Name: SmartStream at Money20/20 Europe – Streamlining data reconciliation with AI
Uploaded: 2023-06-21T16:52:19.267Z
Duration: 360.28

SmartStream-12_06.mp4

SmartStream-12_06.vtt

English

{{00:00:08}}
Hi, I'm Alex Pugh.

{{00:00:09}}
I'm the reporter for FinTech Futures.

{{00:00:11}}
I'm Money 2020 Europe, Amsterdam in 2023.

{{00:00:16}}
I'm with Jethro McDonald from SmartStream.

{{00:00:18}}
We're gonna be talking about

{{00:00:19}}
SmartStream air Hi Jeth, how you doing?

{{00:00:22}}
Do you wanna introduce yourself?

{{00:00:24}}
Yeah, of course.

{{00:00:24}}
Hi, I'm, I'm Jethro McDonald.

{{00:00:26}}
I'm the product manager of AI Machine

{{00:00:27}}
Learning at Smart Stream Technologies.

{{00:00:30}}
Smart Stream Air, it was launched 2019.

{{00:00:34}}
Can you tell us a bit about how

{{00:00:36}}
the techs evolved since then?

{{00:00:37}}
Yeah, of course.

{{00:00:38}}
We've gone from a solution that was

{{00:00:40}}
originally based in the cloud to now

{{00:00:41}}
being a fully cloud native solution.

{{00:00:44}}
And by that what we mean is we are

{{00:00:47}}
able to deploy the technology in a

{{00:00:49}}
way that when we need more processing

{{00:00:51}}
power and things like that, we can

{{00:00:52}}
actually scale up more machines.

{{00:00:54}}
And scale right back down,

{{00:00:55}}
which from a reconciliation

{{00:00:57}}
point of view is really key.

{{00:01:00}}
The bulk of processing power is quite at

{{00:01:01}}
the point that data is loaded and matched.

{{00:01:03}}
And with the way that we've deployed

{{00:01:05}}
the technology and the way that

{{00:01:06}}
it's evolved, we now do that in

{{00:01:07}}
a very cost effective manner.

{{00:01:08}}
Cool.

{{00:01:09}}
So how does it leverage both the

{{00:01:12}}
cloud and artificial intelligence?

{{00:01:15}}
What functions they each.

{{00:01:17}}
Perform, how do they work together?

{{00:01:18}}
Yeah, of course.

{{00:01:19}}
So air in its name, AI reconciliation.

{{00:01:22}}
AI is in its dna.

{{00:01:24}}
So what we look to do with the AI

{{00:01:26}}
components is three main things:

{{00:01:28}}
help clients quickly onboard new

{{00:01:30}}
reconciliations, help them with

{{00:01:32}}
any manual work they might have

{{00:01:33}}
when manually matching data.

{{00:01:35}}
And then also any manual work

{{00:01:37}}
they might have when managing

{{00:01:38}}
their exceptions as well.

{{00:01:39}}
So AI is a key feature of

{{00:01:41}}
what we have in the solution.

{{00:01:43}}
And then from a cloud point We

{{00:01:46}}
utilize that technology in a way that

{{00:01:47}}
allows us to be very cost effective.

{{00:01:50}}
Having a fully cloud native solution

{{00:01:52}}
where we can scale up when we need

{{00:01:53}}
processing power and scale right

{{00:01:55}}
back down when we don't need that.

{{00:01:56}}
Processing power is a very cost

{{00:01:58}}
effective and efficient way of

{{00:02:00}}
deploying a solution like ours.

{{00:02:03}}
It means that we can now do volumes

{{00:02:05}}
that are unheard of in our industry.

{{00:02:07}}
Not the millions more into the

{{00:02:09}}
billions, things like that, right?

{{00:02:10}}
Just by the clever ways that we utilize

{{00:02:12}}
technology to, to help with that process.

{{00:02:14}}
Cool.

{{00:02:15}}
So when it comes to data what sort

{{00:02:18}}
of things can go wrong when data's

{{00:02:20}}
moved from one location to another?

{{00:02:24}}
And how does smart stream air

{{00:02:26}}
tackle those potential problems?

{{00:02:28}}
Yeah, of course.

{{00:02:29}}
One of the things that happens in

{{00:02:31}}
big banks, another institutions, is

{{00:02:32}}
when they're moving data from one

{{00:02:34}}
system, maybe the front office down

{{00:02:35}}
to another system or external outside

{{00:02:38}}
their organization, things can happen

{{00:02:41}}
Processes that they have could fail,

{{00:02:43}}
or data may not be enriched correctly.

{{00:02:46}}
What AIR is really good at is actually

{{00:02:48}}
looking and analyzing data across

{{00:02:50}}
those systems and understanding where

{{00:02:51}}
there are issues, where there are

{{00:02:52}}
problems, and showing those in a very

{{00:02:54}}
quick and clean way for users to then

{{00:02:56}}
understand what issues they have.

{{00:02:59}}
And, how does it turn this data, this

{{00:03:01}}
messy data into Hey, usable data.

{{00:03:04}}
Actionable data?

{{00:03:05}}
Yeah, so we, have a number of.

{{00:03:08}}
Touch points at the kind of

{{00:03:09}}
ingestion side of our solution.

