Securing the Future: How Banks are Reducing Risk With Data and AI

Summary

• Fraud Detection Systems: Leading banks leverage the Databricks Data and AI platform to implement ML-driven fraud prevention, identifying vulnerabilities and protecting sensitive customer data.
• Regulatory Compliance: Unified data platforms enable banks to navigate complex regulations, streamline governance and enhance operational resilience while maintaining security.
• Risk Management Innovation: AI-driven strategies identify vulnerabilities, reduce costs and foster trust, allowing banks to tackle emerging threats without hindering growth.

Securing the Future: How Banks are Reducing Risk With Data and AI

Talk by Cyril Cymbler, Head of Global Banking & EMEA FINSERV, Databricks ; Gordon Wilson, Chief Information Officer - Americas, Sumitomo Mitsui Banking Corporation ; Nitin Kulkarni, CIO, Nationwide Building SOCIETY ; Thomas Sawyer, ED, Enterprise Architecture, Sumitomo Mitsui Banking Corp
Today, executives are focused on managing regulatory scrutiny and emerging threats. Banks worldwide are leveraging the Databricks Data and AI platform to enhance fraud prevention, ensure compliance and protect sensitive data while improving operational efficiency. this video will highlight how leading banks are implementing AI-driven risk management to identify vulnerabilities, streamline governance and enhance resilience. By utilizing unified data platforms, these institutions can effectively tackle threats and foster trust without hindering growth. Key takeaways: Fraud detection: Best practices for using machine learning to combat fraud Regulatory compliance: Insights on navigating complex regulations Secure operations: Strategies for scalable operations that protect assets and support growth Join us to see how data intelligence is reshaping the banking industry and enabling success in uncertain times!
Here’s more to explore:
Economist Impact: Unlocking enterprise AI: https://www.databricks.com/resources/webinar/unlocking-enterprise-ai
The State of Data + AI: https://www.databricks.com/resources/ebook/state-of-data-ai
See all the product announcements from Data + AI Summit: https://www.databricks.com/events/dataaisummit-2025-announcements

FAQs

How can banks detect fraud using machine learning?

Banks use the Databricks Data and AI platform to implement ML models that identify suspicious patterns in transaction data and prevent fraud at scale.

What compliance challenges do banks face?

Banks must navigate complex regulations while maintaining data security and privacy, and unified data platforms on Databricks help streamline governance.

How does AI improve operational efficiency?

AI-driven risk management identifies vulnerabilities, streamlines governance and enhances resilience, enabling banks to operate more efficiently.

What is the business impact of data and AI for banks?

Effective data and AI strategies help banks enhance fraud prevention, ensure compliance, protect assets and support growth in competitive markets.

