AI in Banking: The Risk Beyond the Reward

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Amanda Knight (00:23):
Welcome back to Placing you First, the podcast
that keeps you at the forefrontof insurance insights and
industry shifts.
I'm Amanda Knight, joined by myco-host, Scott Gordon.

Scott Gordon (00:33):
Today we're diving into a conversation that's
generating a lot of buzz therole of AI in banking and
financial services.
While the efficiency and profitgains are exciting, the risks
could be game changing forcarriers, brokers and clients
alike.

Here to break it down for us today is Mark
Waldeck, president of CRCChicago and longtime specialist
in financial institution risk.
This is the Placing you Firstpodcast from CRC Group.
This podcast features news andinsights from a vast knowledge
base of more than 5,500associates who write more than
$30 billion in premium annually.

(01:10):
Plus, we give you the latestinformation on what's happening
at CRC.

This this this is the Placing you First podcast
and now the hosts of the podcastAmanda Knight and Scott Gordon.

Mark, welcome to the show.
Thanks for having me.
Well, you know, as I waspreparing for this episode, I
ran across some interesting infothat I thought I would share
with the class before we take adeep dive today.
Are you guys ready?
I learned some things, somethings I didn't know, and it
could have even been AI teachingme this through Google.
It's hard to say, but did youknow?

(01:43):
Eliza was actually the firstchatbot and she was created hold
on to your hat in 1966.

Was it for a James Bond movie?

59 years ago, yeah, okay.
So I was like this can't beaccurate, but it is.
She was created 59 years ago atMIT.
Was created 59 years ago at MIT.
She was this like simulatedpsychotherapist that operated by
rephrasing users' inputs asquestions, sort of like a
therapist would do to act as amirror to a patient, right?

(02:18):
So it sounds like she is likeabsorbing information and giving
it back to you, but reallyshe's just reflecting it, just
rephrasing things.
She wasn't actually learningnecessarily, but people got
really attached even thoughEliza had no real understanding.
She didn't involve, like, Iguess, actual machine learning.

Did she run off?

Mark Waldeck (02:40):
of punch cards.
I'm thinking the computer wasprobably the size of a Buick
Right.

Yeah, probably A Buick showroom.

Also true.
So here's our second fun factand this will surprise no one
probably ChatGPT, maybe the mostwell-known form of AI that is
kind of popular at the moment.
Chatgpt reached 1 million usersin five days less than a
calendar week.

(03:09):
So, in comparison, it tookInstagram another highly popular
app two and a half months toreach a million users, and it
took Netflix three and a halfyears to reach a million users.
So this makes ChatGPT one ofthe fastest growing consumer

(03:31):
apps period in history.
I mean, I was like that makessense.
I can see that.
And then Erica I don't know.
Like I think we could have awhole nother podcast on why we
name all of these chatbots withwomen's names.
We've got Eliza.
Now we're going to talk aboutErica.
Erica, bank of America's AIassistant, was developed in 2018

(03:55):
.
And we actually discuss her inthe article that goes along with
this podcast.
But Erica has had over 2billion interactions.
That's more than six times thepopulation of the entire United
States and she's managed all ofthem without human agents Just
Erica the chatbot.

(04:15):
And then we'll end our funlittle fact-finding mission here
with a rogue fact.
One chatbot once claimed itwanted to be alive.
Microsoft's AI chatbot Taylaunched in 2016.
It started out okay, but thenquickly spiraled out of control

(04:37):
after Tay started learning fromTwitter.
Which shocker right.
Tay had to be shut down within24 hours.

I think I remember why, I think it was getting
pretty anti-Semitic on thecomments.
Thanks, twitter.
Thanks a lot.

Twitter now known as X.
So thanks a lot.
Twitter now known as X.
So I mean, these facts alonetell us that AI chatbots can
change the game and definitelypresent some substantial risks.
So now that we've all got somesmall talk fodder for our next
cocktail hour, we can move on tothe meat of this conversation
with Mark.
So, scott, I'm going to let youkick it off.

What are you dying to know.
Well, now that we've coveredthe really important stuff, Mark
, can you give us a quickoverview of how banks are
currently leveraging AI and howfast it's moving?

