She was.
She was 16.
It was Target and it was a longtime ago, even from now, a long
time ago, 10 years Because Iwas in grad school at the time
and she had gotten informationto help her buy things that
would aid her pregnancy.
For when the baby came and herparents were like what in the
Lord's name is going on?
(05:18):
And then she was like, oh, Ithink that might be for me.
What a way to tell it is, it is, it is.
Sheryl Chen (05:26):
It's so crazy because I was just telling
Rachel yesterday that the daybefore I was telling a colleague
about LGBTQ issues and then thevery next day I was telling my
counterpart, ian Horn, aboutgoing to Phuket and just going
to a villa doing nothing andjust chilling after the Asia
show, and the very evening Ireceived an ad from Agoda
(05:48):
telling me to go to a gay villain Phuket.
Did Agoda just out me?
Omri Yacubovich (05:56):
Or your boyfriend?
I'm not sure.
Somebody somewhere is askingthose questions.
Sheryl Chen (06:05):
Yeah, so to personalization in financial
industry.
Omri Yacubovich (06:08):
Sure.
So personalization is anothertool to solve problems, right,
in my previous background, Ico-founded a company in the
e-com personalization space, sowe didn't do anything like
Target intentionally, but we didhelp figure out, kind of, if
you're entering an e-commercesite and like pink shirts or
jackets, why should you get ablack jacket as an offer?
(06:31):
So product recommendation wasactually there to solve a real
problem, which is kind of theconfusion among all the options
that you see in the store and Ican give you tens of other
examples not related to FinTech,hence I'll stop but the idea
that personalization is anothertool to solve for a real problem
again in the e-com world, it'slow conversion rates and the
inability of people to find whatthey're looking for.
(06:53):
When we're thinking about AI inthe fintech world or in the
lending world, there are a fewthings that it can do.
One of them is personalizationin a way that it helps the
borrower get to the finish linewith kind of minimum
intervention or minimumiterations with the RM.
Now, with that being said, Ithink that community banks pride
(07:16):
their relationship with theirclients, so they're not looking
to be replaced by an AI chatbotor AI agent, because then they
lose the relationship.
And that's not even the bigbottleneck in small business
underwriting.
Yet the ability to personalizethe experience helps the RMs do
(07:36):
a better job.
So they don't need to draft nowtens of emails a day saying hey
, mr X, you forgot to fill inyour tax returns, can you please
do that?
The system can generate itautomatically on their behalf.
But then all they do is a clickof a button and that email is
sent out.
So they keep full control asthey like.
But they don't need to do allthe writing in order to make a
(07:58):
personalized email based on thecontext of the client, based on
the step that they're in andalso any nuances that are also
related to regulation, as if youneed to unfortunately decline
opportunity, you have to send anadverse action notice and guess
what?
That that one should also bepersonalized, meaning providing
the reasons why do you declineit.
So again, it could all be donemanually, and that's how it has
(08:19):
been done for years, but theability to personalize the
message and use Gen AI togenerate those responses are a
crucial part on thepersonalization side.
Another example ofpersonalization and that's kind
of the more advanced banks, ifyou think about the loan journey
, you typically start bychoosing the product that you
need meaning SBA, loan, autoloan, whatever that is and then
(08:42):
it will take you forward with aa closed set of questions and
documents that you need toaddress In our world.
You should first say who youare.
If it's a business, what's yourbusiness name, what's your
address, how much money you needand what do you need it for and
then the system canautomatically recommend what's
the perfect product for you.
(09:03):
So instead of getting declinedon an SBA because of eligibility
criteria that I don't know,they do minimum 100,000, you
need only 90,000.
That's a stupid reason to bedeclined, but guess what People
are being declined for thattoday?
The system can automaticallysay, hey, this is the right
product for $90,000 for workingcapital.
So that's just another flavorof the ability to personalize
(09:24):
based on data, both declared oneand then data enrichment for
third-party sources,governmental databases, open web
, et cetera.
Sheryl Chen (09:32):
So the thing with AI, as with everything else, is
that we need to draw a very fineline right, a fine balance.
So how can financialinstitutions draw a line between
creepy and being helpful, likemaybe oh, I heard that you're
planning to propose to yourgirlfriend soon.
Like through the iPhone, I'vebeen listening to all your
(09:52):
conversations.
You're planning to propose toher.
So here's a home loan, becauseI think you guys are going to be
moving in soon.
Omri Yacubovich (09:59):
So I think there is data usage and data
abuse.
You don't want to abuse thedata.
Rachel Morrissey (10:09):
Yeah.
Omri Yacubovich (10:10):
That means that you shouldn't collect
everything that you can collect.
