FTF_Jean-Pierre Sleiman_1.mp4

FTF_Jean-Pierre Sleiman_1.vtt

English

{{00:00:04}}
It's  Fintech futures at

{{00:00:05}}
Finovate Europe and I'm joined

{{00:00:07}}
by John Pierre slime of N 26.

{{00:00:09}}
How are you?

{{00:00:10}}
Good.

{{00:00:10}}
And you Paul?

{{00:00:11}}
Excellent.

{{00:00:11}}
I'm very well, thank you.

{{00:00:12}}
Have you been enjoying

{{00:00:13}}
the conference so far?

{{00:00:15}}
Yeah I've arrived yesterday and yeah,

{{00:00:18}}
enjoyed the different conferences I

{{00:00:21}}
saw and discussed with some vendors.

{{00:00:23}}
Yeah, very nice so far.

{{00:00:25}}
Perfect.

{{00:00:25}}
Perfect.

{{00:00:25}}
Sounds good.

{{00:00:26}}
Just get started.

{{00:00:27}}
Let's give a quick introduction to

{{00:00:28}}
yourself then and, your role   at N26.

{{00:00:29}}
Yep, sure.

{{00:00:30}}
So my name is Jean-Pierre Sleiman I work

{{00:00:32}}
for N 26, I joined 3 years ago, and I'm

{{00:00:34}}
head of digital operations, which means I

{{00:00:37}}
support our operations department, which

{{00:00:39}}
includes customer service with digital

{{00:00:42}}
data, AI and automation initiative.

{{00:00:45}}
And before this I, worked on

{{00:00:47}}
similar topics as a consultant

{{00:00:48}}
in a large retail bank.

{{00:00:50}}
Excellent.

{{00:00:50}}
You're here speaking on a panel

{{00:00:52}}
about AI and banking then.

{{00:00:54}}
One of the big talking points

{{00:00:55}}
recently in AI has always been

{{00:00:56}}
Chat GPT, generative AI coming in.

{{00:00:59}}
I guess in terms of financial

{{00:01:01}}
services, how disruptive is that

{{00:01:03}}
technology potentially going to be?

{{00:01:04}}
An uh, should pins be

{{00:01:05}}
looking more at that?

{{00:01:08}}
It's a very good question.

{{00:01:08}}
We discussed it briefly

{{00:01:09}}
in the panel as well.

{{00:01:11}}
I think it's amazing,  what what

{{00:01:13}}
this kind of technology, generative

{{00:01:15}}
AI in general can do with text,

{{00:01:18}}
also images, videos, everything.

{{00:01:21}}
And we are already seeing

{{00:01:23}}
some use cases emerging.

{{00:01:24}}
We're seeing like that it can

{{00:01:27}}
be used to help developers,

{{00:01:28}}
for instance, with their code.

{{00:01:30}}
It can summarize articles.

{{00:01:31}}
It can, yeah, the answer

{{00:01:34}}
complex questions, taking into

{{00:01:36}}
account context, et cetera.

{{00:01:37}}
But I think we are also starting

{{00:01:39}}
to see the limits, basically.

{{00:01:42}}
And like we now can, find countless

{{00:01:44}}
examples of AI healthy nations basically,

{{00:01:47}}
or irrelevant or inconsistent answers

{{00:01:50}}
to questions which kind of show that

{{00:01:52}}
it's, currently difficult to use

{{00:01:55}}
these technologies for a business

{{00:01:57}}
purpose without human supervision.

{{00:01:59}}
And so for me, I would if I had to

{{00:02:02}}
summarize it, I think it's, a last

{{00:02:03}}
mile problem that we're facing now.

{{00:02:06}}
So we have a good we have good

{{00:02:08}}
demonstrators of what AI, In terms

{{00:02:11}}
of taking into account complex

{{00:02:13}}
inputs, getting collecting, let's

{{00:02:16}}
say data from everywhere and then

{{00:02:18}}
synthesizing into a, an output can be

{{00:02:21}}
a text, an image, a video as, I said.

{{00:02:24}}
So it's a good demonstrator of all

{{00:02:26}}
this chain and all these components.

{{00:02:28}}
But we still need to figure out, okay,

{{00:02:30}}
now how do we apply this in a business

{{00:02:32}}
context, especially in financial

{{00:02:34}}
services where we have very high

{{00:02:37}}
expectations in terms of accuracy, in

{{00:02:39}}
terms of compliance, and so I think

{{00:02:42}}
now will be a, at the very first, I

{{00:02:44}}
think end of last year, we were very,

{{00:02:47}}
optimistic now in the at least financial

{{00:02:50}}
services area we're very pessimistic.

{{00:02:52}}
I think some banks already said

{{00:02:53}}
that they will never use this.