Does every bank have a chatbot, or is it just
the big boys like Bank ofAmerica?

Well, it started with the big ones and now it's
starting to drop down into thesmaller, medium-sized banks, and
I think the original intent wasto offer that first layer of
customer assistance.
If it was a mundane issue orneed that, the chatbot could
handle that, and it's expandedto do more and more.

(06:00):
You're not just checking yourbalance or following through to
see if a specific check numberis cleared.
Now it's going probably half adozen to a dozen steps beyond,
and then at what point do youmove from providing factual data
versus interpretive results andcommentary?

(06:21):
And I think you get a differentanswer from every bank on what
they're doing and how they'reusing it.
I think the challenge, though,for us as users if any of us
that have accounts the concernis and I'll take it back to
since we have an insuranceaudience when I have a really

(06:41):
tough account, I want a humanunderwriter signing off on it, I
want it documented, and I wantto know that that guy's
documentation is in my file if Iever find myself in a dispute
with a claim.
Is it covered?
Is it not covered?
Hey, we talked about this.
I shared all this information,and I think the challenge for
some of the banks now is that,due to the lack of guidance,

(07:07):
they may have taken this in adirection that was probably
further than we're ready as botha society, as well as how we're
going to interact with ourlegal system.
If something does go wrong,because it's no longer liability
possessed by an individualacross from you, it's the
liability of the software'sconclusions, or maybe it's even

(07:30):
an error.
I don't know.

Well, and we've also talked about, or you just
kind of mentioned you likehaving a human to sign off on
things and sort of put the stampof approval on.
So what kind of conversationsare you having with retail
agents or their clients aboutAI-powered tools and their
insurance implications?

(07:52):
Are they, even at this point,savvy enough to come to you
first?
Are they pulling you in?
Oh, by the way, we launchedthis chatbot six months ago and
now we need to cover it.
Are they doing a better job ofbeing proactive?
It seems like that can bechallenging when it's something
that's moving as quickly as AIis moving.

Unfortunately no, it's up to me to pull that out,
so it's sort of a standard topicthat I bring up at every
meeting.
Hey, where are we with this?
Are you considering it?
And if so, how?
And then what fence have youput around those liabilities so
that it's not open-ended?
Because there aren't a lot ofchoices for these large

(08:36):
commercial clients, in this casea commercial bank.
So if there's only a couple ofgood tools that they can tap
into, it's no different thandoing business with the cloud or
any other situation.
You kind of have to take it andunfortunately a lot of that
liability ends up back in yourlap.
You can't pass it up the foodchain, unfortunately.

Yeah, and technology.
I remember remember how crazyit was the first time you
realized you could deposit acheck from your phone.
It blew my mind.
I know it's not exactly the AIsubject, but technology in
general and banks it's come sofast.
It just seems like just theother day I was having to sign
it and take it into the bank.

Just the other day I was having to sign it and
take it into the bank.
I listened to a podcast theother day where they were
talking about how they had toexplain to the younger listeners
about how in the drive-thru youhad the tube that you put the
check or the money in and itsucked it into the bank.
I thought that was amazing as akid, because then they sent you
a sucker right when they knewthere were kids in the car and
the younger podcast listenerswere like what are you talking

(09:38):
about?
Because they've never not hadAI or a chatbot or an app to do
all of these things.
Some of them have never evenbeen inside a bank.

And I think remote deposit's a good example in
that to my earlier comment.
If you have a $10 check, that'sfine if you want to take a
picture of it and deposit it.
But if you had a $10,000 check,wouldn't you want to go to the
teller line and get a receiptand know that that money was
received and posted to youraccount, as opposed to did it go

(10:09):
through?
Did it not go through?
I'm still old school.
I don't know how someone intheir 20s feels about that, but
when you get into big dollarsagain, I want a human in the
middle of that transaction.

Yeah, sure, I was one of the last people to use
like a paper check to pay.
I forget there was one billthat I always paid with a check
and my daughter made fun of me.
And I was like OK, I need toget with the times, Sorry.
So let's get back to thequestions.
One of the biggest concernswe're hearing now is how reliant
banks have become on a fewmajor tech providers.