Full transcript

Okay. Hello everyone. Welcome. Welcome to this session. Today, we are going to talk about how a robust data and AI strategy can unlock an efficient risk management for every stakeholder in that field. But before we start, let me introduce myself. I have the privilege to uh lead banking and EMEA uh and for financial services for the rest of EMEA. So, I'm privileged here to be with you today. And I will be the one providing an exotic accent to the session.
So, welcome everyone. I know that again, it's super quiet. It's uh it's right after the lunch break, so we'll try to make it interactive and interesting for all of you, ladies and gentlemen. Before diving into the hot topic of the day, I have a big favor to ask to each of you. We have decided at Databricks now to tailor our solutions to fit basically the challenge of the industries based on what do we hear from your side, what do we hear from our partners, what do we hear from the field.
And uh the breakdown is super simple. How Databricks can help you in generating new revenue streams, better protect the firm, and being more cost efficient. So, if you can just scan the QR code here, and pick up up to five uh elements which which are top priority for you where using data and AI and generative AI can be beneficial to your institutions, please fill it. It will help us again to deliver super focused and tailor solution to help you in answering these kind of challenges.
I will let this slide few minutes, few seconds. Minutes would be hard. Thank you very much. I know that it's a bit of homework again after after lunch. It is helping everyone to keep super awake. And this is basically what we are going to tackle today. As you probably know, risk is a top concern for the banks. In 2024, the cost of risk in the balance sheet of financial institutions has increased about 3 points. And basically what will happen in 2025, it's following a trend which could be around 4.
5 5% uh increase, which is even super important. So, without further delay, I would like to introduce the first keynote speaker of the day. He's one of the most inspiring CIO that I have the privilege to meet. And you know, he serves today as the CIO of SMBC Americas. It's a global top 15 financial institution based out of Tokyo. He leads technology from both strategy and execution sides. Before to joining SMBC, was uh at GPNC.
He was uh leading uh globally data management for assets and wealth management. And before his career in financial services, he served during 20 years as a naval officer. So, you can believe me. When he's talking about leadership or mitigating risk, he knows what he's talking about. Ladies and gentlemen, please welcome on stage with a big round of applause. So, I appreciate the kind remarks. I would say there's a ton of inspiring people here, and hopefully everyone will wake up after lunch.
Yeah, I had a good lunch. This is going to be exciting. I think the panel's going to be awesome. Uh what I'd like to do is first introduce myself. Um, and it looks like Oh, there we go. I can see what slides are there. So, I'm Gordo Wilson, like you said, I'm the Americas division uh, CIO for Sumitomo Mitsui Bank. Uh, we are on a journey and we'll go through that real quick in my short time. A journey of going from where we were to the future and what we want to do is engineer everything we do in a way that's going to protect the firm and Data Bricks is a key component of that journey that we're on.
Little bit about Sumitomo Mitsui Bank. You mentioned it, G-SIB, 2 trillion in assets, truly a global bank. Uh, you know, we have a global banking unit, which is EMEA, uh, the Americas and APAC ex-Japan. That's where we kind of sit. It's a huge growth engine for the firm. And one key note, you can see what else we have there, primary dealer, sales and trading. We did launch a retail digital bank from the start called Genius Bank.
You see that on the list. So, there's a lot going on, a lot of growth. I think one key point though is that the firm has its roots over 400 years ago, which is pretty amazing and you don't make it that far without understanding your risk, without looking at ways to innovate and protecting the the bank itself. So, that's a pretty amazing stat that SMBC has that not many banks I think can say that. Uh, a a little bit about me.
You men- you mentioned my career path. What I want to say really is uh, a lot of people ask, "How did you get to be a CIO?" Or, "How did you even go into banking from that track record?" I'll highlight one item on the Navy. I I'll couple items. I was honored to actually lead a squadron on deployment uh, a Navy squadron as the commanding officer. Sweet Caroline in a prior tour, just for the record, and that's why they threw that in there, was the fight song of another squadron I was in, which is kind of funny.
It was a It was kind of a joke because the CEO wanted it to be Eye of the Tiger. And everyone's like, that's not going to fly. So, we made it Sweet Caroline and it was it was kind of like a little you can always push back a little bit on things. Uh so, that's pretty funny. I'm a Mets fan, not a Red Sox fan, for the record. I would highlight the other item here is the engineering test pilot. I'd say late '90s, I went through what's called the US Naval Test Pilot School.
You actually learn how to acquire, how to integrate, how to test, how to field systems, new aircraft, new platforms with a goal to accomplish a mission, a capability. That mindset has stuck with me since then. Already before then, I was thinking about the mission. Then it became about how do you build end-to-end solutions that are going to get things done. And that's really what it's all about. Uh at the heart, it was always about data.
It was always about technology and always about innovation. I had the pleasure to work at JP Morgan. My old uh CEO I I I will never claim to be as good a key keynote as that guy. He was pretty amazing. I don't know how many people got to see him, but that was always fun. And always very direct. Um and then I learned a lot there about client-facing roles. What was about 10% of my time first was data analytics and tech became 50, 60, 70, 80 about 10 years ago.
