Nasdaq's AI Secret: How They're Dominating the Trade Lifecycle!

Today is going to be a general casual conversation about Nasdaq's intent and strategy to move at scale with AWS to deliver Agentech and other services across our portfolio of products at Nasdaq and in particular our fintech division. So, we'll share a very first principled approach at how we're planning to do that, what we've already achieved, but I should frame that that we haven't been sitting idly by. We've been working with Agentech AI for many years and before that and decades before that working to ensure we have a great data platform with great semantic layers and data models. All of that has prepared us for what we're building now and extending across all of our products, which is an Agentech AI framework.

A little bit about myself first. My name is Mark Murphy. I have two roles at Nasdaq and both have been influenced by the rise of high performance compute and Agentech AI. Uh in a role that we won't touch on today, it's connectivity and co-location. So, for Nasdaq, we have the Nasdaq exchanges, the market services division, where we're providing, you know, the Nasdaq stock market and a host of other exchanges around the planet. Uh in that clients co-locate around our trading systems. They're looking to be proximate to the signals that come off a trading environment. And it's been really interesting to see that the rise of compute and even Agentech AI has now crept into that edge compute space. What we'll focus on today is my other role, which is to oversee the product strategy of the intelligence layer at Nasdaq. And that's a journey, as I mentioned

earlier, that we've been on for a number of years. Starting with data Starting with data and then later moving into a genetic AI. Nasdaq has a rich history with firms like AWS. We're a listing venue where AWS and others can raise capital. And what that affords us is the ability to have deep partner relationships with them. And AWS is certainly no exception to that. And we've worked with AWS for more than 20 years now. And many of you would know that we've installed AWS Outposts to drive the largest markets we have in the US, which is our US options exchanges. And we continue to roll that out across all of our venues globally. And then I mentioned earlier and where we'll focus today, financial technology. Nasdaq over the past 5 to 10 years has gradually expanded through acquisition and organic growth, delivering critical market infrastructure services and software to a wide variety of clients across the globe. So, think of banks, regulatory bodies, exchanges, clearing houses, etc.

To focus in on financial technology, I just want to take you through some of our products, our core products. Verafin, which is a anti-financial crime software. A really interesting piece of software that is installed in thousands of banks today and is part of the back-end process that protects clients and banks themselves from fraud. Even going as far to being able to detect human trafficking. Axiom XL, which is a regulatory reporting solution. Very data rich and something that always has to be absolutely correct. There isn't room for error. And I highlight that here, which will be a reoccurring theme, that the data we deal with, the outputs that we're responsible for, are absolutely critical that they're accurate. There isn't room for interpretation or inaccuracies. Surveillance. Surveilling markets across the globe to ensure that things like insider trading are not occurring. Again, another heavy data workload, something that needs to be absolutely correct, otherwise there's ramifications in the hundreds of millions of dollars. And then in our capital markets technology, we have Calypso, which is driving the back end uh collateral workflows, treasury, etc. for many banks across the globe. Central banks included. And then we have a clip, which is our branding for our technology division that sells trading technology, clearing technology, and CSD technology, again, across the world. Many of the largest exchanges on the planet are driving uh themselves with Nasdaq technology. So, when you look at that in in one cross-section, you you do ultimately have a bunch of core applications that are absolutely critical for what they do, and they all have their own history.

And what we're doing with the intelligence layer that I mentioned earlier is to have a way to unify that and expose the data that's underneath in the core applications in an intelligent way. Now, we were on that journey ahead of Agented AI being repopularized or popularized, but now with Agented AI at the forefront and where it is, we're perfectly positioned to take advantage of it. And I'll emphasize as well that we're not gated by the fact that we're rolling out the intelligence layer to have AI in these products. Today, we already serve clients, around 4-500 clients, with AI workflows or AI solutions across these products. The intelligence layer though is something we're building, as I said earlier, to move at scale, to expand and to provide more optionality for our clients, for our partners, and also for ourselves to deliver Agentyc workers.

These are obvious and you've heard them before, before, but they're incredibly simple and they're not particularly, you know, mind-blowing, right? They're not maybe even not interesting, but they're fundamental requirements to move at scale and to deliver proper, enduring, sustainable Agentyc workflows. Pilots are great. We value them. We invest in them. We do them. But you have to accept that a pilot is a long way from moving to production at scale. And that's something that when we're building this framework, we keep front of mind. We operate in a complex environment, not just one, but many. That's not a challenge, it's not a headwind, it's a condition. Something we live with and work with. And we lean into. We're good at it. For Agentyc AI, that's just the same, but, you know, a continuation. And then I mentioned scale a number of times already. times already. People want innovation, whether that be internally, I'm sure many of you are under pressure in your own organizations to deliver, quote, AI. Get me AI, right? But how do you do that in a sustainable way? And again, that is a focus here. This is a long, long-running objective of ours, right? It's not about being successful in the very, very near term or or having something to demonstrate. We do all of that with our proof of concepts and our our pilots, but how do you bring that into production long term?