(10:45):
So, Mark, what are the risksthere and who ultimately bears
the liability in the event of,say, a data breach?

I sort of touched on it earlier.
There's a short list ofproviders that let's use
processing of bank statements.
There's only about three orfour good players out there and
they know it and, as a result,if you want to hire them to
process your bank statements,most of the liability is coming
back to you.
Fiserv is not going to take theexposure just because they're

(11:17):
processing your statements for afee.
So if they make a mistake, forthe most part it's your problem,
it's your brand, it's yourgoodwill, it's your client base,
and if those people want to sueyou, good chance.
Fiserv's not going to standshoulder to shoulder with you.
You're on your own.
So it is a very small universeand, as a result, it puts a lot

(11:41):
of burden on us to at leastreview the contracts.
And so, um, to answer theearlier question about well, are
the banks contacting me or am Icontacting them?
That's how I engage them.
Are you doing this?
And if you are, I need to seethose contracts.
It doesn't mean I can fix them,but it does pivot the

(12:01):
conversation from gosh.
I didn't realize we had thatmuch exposure.
We were assuming, and now thatwe know this is the insurance
that we're buying adequate fornow the exposure we've been sort
of awakened to.
And I tell everybody let's gothrough and pretend you had a

(12:22):
claim today, because it will besystemic, it won't be a one-off,
it'll be.
All your users, all youraccount holders are going to be
hit, probably at the same time.
So how might we go through thisif we were doing a fictitious
doomsday scenario, becausethat's what really gets people
thinking.

Well, and we've kind of touched on this next
question when we're thinkingabout a chatbot that is powered
by a third party, one of thesemajor tech providers, it
collects and it stores customerdata, we have to think about how
is a bank and their insurerstaying ahead of compliance,

(12:59):
contract risks, like youmentioned, coverage gaps, and it
would seem to me that you kindof have to game it out, just
like you're doing with people,right, because the risk is
invisible to so many people.
You assume and we all know whatassuming does, right?
You assume that if the chatbotexists and it's powered by one

(13:21):
of the largest tech providers,et cetera, that it must be safe,
it must be compliant, and Idon't think that's always
necessarily the smartest way todo this.
So it sounds almost like you'resaying they're not doing a
great job of staying ahead ofsome of these things, unless you
happen to be the one thatthey're dealing with and
bringing it up.

Yeah, I think it's up to the broker and the
general counsel to drive thatconversation, because they see
this as a marketing tool.
If we fail to deliver thesecustomer convenience options,
those same customers will leaveus and go somewhere else, and it
could be a non-sanctioned bank.

(14:00):
We have a lot of new entitiesthat have been created in the
last 10 to 15 years that are notwhat I would call licensed
chartered banks, either at thefederal or the state level, and,
as a result, banks arecompeting with people with a
lower cost structure.
It used to be just creditunions and they're more loosely

(14:22):
regulated, but now it's movedwell beyond that and there's
tons of payday lenders they'reeverywhere.
So, as a result, I think banksfeel a lot of pressure in trying
to keep up.

Oh, for sure.

Let's talk about employment risk with banks using
AI to reduce their headcount.
How might that trigger EPLclaims, especially around like
age and wage discrimination?

I mean I saw wasn't it in the news, like last
week, guys, that Microsoft laidoff like or maybe it wasn't
Microsoft, maybe it was somebodyelse, but I feel like it was
Microsoft laid off like 9000.
I think it was Microsoft Intheir drive toward more AI
implementation.
It could be lay off a smallnumber, could be a large number
like that, but it seems like tome you could end up in some

(15:14):
litigation over it, implementingsomething like this and
eliminating.
Are there rules about that,mark?
Can you just do that?