And you realize that it's all about the data. He mentioned it in his keynote today. That data is the biggest item to tackle for any bank. And he's right. Then I came to SMBC, was the chief data officer, and then now CIO. So, it's been a great journey and we're on a growth journey. What I want to do is focus you on a couple things here. And these were like ingrained in my mind in flight school and I still think about them every day and in the context that we do.
Aviate. That's the basics. That's like keeping the airplane on altitude. It's taking off. It's landing. So, imagine yourself on a mission and it's like a high-value target somewhere in some not-so-nice country. Or you're just going from point A to point B. But let's go to the exciting one. High-value target. Kind of like reg reporting for the Fed. Same thing. So, you end up Aviate means you need to have the basics. The airplane's on speed, on you know, on altitude.
You know how to get there. You know if there's a mission system that fails, if you have an engine fire, you have to handle that. That's that you have to do this for safe uh you know, just safe conduct of the flight. Navigation though seems interesting. Navigate's not just about staying on the flight plan. Navigate means optimizing all your information and your your system, which is happens to be that aircraft, to put it in the right spot to maximize the potential capability of your onboard sensors or even offboard sensors to find that target.
You can also look at threats, risks, cyber-type attacks. Those kind of things come in in that same Navigate. This is where you actually accomplish your mission. Then the last one is Communicate. How do you tell your call it your customer, "This is the target." Or maybe it's not the target. And what's it doing? Is it ex- Does it realize that you're there even somehow surveilling it? And then what actions are going to be taken?
So, then there'll be this loop again of communication. But in the bottom line is if something goes bad at Aviate, you go back to that. That's the primary thing. Same thing applies if you think about it as a tech organization. We have to think about things the same way. What are the basics? And on this slide, I think you'll see some things that should be familiar to you. I'm sure many are in the same point. We have a lot of fragmentation of our sources.
We have silo data. We have information that's really hard to wrangle and put together. If you figured all this out, please raise your hand at your bank cuz you're in like a nirvana state and we'd all love to learn from you. And even if you're there, it's probably changing. That bar is going to change in a day. So we have to continue to wrangle with these what I would call industry challenges. The other item is the demands are increasing, whether we like it or not.
And Jamie mentioned it today about regulatory challenges, right? But that's they're they're not going to slow down. They're going to be there and hopefully we'll be better at it and we are trying to be better at it. But there's also demand from our stakeholders, our people internally as well as our customers that we can provide them the information they need to be successful in an efficient way and they not wasting their time.
I'm going to give you another airplane thing cuz I love airplanes. But and by the way, I used I used some GenAI to make this. And it just it just magically happened. On the left side is kind of a depiction of what a cockpit used to look like. And those that maybe fly any pilots out there? Anybody with commercial or no? Ah, there's one. All right. So what I would call steam gauges, literally 70% 60% of my time might have been spent spent looking at these dials, looking at oil pressure fluctuating, looking at the fuel, having to scan all of those things that kind of look like they're from the '60s cuz they probably are from the '60s.
They still are out there. The systems underneath that are still there when you go to the right side. And this is where you start looking at a digital cockpit. And what you notice on that cockpit, look at all the information in these big glass displays. I don't need to look at the engine oil pressure. I don't need to look at any of that. I'm now focused on the navigate part. The aviate part is much easier. And my mind share as a pilot, especially in that same high-value target scenario, I'm not worried as much about airspeed and altitude.
I'm still concerned. And if it goes bad, it'll pop up on the glass because we've tested it. It's not good quality. Something bad's happening. And then I would revert back to that aviate mode. But you can spend more of your time in that navigate and communicate. You can pipe it off with real technology. Instead of having to talk and give a lat- latitude and longitude, you can actually just send it and transmit it. So, that's pretty cool.
But the key to this is trust. You have to actually trust this. And so, how do you trust it? It's the way you engineer it. And the testing that's involved prior. And if you don't do that, you will not have the trust of the pilot. You won't have the trust of the team that's using that information. And worst case, you get yourself into a unsafe type situation, a big risk situation. So, what are we doing at SMBC? This is pretty cool because what I would say is we're in a transition from a legacy steam gauge cockpit to a high-end digital cockpit, even better than that picture, you know.
That was just That was an artist depiction. Uh And what you'll see is you're using that same same foundation. The aviate part what the is the key. If we don't get that basic part right, no one will trust what we're providing the rest of the way on the stack. So, we've been bringing our portfolio together, leveraging the Databricks Lakehouse. We bring in sources of record. Over 25 are in there now. We bring the product views in, the domain views.
We get a consistent way of bringing that together. And we do it in a governed way where we can see things like lineage. We can see in an automatic way quality. We have work to do. Don't get me wrong, but we have this foundation set and we're building to it. When we get to the navigate part, this is where it gets real. Does our our chief risk officer really get what he what he needs for counterparty credit risk? Does he trust the information in the data that we're providing to him?