{{00:03:11}}
So outta the box, we support loads

{{00:03:13}}
of different formats but even formats

{{00:03:15}}
that we don't support, we have

{{00:03:16}}
integrations that will allow us to adapt

{{00:03:18}}
and ingest that data so it can take

{{00:03:20}}
unstructured data and whatever format.

{{00:03:23}}
It will normalize that into a

{{00:03:24}}
clean way onto a customizable

{{00:03:27}}
schema based on the data.

{{00:03:28}}
So each rec can have a different schema

{{00:03:31}}
and then when it's done that it's able

{{00:03:32}}
to then correlate and check something

{{00:03:34}}
on side A versus something on side B.

{{00:03:36}}
Because it's normalized at data,

{{00:03:37}}
it can then tell you where there

{{00:03:38}}
might be breaks or differences.

{{00:03:40}}
Very clever.

{{00:03:41}}
How does it tackle the issues

{{00:03:43}}
around black box solutions?

{{00:03:47}}
There's a lot of talk with

{{00:03:48}}
AI that they are black boxes.

{{00:03:50}}
People don't really know how they work.

{{00:03:52}}
They, spit out a, an answer.

{{00:03:55}}
And we are not quite sure how

{{00:03:56}}
they arrived at that conclusion.

{{00:03:59}}
Is, it more transparent than

{{00:04:01}}
other black box solutions?

{{00:04:02}}
Yeah, of course.

{{00:04:02}}
It's, that's one of the main concerns

{{00:04:04}}
that people have when they talk about

{{00:04:05}}
these new technologies and AI and, ml.

{{00:04:08}}
We've invested probably more time on

{{00:04:11}}
the transparency of the decisions that

{{00:04:12}}
we've made than actually the, AI itself.

{{00:04:14}}
Because as an end user, you want

{{00:04:16}}
to be sure that what has been

{{00:04:18}}
decided by the machine is correct.

{{00:04:19}}
So what we do every step

{{00:04:21}}
of the way is we show them.

{{00:04:22}}
What data we believe correlates

{{00:04:24}}
with what on the other side.

{{00:04:25}}
So you can physically see that.

{{00:04:26}}
We also then show you why

{{00:04:28}}
we've driven that decision.

{{00:04:30}}
Basically the percentage of how

{{00:04:31}}
confident we are that this bit of

{{00:04:33}}
data matches with this bit of data.

{{00:04:35}}
And with it being fully transparent, even

{{00:04:37}}
if you as a user as aside, oh no, I don't

{{00:04:39}}
believe this is correct, you can just

{{00:04:40}}
query it quickly and easily changed that.

{{00:04:43}}
And then one of the other things that we

{{00:04:45}}
have as well is say if you are matching

{{00:04:47}}
data and you've got like a small substring

{{00:04:49}}
on one side against something else,

{{00:04:51}}
we highlight and show you and say, Hey

{{00:04:52}}
look, this reference here is actually

{{00:04:54}}
here on the civil record as well.

{{00:04:55}}
So every step of the way it's

{{00:04:57}}
very transparent to a user exactly

{{00:04:59}}
what the system's doing and

{{00:05:00}}
why it's made those decisions.

{{00:05:01}}
It always shows you it's working

{{00:05:03}}
like last homework kind of thing.

{{00:05:05}}
Exactly.

{{00:05:05}}
So it's always show your answers.

{{00:05:06}}
Yeah.

{{00:05:07}}
Yeah, exactly.

{{00:05:09}}
What other industry sectors do you

{{00:05:11}}
think this solution could be useful in?

{{00:05:15}}
Yeah, so we as a firm have

{{00:05:16}}
traditionally been working with

{{00:05:19}}
banks from all different sizes.

{{00:05:21}}
But what we're seeing with this solution

{{00:05:22}}
now, with the way that it's deployed

{{00:05:24}}
and the way we use technology so we can

{{00:05:26}}
actually open up to more clients different

{{00:05:29}}
corporates that also have a matching

{{00:05:31}}
problem or a cash problem as well.

{{00:05:35}}
We are looking at really

{{00:05:36}}
expanding into those spaces.

{{00:05:38}}
And we think we've got a really good

{{00:05:40}}
solution that is both very good at what

{{00:05:42}}
it does and cost efficient for those users

{{00:05:43}}
that maybe don't have the same kind of

{{00:05:45}}
wallets as a, tier one investment bank.

{{00:05:48}}
Great.

{{00:05:49}}
That's it from us at Money

{{00:05:51}}
2020 Europe in Amsterdam.

{{00:05:52}}
Thanks, Jethro.

{{00:05:53}}
Thank you.

{{00:05:54}}
A pleasure.

{{00:05:55}}
Thank you.

I'm Money 2020 Europe, Amsterdam in 2023.

I'm with Jethro McDonald from SmartStream.

the point that data is loaded and matched.

technology to, to help with that process.

FinTech Futures

Expert Insight

At this year’s Money20/20 Europe conference in Amsterdam, FinTech Futures caught up with Jethro MacDonald, Product Manager of AI and Machine Learning at SmartStream Technologies, to discuss how the company utilises artificial intelligence and the cloud to power effective data reconciliation and how its SmartStream Air offering can compare and validate unstructured data of all types to create high-quality sets of results or exceptions to investigate.