You can do it, but the employment laws that exist
are the same, whether it'sAI-driven or any other reason.
You have to go about itmethodically and you need to
provide severance.
There needs to be propermessaging on the front end.
And the way I kind of got onthis platform originally was I

(15:50):
was in a meeting with one of myclients and the CEO was saying
in a very non-threatening wayyou know, we think we're going
to improve our expense ratiobecause we'll be able to use AI
chatbots more and that meanswe'll need fewer employees and
of the employees we keep, wewon't have to pay as much

(16:11):
overtime to them because thechatbot will take care of that
for us.
And the danger with that type ofcomment is it can be twisted
quickly to oh, you're just goingto use this to get rid of all
your older employees who are tooexpensive.
It may be an innocent attemptby the bank to become more
competitive.
It's going to be twisted by theplaintiff's bar as you use this

(16:33):
as a weapon to get rid of ourolder employees or our minority
employees or whoever.
I just think banks have to bereally protective of their brand
when they're doing these things, because it'll get used against
them quickly.
Even if it was never the intent.
We didn't employ AI to try andget rid of employees.
It happens to be a byproduct ofwhat we're doing and I think we

(16:57):
, as our insureds, need to bemore sensitive to that brand and
the goodwill that goes with it,because it could really harm us
and we might not recover fromit.

Well, and you mentioned Microsoft.
Remember when Microsoft's, theextent of Microsoft's AI was
Clippy Remember.

Clippy yes, I remember Clippy when he'd wave
at you, can I help you?

Well, they've come a little bit further than that
now, yeah.
And we've seen the Microsoftand other AI chatbots give
wildly inaccurate information,to put it mildly.
And you know we're talkingabout E&O, which stands for what
?
Kids, errors and Omissions.
So how do you see somethinglike that these chatbots giving

(17:43):
bad information?
How does that intersect withtraditional E&O coverage?

Great question.
The policies are silent in thisspace, so as a result, it's
very difficult to say where abanker's professional liability
policy that's sort of a bank'sE&O coverage versus their cyber
protection might trigger.
I still think it belongs withinthe banker's professional realm

(18:10):
.
That's where the coverageshould reside.
The fact that the medium inwhich it's being delivered
shouldn't really matter in myopinion, but one could argue if
there was a software errorembedded within the chatbot,
does that take us back towards acyber discussion?

(18:31):
Don't know the answer.
The losses, historically, oncewe get there, will answer that
question.
There will answer that question.
But what I tell everybody is Ican see chatbots moving in a
direction where they either havea file on you because you've
used it before as an accountholder of Bank of America, and

(18:53):
in that example Erica says hey,we've noticed that as long as
your income hasn't changed muchand you haven't taken on a lot
of debt, you'd be eligible for amortgage up to X.
Or have you thought about ahome equity loan?
That's where I think we'regoing to find ourselves in
trouble, because the chat bot ismaking assumptions on what your
capacity might be to take onmore debt.

(19:16):
There's a million detours alongthat road, and most of them are
bad.
They're all going to lead tobad conclusions.

Well, it kind of answers our next question I was
going to ask so are we headedtoward this wave of disputes and
denied claims?
Because as far as I understandit, it's not like we're seeing a
huge wave of these claims yet.
It's not like we're seeing ahuge wave of these claims yet as
far as chatbot errors or EPLclaims, et cetera, but it sounds

(19:47):
like they're coming, and untilthose are adjudicated or settled
or whatever, I don't know, Iguess the answer is no one
really knows yet how policies orcoverage or things are going to
respond.

Yeah, and remember , there's no regulatory guidance
either.
So the bank regulators, theydon't understand or they don't
really know how to guide andadvise all of their clients.
So the banks are sitting backtrying to develop best practices
but there's no one reallygiving them good direction.
So I've said to everyone youneed to try and be best in class

(20:25):
so that when the regs come outyou're best positioned to
already be in compliance.

That kind of leads into your final question,
Scott.

I was going to say we covered this then.
The regulatory side is stilldeveloping, but it's coming.
So what should banks and theirinsurance partners be doing now
to prepare?