So if the bank collects onlythe data that they need in order
to assess the risk for acertain business or a certain
borrower, that won't be misusedin the fine line you're talking
about, but it will be used tomake sure that it's a good
(10:32):
borrower and, furthermore, Ithink the more data that they
have over time, the wider rangeof borrower they can serve and
basically serve more underservedcommunities that, if you just
tie your models to FICO score,as a lot of them do on the
consumer side, you're limitingyour ability to lend to
potentially good borrowers.
Rachel Morrissey (10:53):
Right, especially with some of these
less served communities.
Omri Yacubovich (11:01):
Underserved.
Sheryl Chen (11:02):
Underserved communities.
Rachel Morrissey (11:04):
When I'm thinking a little bit about this
idea of small businesspersonalization, what do you
think that small businesseswould look to to the banks?
What are they expecting whenthey say, when you say a
personal experience for smallbusiness, because the personal
experience for you or me is notnecessarily exactly the same
(11:24):
thing as a personal experiencefor a small business that you or
I would run right.
So what would a small businessthink of as a personalized I
mean loan?
you know, being told nicely andlegally that you didn't qualify
or you did qualify, or that youmight qualify for something that
was smaller or whatever, is onething, but what about some of
(11:44):
the others Like are they lookingfor stabilization?
Others Like are they lookingfor stabilization, are they
looking for more informationabout cash flows and treasury
management?
What is the kind ofpersonalization that you're
really talking about for that?
Omri Yacubovich (11:59):
So, first, behind any small business, you
have you and I standing behindit.
That's where they're runningthe show and the ones that are
applying.
So the personalization is kindof tuned to how we feel about it
and our one that are applying.
So the personalization is kindof tuned to how we feel about it
and our experience in thejourney.
That being said, the ability topersonalize the offering is
tied to the data that bank has.
(12:19):
So, instead of waiting forsomeone to come and say, hey, I
needed cash yesterday which is ascenario that banks actually
would not like to give you thatcash necessarily being proactive
about it and saying that thisbusiness maybe the hospitality
business that you discussed theyhave some cycles.
So, before the next cycle,saying, hey, it seems like a
great business, you're probablygoing to hit the low season soon
(12:41):
.
You need some cash to bridgecash flow issues.
So that type of personalizationin a way allows the banks to be
more proactive and kind of morehelpful for their clients.
Furthermore, you see some toolsthat embedding tax returns into
the ecosystem, right.
So basically, it provides moretools to the business owners to
(13:01):
run their business better.
Back to the more simplistic wayof thinking about it business
owners are busy running theirbusiness during daytime and
whenever they're free to dealwith their financing, their
bankers are probably alreadyasleep.
Typically, they live ondifferent time zones, now how do
(13:23):
you bridge that?
Because I think every bankwants to be useful and be
accessible to their clients, andhaving the ways to do it in
parallel, asynchronously and,even without an RM, provide some
information or help the ownerget through the finish line,
(13:44):
that's another huge value thatthey can get.
Rachel Morrissey (13:48):
So it's a way for them to actually personalize
it to the time of the actualowner's need, as opposed to
worrying about traditionalbanker's hours.
Omri Yacubovich (13:56):
Call you.
9 am your first meeting.
Rachel Morrissey (13:59):
The call center that's like, we're open
from 9 am to 5 pm on.
Midwest time.
Yeah, I can see that that wouldbe very interesting.
Omri Yacubovich (14:10):
And you're touching on another important
point, those automated systemsthat are sending emails and
texts.
They typically have kind of arule base.
Like you said, 9 am to 5 pm.
Having the ability tounderstand when someone is
opening their emails or when dothey engage with their platform,
when do they click the clicks,allows the system to become
smarter and know when to try andengage.
So, whenever they're back athome Now we're not tracking
(14:33):
their geography, but the systemcan track the time of the day
that they're interactive, so thesystem can better approach them
during these hours and say hey,you forgot to fill out your tax
returns.
This is how you do that forgotto fill out your tax returns.
Rachel Morrissey (14:49):
This is how you do that.
What do you think aboutsomething like an AI almost
being, or the banks using thisinformation to personalize?
But, for example, you said thatthere's businesses that are
cyclical.
They can be cyclical yearly,they can be cyclically quarterly
(15:10):
, but they know that they'retalking to their client when
times are in the fat, they'regood.
And what do you think about abank that says, hey, you're not
always going to be in this spot,this is a good time to put away
.
This is like what is the, whatwould be the kind of limits
(15:31):
around that, so that you are notkind of overextending yourself
there?
