Name: Code & Capital - Ai as a co-pilot in financial services
Uploaded: 2024-06-12T13:06:29.562Z
Duration: 374.84
Description: Code & Capital explore the uses of Ai being a co-pilot in financial services at Finovate Spring 2024 with Sarah Hinkfuss Partner at Bain Capital Ventures.

Code & Capital - AI as a Co-pilot in financial services.mp4

Code & Capital - AI as a Co-pilot in financial services.vtt

ENGLISH

{{00:00:17}}
So banks, just like all large
organizations, are at the forefront

{{00:00:20}}
of thinking about the real world
applications of generative AI.

{{00:00:24}}
And some of the first places that we
see those applications come to full

{{00:00:28}}
life is actually in the spaces that
also apply to other enterprises.

{{00:00:32}}
And so that could be customer
service, or it could be new

{{00:00:36}}
product development as well.

{{00:00:37}}
And those areas have more full feature,
horizontal AI apps companies that are

{{00:00:44}}
ready to actually serve the customer.

{{00:00:46}}
And so in the customer service example,
for instance, we see that what you can do

{{00:00:50}}
is actually take the entire compendium of
all of the information on all of the FAQs

{{00:00:55}}
that you've written, all of the back and
forth emails and calls, and use that to

{{00:00:59}}
train agents who can then respond to the
first order level of questions in order

{{00:01:04}}
to Say, in order to open up time for your
customer service reps to answer the more

{{00:01:08}}
complex questions for your customers.

{{00:01:17}}
Yeah, we think about the opportunity
for generative AI and financial services

{{00:01:21}}
as going across the entire value chain.

{{00:01:23}}
And so that starts from the distribution
of the financial service product itself.

{{00:01:27}}
So how our customer is actually
being made aware of it.

{{00:01:31}}
And so what is the way that a
customer can actually ask questions or

{{00:01:35}}
interrogate what that product is and
understand it in layman's language.

{{00:01:39}}
It's also true in the way that
the product is actually made.

{{00:01:42}}
So what is the manufacturing of
that financial services product?

{{00:01:46}}
And that's where, as you ask about,
it's the underwriting process.

{{00:01:49}}
So one example in underwriting would
be taking all of the discrete data

{{00:01:54}}
that an individual is submitting.

{{00:01:56}}
on the, on their behalf for a loan
application and pulling that data and

{{00:02:00}}
actually putting it into a structured form
and then analyzing the elements of it that

{{00:02:05}}
are related to different fraud components.

{{00:02:07}}
And then it's also leveraged
in servicing as well.

{{00:02:10}}
And so the way that a customer who has a
product can follow up and ask questions,

{{00:02:14}}
or if they want to make a modification
to the loan, or if there's a place that

{{00:02:17}}
they're moving, all of those changes.

{{00:02:19}}
And the servicing of it can actually
be done in a much more efficient

{{00:02:22}}
way as well using generative AI.

{{00:02:30}}
So transparency and fairness,
I think two are the most

{{00:02:33}}
important considerations for AI.

{{00:02:35}}
And I think what's interesting is often
in this conversation, people assume

{{00:02:39}}
that AI will be worse than humans.

{{00:02:41}}
The reality is that the baseline
that we're starting from has a lot

{{00:02:45}}
of bias already today in the way that
humans are evaluating financial data.

{{00:02:49}}
and making decisions around who
should have access to a product,

{{00:02:52}}
at what level, and at what price.

{{00:02:55}}
And I actually think about the opportunity
being greater to infuse fairness within

{{00:03:00}}
all of the distribution, production,
and servicing of financial services

{{00:03:04}}
apps, because all of those models will
be under higher scrutiny, and you can

{{00:03:09}}
test or evaluate against different
metrics to ensure that fairness is

{{00:03:13}}
used throughout the entire process.

{{00:03:15}}
and not just today in the particular
regulatory steps where that's true.

{{00:03:26}}
Some of the key benefits for AI in
financial services are to take what

{{00:03:30}}
is a very laborious, services heavy
product today and create efficiencies

{{00:03:36}}
to streamline and automate the process.

{{00:03:38}}
The reason why that matters so much
is because so much of the cost and the

{{00:03:43}}
difficulty in extending financial services
products to all businesses and people who

{{00:03:48}}
need them is because of the cost that goes
in from the actual services components.

{{00:03:52}}
And so as I think about the benefits, it's
actually making something that is hard to

{{00:03:56}}
understand, but critical for the economy,
much easier to understand and able to be

{{00:04:00}}
in the hands of everyone who needs it.

{{00:04:02}}
In terms of the risks or the complexities,
there has to be auditability.

{{00:04:07}}
And so once I have the data, once I
have a system that's actually embedded

{{00:04:12}}
within my overall organization, I
can't just assume that there's a

{{00:04:15}}
black box that tells me what to do.

{{00:04:17}}
But instead, the solutions that are
able to actually show me the precedent

{{00:04:21}}
information, where it came from and
what it means, that actually drives

{{00:04:25}}
the decision, those are the ones that
those are the ones that will be adopted

{{00:04:28}}
within these organizations and be
approved by the regulators as well.

{{00:04:37}}
So I have a little bit of a hot
take on the first question around

{{00:04:41}}
how to address the obstacles.

{{00:04:43}}
And that is one of the challenges, I
think, in the first wave of how we thought

{{00:04:47}}
about implementing generative AI was
that we still wanted to retain control.