I mean I would start with contract review.
I would start with identifyingwhere the liability resides
unfortunately, mostly with theconstituent banks and then I
would encourage them to buildout the most robust controls
that they can do so that it'sready if and when regulatory

(21:09):
requirements come out, becausethat's usually the way the game
plays is by the time the regscome out.
The banks have already beenoperating in an environment with
that tool AI being one of themfor a while, so they've
developed a workflow, et cetera.
Now they've got to change allthat, otherwise they're not in
compliance.
I still think there's a hugereluctance by our regulators to

(21:32):
get anywhere near the techsector.
We've seen it in other wayswhen it just comes to even the
placement of cyber insurance forthe banks.
That bank regulator typicallyis just checking a box.
Do you have a policy, yes or no?
Is it any good?
Is the limit adequate?
Is it covering?

(21:53):
You know, maybe there's aransomware supplement?
I don't think the regulatorsare drilling very deep.
They're just.
This is part of their overallaudit when they visit a bank and
it's probably a very quick yeah.
I looked at a deck page and Ican see that they buy a policy
from a carrier.
That's all they need to know.

I bet you could get lost in the weeds in that
very quickly, though, when youdrill down.

Yeah, yeah, it can be you get a little dizzy
trying to sort through all the.

But Mark is a specialist at that.
He lives in the weeds with likehe looks through all of this.
I know he does, Cause we'vetalked about it.

Yeah, it keeps me up at night.
But the good news is I thinkthere's, uh, there's, there's
plenty of work here to for us todo as brokers and, uh, I think
we have the right culture at CRCthat we're going to slow down
and do it once and do it right,because I pick up a lot of
programs from my competitors andI'm shocked sometimes that I'm

(22:52):
surprised the cyber just doesn'tget more scrutiny, and I view
AI as an extension of that.
People just don't.
I think they're looking at ittoo much, as our placements are
in silos and unfortunately,they're all interconnected and
the banking industry is a greatmicrocosm to that Everything

(23:12):
that you're going to have onewrongful act but you're going to
have multiple policiestriggering, so I think that the
AI discussion falls into that aswell.

And I'm sure it's something we'll talk about
again with you.
Mark, I have no doubt We'll seeyou again in another podcast.

Along the way, Unless this all goes away in the
next few months, right?

Right, that's totally happening.
Well, mark, before we let yougo, it's time for a quick
lightning round.
We like to call it rapid fire.
So we've got three rapid firequestions for you.
All you have to do is saywhatever comes to mind first.
So question number one what'sone piece of tech you can't live

(23:55):
without?
Is there one?

That's easy, just your iPhone.

If you weren't in insurance, what job would you
be doing?

I almost went to law school and this is a detour,
so I'm glad I took the detour.

Although, with your attention to detail, I'll
just go ahead and say I thinkyou would have been a great
lawyer.
All right, last one what's yourgo-to coffee order or workday
ritual?

Yeah, I'm not really a coffee drinker, but I
would say that that daily ritualis.
I like to get started realearly, so all the email traffic
that's already backing up, I'mready for it.
And that way, by the time I getto the office, if there's a
fire, I've already dealt with it, or at least I've already
thought it through so during mycommute with it, or at least

(24:40):
I've already thought it throughso during my commute I'm already
, you know, letting the wheelsturn, so I'm not panicked,
whereas if I don't get startedearly, next thing you know you
got a hundred emails and they'reall rushes.

Ease up people.
They keep him awake at night,y'all Ease up on Mark.

Okay, hold off on that email till the next day.
I mean, don't ease up, justmake it a little more pointed
into the point.
Mark doesn't have all day toread paragraphs full of info,
that's right.

Well, Mark, thanks for joining us and
sharing your insights.
We appreciate it.

Appreciate it.
Thanks for having me.
The risks surrounding AI andbanking are complex, but it's
clear.
Agents and brokers who stayinformed will be better
positioned to protect theirclients.

For more insights , check out our tools and intel
articles on crcgroupcom andconnect with your CRC specialty
broker to stay ahead of emergingrisks.
Thanks for listening.
We'll see you all next time.

All Episodes

Episode Transcript

Popular Podcasts

Stuff You Should Know

Dateline NBC

The Burden

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}AI in Banking: The Risk Beyond the Reward

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Stuff You Should Know

Dateline NBC

The Burden

All Episodes

AI in Banking: The Risk Beyond the Reward

Stuff You Should Know