{{00:02:56}}
I think we'll get to a third step,

{{00:02:58}}
which will be a little bit in between

{{00:03:00}}
where all the stakeholders, us

{{00:03:01}}
financial institutions, regulators,

{{00:03:03}}
many third parties will sit and

{{00:03:05}}
start to think, okay, where can

{{00:03:07}}
we apply this kind of technologies

{{00:03:10}}
in a secure and a compliant way.

{{00:03:13}}
Excellent, excellent.

{{00:03:14}}
Conversational AI has been around

{{00:03:15}}
for a while now in terms of like

{{00:03:16}}
chatbots and things like that.

{{00:03:18}}
Given how quickly the space that evolves,

{{00:03:20}}
so is there any kind of new developments

{{00:03:22}}
there that you can tell us about?

{{00:03:23}}
Well yeah, I think it will relate to

{{00:03:27}}
the, questions just before I think one of

{{00:03:30}}
the very interesting things that we see

{{00:03:32}}
emerging conversational AI is something

{{00:03:33}}
called large language models, LLMs.

{{00:03:36}}
So it's it's one of the components

{{00:03:38}}
of Chat GPT for instance.

{{00:03:39}}
And it basically enables to take into

{{00:03:43}}
account more complex inputs from from

{{00:03:46}}
the human users and take into account

{{00:03:49}}
more contextual, let's say inputs given.

{{00:03:52}}
And also manage conversations at my

{{00:03:54}}
shift, jump from a topic to another and

{{00:03:57}}
basically manage this, whole complexity

{{00:03:59}}
with more complex and powerful models.

{{00:04:02}}
So I think, yeah, this is something that

{{00:04:05}}
will, that is starting to be applied

{{00:04:07}}
in the conversational AI area and is

{{00:04:10}}
definitely one of the next big things,

{{00:04:12}}
if not one of the current big things.

{{00:04:16}}
Excellent.

{{00:04:17}}
So you've got neon at N 26

{{00:04:18}}
that's your own chat  bot.

{{00:04:21}}
With my experience with chatbot,

{{00:04:22}}
sometimes, you know, requests can be

{{00:04:24}}
quite niche in terms of what they're

{{00:04:26}}
asking and some people obviously still

{{00:04:28}}
prefer being able to speak to a human.

{{00:04:30}}
But I mean, what kind of

{{00:04:30}}
success have you had with that?

{{00:04:31}}
Have you been able to tackle some of

{{00:04:33}}
these like niche elements as well?

{{00:04:34}}
No, definitely.

{{00:04:36}}
So we started working on conversational

{{00:04:39}}
AI at N 26 more than four years ago.

{{00:04:44}}
And at that time it was at some

{{00:04:45}}
point overhyped and then overlooked.

{{00:04:48}}
Kind of a bumpy road for the technology,

{{00:04:50}}
but we kept believing it in it.

{{00:04:51}}
We kept investing time and effort to

{{00:04:55}}
because we could see the potential,

{{00:04:56}}
because yes, there are niche questions,

{{00:04:59}}
especially in financial services,

{{00:05:01}}
but there are also more basic

{{00:05:05}}
situations where we can basically

{{00:05:07}}
provide an answer without a human.

{{00:05:09}}
And so we built like several bots that

{{00:05:13}}
are now dealing with customers directly

{{00:05:16}}
on a day-to-day basis in five languages

{{00:05:18}}
we manage to integrate to our systems,

{{00:05:20}}
which means that they can, they know like

{{00:05:23}}
some characteristics of our customers.

{{00:05:26}}
They can also act, sometimes

{{00:05:27}}
trigger some actions directly

{{00:05:30}}
if the customer requests it.

{{00:05:32}}
And and yeah, as I said, in managing

{{00:05:35}}
millions of conversations a year

{{00:05:37}}
automating the third of them, so I

{{00:05:39}}
think it's pretty amazing achievements

{{00:05:41}}
and it's still progressing.

{{00:05:43}}
So yeah, now I see a pretty bright

{{00:05:45}}
future for this kind of technologies.

{{00:05:47}}
Excellent, How do banks such as  yourself

{{00:05:49}}
then ensure accuracy when they're

{{00:05:50}}
developing this kinda technology?

{{00:05:53}}
Of course, yeah.

{{00:05:54}}
Accuracy is super important.

{{00:05:56}}
As I said before we're in a

{{00:05:58}}
highly regulated industry.

{{00:06:00}}
We have also customers with

{{00:06:02}}
very high, who expect a lot

{{00:06:06}}
from us, very high expectations.

{{00:06:08}}
Especially as we're dealing with

{{00:06:10}}
something that is very important

{{00:06:11}}
for them, which is their funds.

{{00:06:12}}
And so I think we need to, ensure

{{00:06:16}}
accuracy and that what we, anything we

{{00:06:18}}
provide to the customers is relevant.

{{00:06:20}}
I think we need to think about

{{00:06:23}}
AI as an operating model.