Does our finance our CFO get the right reg reg reporting information to deliver Those that know the 2052A liquidity report, it's a really fun one. Can we provide that uh that to him in a way that he trusts it? Cuz what happens otherwise, it's like the steam gauge today. Hundreds of people reconciling the data, doing almost, you know, massive amount amounts more work than actually analyzing it and using it. So, we are now leveraging this lakehouse and the stack you see on the right to do that.
Now, how do we get trust in the data? Well, that's that medallion architecture and it's the data quality, the data lineage. Is it coming from the right place? This will prevent those silos in the future. And that bronze, silver, gold, I think this group, I know we don't need to explain that, but it's amazing. If you don't put the bronze in, you're never going to get to silver, you're never going to get to gold, and people will build their own way cuz they need to get their mission done.
So, the goal is to keep that fully together. And at the bottom, I would say, you know, the communicate part is even more critical. You know what? I'll I'll I'll take it more critical in the than the aviation example in a lot of ways. We need to tell these users how is the quality of what they're getting. And if it's not good, they need to know rather than just accept it. So, we're automating data quality rules. We're using AI, using generative AI.
We're going to talk about that later uh I'm sure in another forum, too. But, it's really exciting that you don't have to spend your time manually clunking through all of that information, but the end user then can see when it's not great. And if it's not great, you go back to aviate mode. You go back to the source. You fix it. You have to know you have to have knowledge in order to do that. So, as I I'll close just with a couple things.
One, I think we are actually there. We have the platform built. I'm looking at some people or somebody right there is helping you continue to build and leverage it. We feel like we have the right framework with Databricks at the heart to enable us to get that aviate part really clean, very well governed, and make us focus more on the navigate, the mission, getting things done not only for risk, but also for the front office, for everybody.
And the interesting thing is with our EMEA partners, our APAC partners, and the global banking unit, we're going to do this in the same model across our entire enterprise. That's not a small task where starting I'd say I'd say we're in the initial stages of that second part of that journey, and it's going to be great. I think we have a lot of work ahead of us, but we have a lot of foundation set. And the key really is what we said at the beginning, that you have to engineer it today in order to protect the firm for tomorrow.
And it's the same thing as that cockpit. If you didn't engineer it well, it's going to be unsafe. Same thing here for us. So, thanks for the time, and I know we have a a great panel coming up. Thank you. That was amazing. Thanks. Wow. It was so inspiring. I've told you, you know, the parallel between uh aviation being a a a naval officer and mitigating risk, anticipating risk is super inspiring. So, thank you very much.
To pursue this discussion, ladies and gentlemen, I have the pleasure also to welcome on stage two amazing speakers. I will welcome on stage Nitin Kulkarni and. Please give a round of applause. So, as you can see you don't see it now, but uh didn't hesitate between let's say yes, going to Paris Fashion Week or wearing the Unity catalog socks. We shall basically only for top practitioner, you know, even Jun Tanaka, who is leading global FSI and public sector, told me I have all the swag for Databricks, but I do not have these socks.
What is the secret? You got to know where to find them. So, ladies and gentlemen, so just to introduce our our speaker today, so I have the pleasure to welcome. Thank you. So, maybe if you can introduce yourself. Like you guys say, nice to see everyone. I know uh I it'll be difficult to follow, you know, fantastic keynote from the analogy of, you know, aviation industry. Uh I have a relatively smaller role to play than.
I am a CIO of data and AI at Nationwide Building Society. We are a mutual, you know, in UK as what it means essentially is that we don't have any shareholders as such. We essentially work directly with our members and share benefits directly with them. So, you are not private and not as Databricks. Uh and uh, if you can introduce yourself as well. Sure. Uh pleasure to be here today and uh thank you for the opportunity to speak to the group here.
I've been with SMBC for 24 years now uh in a variety of capacities and I've seen the journey of our organization over time kind of really evolve. And I have to say it's it's an exciting time now because to see really the evolution of data from where we were when I started with the organization to where we are now and knowing what's coming in the future, it really is transformative I think for the organization. As Gordo said, you know, we have a we have a very strong footprint footprint right now in the Americas.
We're expanding that rapidly into to EMEA and now we're starting to bring that footprint out to the global level. So it really is going to bring our organization to the next level. So we're set. Thank you very much. So I will start with my first question gentlemen is what are according to your different experience the key data challenges that all the banks today are facing when I would say building AI driven risk management models.
And how did you address all these challenges, these issues like for example data quality, integration and governance? Maybe Nithin you can start. Yeah, absolutely. I think you got it. I think uh as you rightly highlighted look from a data quality standpoint, it is one of the biggest you know key things. What happens is you know in terms of you know DQ is organizations often you know go after and say that I'm going to you know go and you know fix this entire problem or data quality has to be perfect you know and after that you know the world is going to look you know fantastic.