{{00:04:51}}
So we only wanted to use the co pilot that
sat next to us that we could ask questions

{{00:04:55}}
to when we wanted, but we didn't trust
that co pilot to actually do work for us.

{{00:05:00}}
And what I've seen in conversations with
hundreds of companies now is that when

{{00:05:04}}
we actually give up more control to AI.

{{00:05:06}}
and allow it to automate our work,
we're able to see greater benefits.

{{00:05:11}}
So we're able to see a higher ROI on our
time, we're able to see actually a safer

{{00:05:16}}
and more consistent application of AI, and
we're able to see an easier implementation

{{00:05:21}}
because it requires less change management
when we're actually automating our work.

{{00:05:25}}
In terms of how you continue to stay up
to date, the best thing that I can say is

{{00:05:29}}
to figure out how to experiment with it.

{{00:05:31}}
So to do it in a safe way, in a closed
garden where you're not concerned about

{{00:05:36}}
your data or the data of customers, of
course, being shared with anyone else,

{{00:05:41}}
but allowing different teams across the
organization to try it and use it with

{{00:05:45}}
their own data and their own processes,
and then have a very fast evaluation

{{00:05:50}}
process internally to determine which of
those applications actually produce real

{{00:05:55}}
ROI and continue those and roll those
out across the organization and kill

{{00:05:59}}
those that are actually not successful.

{{00:06:01}}
But the best thing that I can say is
instead of having it be commanded by one

{{00:06:04}}
place in the middle, it's really important
to let the people at the front lines

{{00:06:08}}
actually try and see what works for them.

So banks, just like all large
organizations, are at the forefront

of thinking about the real world
applications of generative AI.

And some of the first places that we
see those applications come to full

life is actually in the spaces that
also apply to other enterprises.

And so that could be customer
service, or it could be new

And those areas have more full feature,
horizontal AI apps companies that are

And so in the customer service example,
for instance, we see that what you can do

is actually take the entire compendium of
all of the information on all of the FAQs

that you've written, all of the back and
forth emails and calls, and use that to

train agents who can then respond to the
first order level of questions in order

to Say, in order to open up time for your
customer service reps to answer the more

Yeah, we think about the opportunity
for generative AI and financial services

And so that starts from the distribution
of the financial service product itself.

So how our customer is actually
being made aware of it.

And so what is the way that a
customer can actually ask questions or

interrogate what that product is and
understand it in layman's language.

It's also true in the way that
the product is actually made.

So what is the manufacturing of
that financial services product?

And that's where, as you ask about,
it's the underwriting process.

So one example in underwriting would
be taking all of the discrete data

on the, on their behalf for a loan
application and pulling that data and

actually putting it into a structured form
and then analyzing the elements of it that

are related to different fraud components.

And then it's also leveraged
in servicing as well.

And so the way that a customer who has a
product can follow up and ask questions,

or if they want to make a modification
to the loan, or if there's a place that

And the servicing of it can actually
be done in a much more efficient

So transparency and fairness,
I think two are the most

And I think what's interesting is often
in this conversation, people assume

The reality is that the baseline
that we're starting from has a lot

of bias already today in the way that
humans are evaluating financial data.

and making decisions around who
should have access to a product,

And I actually think about the opportunity
being greater to infuse fairness within

all of the distribution, production,
and servicing of financial services

apps, because all of those models will
be under higher scrutiny, and you can

test or evaluate against different
metrics to ensure that fairness is

and not just today in the particular
regulatory steps where that's true.

Some of the key benefits for AI in
financial services are to take what

is a very laborious, services heavy
product today and create efficiencies

The reason why that matters so much
is because so much of the cost and the

difficulty in extending financial services
products to all businesses and people who

need them is because of the cost that goes
in from the actual services components.

And so as I think about the benefits, it's
actually making something that is hard to

understand, but critical for the economy,
much easier to understand and able to be

In terms of the risks or the complexities,
there has to be auditability.

And so once I have the data, once I
have a system that's actually embedded

within my overall organization, I
can't just assume that there's a

But instead, the solutions that are
able to actually show me the precedent

information, where it came from and
what it means, that actually drives

the decision, those are the ones that
those are the ones that will be adopted

within these organizations and be
approved by the regulators as well.

So I have a little bit of a hot
take on the first question around

And that is one of the challenges, I
think, in the first wave of how we thought

about implementing generative AI was
that we still wanted to retain control.

So we only wanted to use the co pilot that
sat next to us that we could ask questions

to when we wanted, but we didn't trust
that co pilot to actually do work for us.

And what I've seen in conversations with
hundreds of companies now is that when

and allow it to automate our work,
we're able to see greater benefits.

So we're able to see a higher ROI on our
time, we're able to see actually a safer

and more consistent application of AI, and
we're able to see an easier implementation

because it requires less change management
when we're actually automating our work.

In terms of how you continue to stay up
to date, the best thing that I can say is

So to do it in a safe way, in a closed
garden where you're not concerned about

your data or the data of customers, of
course, being shared with anyone else,

but allowing different teams across the
organization to try it and use it with

their own data and their own processes,
and then have a very fast evaluation

process internally to determine which of
those applications actually produce real

ROI and continue those and roll those
out across the organization and kill

But the best thing that I can say is
instead of having it be commanded by one

place in the middle, it's really important
to let the people at the front lines

actually try and see what works for them.

FinovateSpring

Finovate

Code & Capital

Expert Insight

Code & Capital explore the uses of Ai being a co-pilot in financial services at Finovate Spring 2024 with Sarah Hinkfuss Partner at Bain Capital Ventures.