{{00:06:24}}
It's not just a fancy tech

{{00:06:26}}
thing that can do amazing stuff.

{{00:06:28}}
It's something that requires a pretty

{{00:06:32}}
clear like governance, pretty clear

{{00:06:34}}
roles and responsibilities mixing

{{00:06:37}}
and different areas, sometimes

{{00:06:40}}
external vendors, sometimes not,

{{00:06:41}}
but at least internal tech companies

{{00:06:43}}
and it product and business teams.

{{00:06:48}}
Certainly compliance at some point

{{00:06:49}}
to make sure that everything we

{{00:06:50}}
do is according to the rules.

{{00:06:54}}
And then what I would really like

{{00:06:57}}
to insist on, I think what is

{{00:06:58}}
super important is to also take

{{00:07:01}}
into account what happens after.

{{00:07:04}}
So releasing a nice bot, okay,

{{00:07:06}}
this is cool, but it's just

{{00:07:08}}
the beginning of the story.

{{00:07:09}}
In the end, your internally your

{{00:07:13}}
company will change, the tech

{{00:07:15}}
will change, the products will

{{00:07:16}}
change, the people will change

{{00:07:17}}
externally, things will also change.

{{00:07:19}}
The economic context will change the

{{00:07:21}}
customers, they will change, et cetera.

{{00:07:22}}
etc.

{{00:07:23}}
And whatever ai and conversational

{{00:07:26}}
AI in particular, you will release.

{{00:07:29}}
It'll need to remain relevant.

{{00:07:31}}
And to do this, you need

{{00:07:32}}
to put in place controls.

{{00:07:33}}
You need to put in place mechanisms

{{00:07:34}}
that enable you to audit.

{{00:07:36}}
Make sure that whatever you provide as

{{00:07:38}}
an AI output is relevant to the users

{{00:07:41}}
and correct if needed, and correct fast.

{{00:07:45}}
If you detect something wrong,

{{00:07:47}}
you can correct it in a matter

{{00:07:48}}
of hours or days maximum.

{{00:07:50}}
And I think this is what enables the

{{00:07:53}}
to, remain relevant and accurate.

{{00:07:55}}
Excellent, and just finish off

{{00:07:56}}
then, do you see a future where

{{00:07:57}}
all digital interactions can

{{00:08:00}}
be fully automated by an ai?

{{00:08:02}}
Again, could you just have a conversation

{{00:08:03}}
with an AI chat bot and you know, not

{{00:08:04}}
even know that you're speaking to?

{{00:08:09}}
Well, I think financial services

{{00:08:10}}
are a complex space as a whole.

{{00:08:15}}
It's, first of all it's, difficult

{{00:08:17}}
and there are many niche problems

{{00:08:20}}
that can occur as a user, B2B or b2c.

{{00:08:25}}
And second it's it's an industry where

{{00:08:28}}
the relationship is very important in the

{{00:08:31}}
end It'll make a big difference whether

{{00:08:35}}
you have a good relationship with your

{{00:08:36}}
bank or financial institution or not.

{{00:08:38}}
It'll be a a major criteria

{{00:08:40}}
to to decide to stay alive.

{{00:08:42}}
And for these two reasons I am

{{00:08:45}}
pretty convinced that we will

{{00:08:46}}
always need humans in the loop to

{{00:08:49}}
deal with these very complex issues

{{00:08:51}}
or very ad hoc niche problems.

{{00:08:55}}
And to create and maintain this this

{{00:08:58}}
relationship and also commercial.

{{00:09:02}}
Now I'm also convinced that these

{{00:09:04}}
human interactions will more and

{{00:09:05}}
more be assisted and augmented with

{{00:09:08}}
AI with recommendations on the best

{{00:09:11}}
resolution path or on the best product

{{00:09:14}}
or recommendation for the customer.

{{00:09:16}}
And to, yeah, also answer the

{{00:09:18}}
last part of your questions.

{{00:09:19}}
Of your question, sorry.

{{00:09:22}}
I believe in transparency.

{{00:09:23}}
So I think that whenever we should be

{{00:09:25}}
very clear when the users are interacting

{{00:09:29}}
with the machine or with the human.

{{00:09:32}}
Okay.

{{00:09:32}}
Well thank you.

{{00:09:32}}
This taking the time to

{{00:09:33}}
speak with Jean-Pierre.

{{00:09:34}}
It's been a pleasure.

{{00:09:35}}
Thank you.

{{00:09:36}}
Excellent.

{{00:09:36}}
Thank you.

head of digital operations, which means I

examples of AI healthy nations basically,

At the FinovateEurope 2023 conference in London, FinTech Futures sat down with Jean-Pierre Sleiman, head of digital operations at N26, to discuss how artificial intelligence (AI) applications such as conversational AI and generative AI have disrupted the financial services sector and how the tech could evolve going forward. 