It never happens that way right? We started very small. We said look we are a retail organization. Let's look at for us to function what are the core critical data elements that we to look at. This could be a simple thing like your KYC variables or it could be like your PD, AD, LGD in terms of you know risk models right? So we looked at those particular elements and said that how do we start monitoring them ground up right from source to target.
And this is where we use Databricks in the middle as well with Unity Catalog you know giving us all that particular leverage. That's the first you know portion. The second part I would say is the data governance element you know which is very very fundamental. Having the right owners you know and associating you know ownership you know to data is very key. We embarked on a journey and made sure that you know there are right owners, there are right data stewards.
We create a culture of you know data governance is not an afterthought once you know you have done all your warehousing and lake houses. It's really you know starting you know at the you know at the point of initiation of the journey itself. So those are the two key you know fundamental things. In terms of risk models you know one of the key things you know that we landed was the third party models you know is somewhere we saw it that hey whenever you buy a product these days what happens is there are a lot of ML models that come and board with it.
We said that hey how do we look and make sure that the same control and rigor is applied to them as well. So we worked extensively with our procurement guys and said that hey whenever you're purchasing some of these things let us not you know use you know have have a different dimension of you know kind of you know control standard. Let's have a similar standard into those models as well to make sure that it's all consistently applied you know across the entire area.
So yeah I mean look DQ and data remains the biggest challenge in my you know even today. It's just that you know how you implement on these particular practices and culture is what makes the difference in the organization. ? And I think the other key challenge that we see that's and from speaking to a lot of peers here it's it's not unique to financial services and it's not unique to AI models is that integration aspect.
So having that core data foundation. So you know we didn't have the systems that we have today to really build that data lake architecture. We had fragmented systems, we had data silos, we had enterprise data warehouses that weren't truly enterprise level. They were still kind of fragmented by business units. You know different different regions, different parts of the organization. Now we have that technical capability to bring that data in and that's truly the foundation whether it's your traditional you know analytics workloads, reporting workloads, or your risk driven AI model workloads.
Having that centralized lake house architecture with the medallion model and that data quality layered into that that's how you can make those actionable insights. Well and what is super interesting as well is you are both operating different markets. You are operating in the UK, you are operating broadly in the US, EMEA, and of course Asia. Um we are living in a super highly regulated environment. And what is absolutely key is to be able to track the data and to be able to make the data available if you have a control coming from uh regulatory authority.
So basically, how do you ensure this transparency and explainability, especially in the AI models? Uh and I think it is important to know how you are doing that, especially for compliance model, uh for fraud fraud prevention, for example. Um I think it should be super interesting for our audience to learn a little bit around this. first. So I think the key there is uh having a mature model uh risk framework. That's really the foundation for managing that risk around it.
Um I don't think most organizations, you know, AI is is relatively new. It's ML's been around for a while, but but now the the speed at which AI is advancing, I I don't think the model risk management frameworks have necessarily kept up with that. So uh you need to look at the, you know, have a program for analyzing the fairness of the models, um looking at the at the potential bias for those models, uh the efficiency of those models, and those governance frameworks around the transparency and the and the modeling of those.
And having the appropriate programs in place to be able to assess really what that what that um you know, the both the ingestion side and the output side of those models looks like. And then do you Yeah, so I think if you if you within the UK, you know, regulation, uh we have the SS123, you know, a principle, which essentially gives a very nice, you know, model, you know, risk management, you know, framework. So what we did was, you know, we used some of those basics and said that, "Okay, what do we need to build around it?"
Right? Now, if you look at, you know, purely from a model tearing standpoint, it's very fundamental because if you use and go by a very traditional way of looking at a AI model, every AI model will go at come as a highest tier because it's external at the end of the day. So, we said that, "Okay, we can't work like this way." So, we looked at the materiality and complexity of every model and made a tearing system within organization and said, "Okay, there's a tier 1 2 3 4, you know, system that we're going to create."
For every tier, what we then said is that what kind of controls you need to apply which align, you know, to these particular principles, transparency, explainability, all all of all flavors of all these particular principles as well. So, that's the first exercise that we did. The second thing that we did was then whenever an AI model comes, we said that the first, you know, thing you do is, you know, tier that particular model.