Omri Yacubovich (15:36):
I think that's conflict of interest.
Question right.
Rachel Morrissey (15:39):
It does feel because we were talking about
Amazon.
Right, that was part of theintroductory to the, to the in
the inside the app for this, andit pretty.
I mean, yes, what Amazon doeswith data is quite amazing, but
it's pretty easy, like she said,to have some input about.
You know, I'm interested inthis and I'm interested in this
(16:01):
and they put two and twotogether and they put that you
know shiny jewel in front of youand then you're like you know,
I'm going to click and impulsebuy, but that's the mission of a
.
Does that fit the mission of abank?
Not at all, right.
So how does the personalizationfit the mission of a bank?
Omri Yacubovich (16:18):
So I think the limit that some bankers see in
terms of personalization theythink about the relationship
managers that are interactingwith clients and that's what
they interpret as relationshipor personalized relationship.
My claim is that that's not apersonalization.
It could be a person deliveringthe message but using the data
(16:39):
like you talked aboutseasonality and other stuff to
have the rm call you at a timelymanner saying, hey, I, I see
you, I understand your business,this is what you need and this
is why you need it is like athousand times better than just
calling you out of nowheresaying, hey, we have a new loan
to offer you.
It's like oh, my gosh yeah.
Rachel Morrissey (17:01):
And why you would need it at any time.
It has nothing to do with theproduct design.
Omri Yacubovich (17:06):
The compelling event is something that differs
from one client to another,especially in business.
It has a lot of data pointsthat could indicate that that
could be someone's looking toexpand their business or just
bridge any cash flow issues, ormaybe the entire industry is on
its way down.
Think about COVID andrestaurants.
That's a very simple example.
(17:29):
But what if banks could reachahead of time and say hey guys,
you need to cut off your costs.
That's a trajectory, notnecessarily that a single
business owner can see thefuture that far, but I think
that's a great tool for bankersto really be trusted advisors
and not just salespeople.
Sheryl Chen (17:48):
Yeah, so this is also circling back to data usage
and what do we call it?
Misusage.
Omri Yacubovich (17:55):
No, no Data usage and this is abuse Abuse.
We coined a new term CorrectAbuse Abuse yes.
Sheryl Chen (18:02):
So I also wanted to , when we talk about data usage
abuse and also, like we alsowant to like take a peek on what
people have been they have beensweeping under the rug, right
so, when it comes to workingwith financial institutions,
banks what are the actualreasons for data silos and lack
(18:24):
of personalization that'shappening right now?
Omri Yacubovich (18:27):
I do think that they started personalizing
things.
So the data silos exist becauseit needs to exist, because of
regulatory, of course.
Even if you think about usingGen AI in the context of lending
, it has to be siloed.
You don't want one client'sdata to leak to another client,
(18:48):
so we're looking at it always inthe scope of specific
application or applicant datasilo that allows to create those
insights, not just forcommunication purposes or for
promotions, but also to actuallyassess the risk and help the
underwriters.
And that's where the big valueis helping the underwriting team
to be more efficient, smartabout it and really understand
(19:10):
the business.
Because when it comes tobusiness lending and I think not
a lot of people are aware ofthat that's not as easy as
underwriting a consumer loan,because for a consumer, you take
their FICA score and a few moredata points and that's it.
And it's not as lucrative asunderwriting $100 million deals
that you do whatever you do andthen you give them a check at
whatever rate and hope for thebetter doors of more established
(19:33):
businesses.
The small business sector isbuilt out of thousands of
different business types, sothink about makes codes.
You've got I don't know howmany, but thousands of them
right, and that reflectsdifferent business
characteristics.
Now, if you're an underwriterat a bank, nobody can expect you
to be an expert at 20,000different industries, and I
(19:55):
think that's another kind ofimplication for, yeah, a way to
use AI that allows you tobasically teach you and make you
more knowledgeable aboutindustry's pitfall or
benchmarking or anycharacteristics that otherwise
have no way to assess.
Rachel Morrissey (20:13):
Yeah, it'd be a great way to get extra insight
into any industry where thebanker is supposed to.
You know, the ideal is, ofcourse, the banker is an advisor
and can really help anybody andunderstand the industry.
But nobody can be an expertabout every kind of small
business or every kind ofindustry.
So this idea of this kind ofdata insight helping feed and
(20:37):
teach the underwriters and makethem become experts in more
kinds of businesses is kind offascinating.
If you think about it.
It really changes that becauseit would eliminate or it could
not necessarily, but couldeliminate a lot of biases.
Based on my experience with oneindustry and what I would think
of that, it might have verylittle to do with the realities
of another industry and createnew biases, though.