So, everybody exactly knows what principles we need to follow to make sure that the control standard is essentially, you know, signed off as a part of, you know, that particular process, right? That's the core ethics, you know, that was done as a part of this. And that really helps you because the moment when you have the tearing, you know, the kind of controls you need to imply and you know, you know the kind of compliance you need to essentially, you know, get to.
And that makes the process a much more seamless. But as Storm said, it's a very early days. Uh regulatory, you know, landscape is changing as well rapidly. So, we need to keep up with that as well. So, that's the third, you know, party thing, you know, that we did is we do something called as a horizon scanning in this particular, you know, world as well. And we try to see that, you know, that team is dedicatedly focused on ensuring what are the regulations coming up front and we we are trying to at least prepare ourselves in for that particular regulations if not, you know, fully ready.
So, that's the way, you know, we are approaching this particular space. Thank you very much. You know, it's interesting. One of the questions I like to ask our our model risk management people is what how are they assessing those models, right? I I don't think I how many model risk management people do we have in the room? Anyone? One? No. We have one? Yeah. All right. Keep keep me honest here if I say anything wrong.
There's a There's a couple of different frameworks for kind of analyzing AI models. There's there's shaft and there's lime. So so so so shaft is an interesting one. It's actually based in uh game theory. So the idea was it was a it was a model used to uh you had collaborative games. You had a bunch of players and the players would go on a quest. And if they were successful in their quest, they would receive currency within the game.
So the idea there is that there is there's a certain number of input elements and each one of those input elements each player would would contribute more or less depending on that outcome. And so they were assigned a value and then based on the value of that input to the game, they would receive a larger or smaller portion of the payout. And so that model framework has now been applied to analyzing AI for efficiency where you have you you can feed stuff into a model.
And what we do is we feed it a a predetermined set of inputs that we know the value of. And we know what the expected outcome of that model should be. And now through that model monitoring in the those those quantitative metrics, we can feed those inputs in. We know what the outcome should be and we can compare the actual outcome of that model to what our expected outcome is and we can use that to determine if the model is valid, determine if we need fine-tuning in the model, or kind of create that continuous feedback loop.
And I think that's one of the key ways we address that kind of model risk is is having that one it's the human in the loop. We heard that this morning of doing that analysis. And and on the back side of that is having that continual feedback loop to allow us to refine those models cuz it's not that they're very much a black box, but we need to be able to explain to our regulators what what are we putting in and what are we deriving out of that.
Yeah, the model has to say what it sees on the bin. That's the whole point. Mhm. Yeah, that's for sure. We we are basically gentlemen living in in a world where basically everyone should be accountable for the risk. Every department. And this is super important because I think it's not the case everywhere from what we can hear. So, I would be super interested in having your perspective, guys, on this very specific topic.
So, what kind of strategy did you use in order to break the silos? Not only the technical silos, but also the silos to provide some tools for even the business, even the middle office, to take into account that risk is really part of the overall strategy. It's absolutely key for all the institutions. I think it should be super interesting to have your vision on that maybe. Yeah, absolutely. I think look even before the risk, right?
The first baseline thing, you know, that at least we did was data. It is absolutely fundamental. So, we did a we did a top-down teaching, you know, right at the board level and walked them into this particular journey to say, "Hey, look, you know, here's an investment we all have to do together." I don't think so, you know, we are in a world today where you no longer have to convince anyone that, you know, data is the lifeblood.
I think it is understood. You just need to take, you know, the entire organization with that journey. This cannot be just a a technology initiative, you know, that you can build all these cool platforms where there's no end user on using it. That's the first thing, right? If you look at the second, you know, part of that particular journey, what we did was purely in terms of, you know, usage of risk and AI. We set up a construct called as AI Council within our organization.
The idea behind that was, you know, getting all of the core entities within the organization, be it be security, be it be a procurement teams, you know, be it be your legal advisers, all of them in, you know, one particular group. And we explained them each and every use case that you're trying to actually build and make value out of it. That helped us a lot. And one of the core principles, for example, for GenAI when we started the journey, we came out and said that, "Hey, everything we are going to do is going to be human in the loop."
You know, to begin with, we firmly believe that GenAI should be used as a co-pilot because that's where the maturity of GenAI was at that particular point in time. So, having those constructs, educational, you know, things, you know, really helps right up front. And that has essentially helped us, you know, we are one of the very few institutions who actually have landed GenAI use cases in production. And we continue to do so.
That's my That's the way we have approached it at this point. Yeah. , maybe do you have a perspective as well? Sure. So, I I I think for us the first thing was understanding is that I I feel like within IT, data was always owned by tech. Right? It's It's the data. We have the databases. We have the systems. It's not. It's It's the business's data. And it's not Even within the business, it's not It's my not my data. It's the organization's data.
So, the first part of breaking down that silo was making everybody understand that the data that is created That's the That's an asset. That is value to the business. And by being able to take that and I and I feel like data democratization is a little bit of an overplayed term, but but it rings true. It's It's bringing all of that data in because it is the organization's data. And so, once you have that footprint of the data, now you can put your models on top of it.
You can start to see graph relationships and and other insights that when you're looking at it with those you know, with those siloed kind of business construct lenses, you didn't have that capability. Then we put our governance on top of that. And of course, there's the Unity Catalog. Um Unity Catalog is is our is our governance control, right? That gives us our our ability to do our quality controls, our our authorization and authentication, um our our you know, our catalog of the data.
It is just that. It is that unity of all of our aspects of whether it's operational data, business data, market data, um master data, we're able to create that and then manage that in the Unity Catalog construct. Well, that's And maybe to finish, I think I have a last question. Question regarding real-time and streaming. I would be super interesting to know how you are leveraging real-time, especially real-time with some very specific insights, before they impact their your real business operations.
Time of anticipating, preventing, and to try to trigger some actions before something risky occurs. Maybe maybe? So so I think this is where the value of the AI models really start to come to play. You you have with real-time data now, you have so much data streaming in. It's not something that in the traditional, you know, kind of kind of batch updates and stuff, you you had a a defined set of data and a and a reasonable amount of resources that could comb through that data and make actionable insights.
Now the volume of all that signals intelligence, all of that data coming to the platform, you know, whether it's again, market data, operational data, there's there's such a great volume of data. That's where the AI models start to shine because now I can use that to filter the noise. I can look for those anomalies. I can I can help, you know, see market volatility. I can help to, you know, see potential credit risk or or liquidity risk.
And I can highlight those anomalies and now I can have that same group of people What's his name? receive actionable insights. And now they can make that human in the loop decision to inform the business and make business decisions based on that filtered set of the much larger real-time streaming data that's coming into the platform. Anything? Yeah, I think I think it's also if you look at different perspective and say it's not just impact on business or let's look at the impact on society.
Traditional you know models imagine that you know someone is applying for a mortgage and you're declining them because you don't have latest data about that customer. It's it's very unfair. We are looking at you know credit risk modeling in that particular context to say you know how can we help you know people? Of course within the risk appetite that an organization has to carry. But you know use more of these particular streaming data inputs to say hey, can we make sure that you know people who are eligible should actually be eligible for these particular things, right?
Otherwise we're not doing our function as a as a banking institution. So credit scoring is where we really see you know traditional models always have relied upon you know batch data as you know as said. We really feel that having industry data, you know, more real time you know, data from our bureaus will essentially change the game in that particular space. That's where the focus is today. The other element also if you look at you know, scams, you know, which is a big problem in in you know, not just the UK but I think across the globe.
We are looking at our transaction you know, kind of you know, monitoring you know, solutions as well to look at how can streaming data essentially in moment essentially capture whether a transaction is fraud or not. So, that's that's another area you know, that we're looking at. These are That's the reason I say these are more society impacting you know, you know, features as well. We are not really looking at how it benefits us, how it benefits you know, back to the society as well.
So, those are the two key areas where we're looking at you know, how we can use you know, real time information at this moment in time. Thank you very much gentlemen. It was great mitigating hearing you just moderating this panel with you. It was super inspiring. Thank you for sharing your perspective. Thank you again. Thank you. Thank you. Pleasure sir. It was a great pleasure. So, Before before we leave, just a big big thank you to our sponsor Deloitte and AWS because without those we would have been not possible for us to have this great session today.
So, big thank you to you ladies and gentlemen and here are some recommended session that you can go. So, please you can see that this is just the industry forum. There are a lot of very interesting things. Yes, use your mobile. You can take it to the pictures and the last but not the least is you can download here our booklet on how data intelligence can transform financial services around risk management, driving new revenue and being more efficient.
My last slide is if you can complete your surveys just to help us to basically you can see that in your apps. Improve ourselves, get your feedback and of course it's a great pleasure to have you here. Thank you for your trust. Wish you a super pleasant day today and tomorrow and welcome again in San Francisco. Thank you everyone.

Learn more about the Databricks Data and AI platform.