Omri Yacubovich (21:00):
Create new biases.
Rachel Morrissey (21:00):
So there we go .
That leads to the next question.
So talk to us about that.
How does it create new biases?
Omri Yacubovich (21:06):
So, before creating new biases, I think
it's not about educating theunderwriters but putting the
spotlight on anything theyshould be aware of before they
approve or decline kind ofopportunities, because it's
endless Extending theirknowledge.
Rachel Morrissey (21:21):
I'm not trying to indicate that they're not
doing their job.
Omri Yacubovich (21:24):
When it comes to biases, I guess, yes, people
are biased, so if they had badexperience with a loan that went
south in a certain industry,they're probably going to look
at it in a certain angle that isless favorable.
So when you automate some ofthese decisions or
recommendations, you definitelycan get rid of these biases.
(21:44):
And yet I think from theregulator perspective, you still
want underwriters to sign theapproval and say you can do it,
or at least the parameters thatinfluence the decision.
Even if it's fully automated,the new biases could be the data
sets that the AI was trained on.
(22:05):
So if it's trained on a certaindata set, that could create
some biases, and obviously thereare ways to avoid that.
Rachel Morrissey (22:14):
There's ways you should avoid that, but it's
an interesting problem solutionthat leads to another problem
that needs a solution.
Omri Yacubovich (22:23):
That's why I think that AI will not take our
jobs.
Rachel Morrissey (22:26):
That's why.
Omri Yacubovich (22:31):
For now?
For now, I think people don'tlike the equation of the
industrial revolution to the AIrevolution, but I think there
are some similarities andoverall we need to be optimistic
and I believe we might havemore free time.
That's a great thing.
Rachel Morrissey (22:48):
I remember when I was studying economics,
they've talked about Keynesiantheories, is it?
Invisible Hand no that's AdamSmith, but in Keynesian theories
, part of his thing was thatcapitalism was going to be so
efficient that we would all havea lot more time on our hands.
(23:09):
We'd only work like a 15 hourweek and then all of a sudden,
because the production would betoo high, and then we would only
have to work a 15 hour week andwe'd have a lot more leisure
time to educate ourselves orread or do things that we love,
and we wouldn't need to work somany hours.
I'm not giving him a star forthat.
(23:30):
I think he underestimatedcertain things when he came up
with that, but I still thinkthat there's a little bit of an
ideal there that we could reachfor, which is, if we are going
to have tools like this, maybewe should be thinking about how
we really want to apply it toproductivity and what that will
actually mean for human beings.
Omri Yacubovich (23:49):
But I think that part of capitalism people
are trying to get the most outof everything.
Rachel Morrissey (23:54):
That's why I thought it was a kind of an
overestimation.
Omri Yacubovich (23:57):
But even in today's world, if it can be more
efficient, and I can tell youthat this startup, lama, is way
more efficient than my previouscompany.
Not because people are workinghard.
They are working hard, but theyhave way more tools to achieve.
Sheryl Chen (24:10):
They're working a lot smarter.
Omri Yacubovich (24:13):
Exactly hard, but they have way more tools to
achieve.
They're working a lot smarterExactly, but they're working as
hard but smarter, so they canproduce way more.
So, in a capitalist world,we'll just try to get more.
So I don't think we'll freemore time.
Rachel Morrissey (24:25):
I don't think we're going to end up with a
15-hour work.
Sheryl Chen (24:26):
Yeah, you're going to spend all the other time
coming up with like 15 otherside hustles.
Rachel Morrissey (24:32):
I spent all the other time coming up with
like 15 other side hustles.
That's probably true.
Okay, well, we are at time.
That was really fast, is thereanything?
You want to tell us before wego.
Omri Yacubovich (24:48):
Thank you again for having me.
Rachel Morrissey (24:49):
Thank you so much for joining us.
Omri Yacubovich (24:51):
It's a great show.
Rachel Morrissey (24:51):
Oh, thank you.
Yeah, I hope you so much it's agreat show.
Oh, thank you.
Yeah, I hope you all enjoyingyour time, okay, so, uh, thank
you everybody.
Thank you, Sheryl, for being myco-host for having me.
Sheryl Chen (25:03):
I wish we had more time.
I wish we did too and thank you.
Rachel Morrissey (25:07):
Uh, we want to thank our live audience.
We want to thank our podcastaudience.
We want to thank our podcastaudience.
If you guys think you have agreat idea for a podcast episode
for the Money Pot, please goahead and email us podcast at
money2020.com, and we would loveto hear from you.
And to all of our listeners, goahead and tell everyone to
subscribe.
Have a great day.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com