So, I mentioned that we have fragmented data data um across our products, even sometimes within our products. This is something that we've invested in for a long time and we continue to do. And we're pushing that up to the intelligence layer. We're also investing heavily in data lineage and enterprise level governance. So, imagine a scenario where a regulator walks in and says, I can see that you made a collateral call uh using your Calypso application and that part of that was an agentic workflow or an agentic worker. Explain to me what happened. And they're not talking to the agent. I mean, maybe in the future they will, but at this point you would imagine they're talking to a human and the human will need to go and and pull that data and be able to explain what happened. So, explainability of these workflows is absolutely critical. And I think that's obvious to everybody, but the steps and architecture you need to put into place is something you need to plan ahead of time. It's not an afterthought. From an AI enablement point of view, scalable deployment, I've mentioned it several times. It's just so important, though. Clients need to be able to trust what you're putting into place and know that it's production ready. And ensuring that models are trained on full amounts of data, wholesome data, not partial data. And that's why the intelligence layer has the ability to ingest other data sources. And in doing so, it can cover not just Nasdaq products that are relevant to a workflow, but it can ingest the other data and have that as context as it learns.

One of the things and and how we think about AI at Nasdaq generally and our clients have the same view is that there's two sides. There's an opportunity and there's a there's a risk. On the opportunity side, you can be more efficient, you can develop new products, you can do different things in a more productive manner. Which is great and I think everybody gets that. With this shift and change, I personally wouldn't believe that there's an option to not participate. So really there's an opportunity cost of not participating, of not acting, not responding. So when we talk to clients and what we're preparing for to help clients manage is it might not be that you want to be fully on the offense, but doing nothing is going to expose you to being defensive or being disrupted. So it's important that we make moves with our clients and for our clients to prepare for that.

A lot of what we've done in the past at Nasdaq and continue to do is position for the future. And sometimes you don't know exactly what you need to do or what the outcomes will be. And it's for that reason that when we think about positioning for the future, we're doing that for ourselves, but more importantly we're doing that for the client. What will the client want in the future? And you won't always get the exact right answer, but you can usually surmise that there's going to be a situation where the multiple things are likely to be true. And that's where we try to bring our product strategy and our road map to arrive at a place where we have some optionality, we've built foundational architecture on top of AWS, and that readies us to then respond to whatever's shifting in the market and where demand normally moves.

So Nasdaq intelligence or the intelligence layer. At the basis of this, um, we're delivering delivering again basic things, but threaded together in in a proper way, in a harmonious way, is valuable and important. So data collection and management. Mapping and transformation. It sounds simple, but many applications, particularly in fintech, are sometimes missing data models or semantic layers, or people have created their own that aren't consistent with others, maybe even in the same organization. So, we can bring that to our clients, and we can bring that to the community of clients that we have. Validation and reconciliation. Reporting and analytics. Some might argue that's, you know, is that interesting? that interesting? I can tell you who it is interesting for is people who work in that space every day, and that have the burden of having to produce reports that have meaningful outcomes if they're not correct. It's a a well-used example, but through the COVID period when the Nasdaq exchanges were operating at four times volume. We in the Nordics had not yet implemented the Nasdaq intelligence platform, the Nasdaq intelligence layer, and therefore we had on-premise servers running overnight, but they weren't completing their jobs as they usually would by the time the market opened the next day. It's a basic example, but it created a huge challenge. At that time it was solved with Dell servers being backed up into a data center and loaded off and plugged it, right? And that was not easy to do. Now, that same premise is true for Calypso or Axiom or any of the applications that I mentioned earlier, it would be to just a different scenario. So, the intelligence layer is is cloud-native and is able to support scaling up and down. And then distribution and billing. Many of our clients struggle today to bill and provide statements, and I think there was a session that might actually be on right now that's talking to that exactly. So being able to provide a modern open platform that allows you to bill for certain products or to calculate billing. It's often a a constraint that our clients have to create new products and services and to monetize them. It's that plumbing that they're missing. So again you've got this modern compute and intelligence layer that enables a variety of different things but most of all a variety of different options for our clients.

I mentioned it earlier but I'll mention it again. it again. The semantic layer and the data models really are at the heart of this. Without context it's impossible to do anything with this platform or the data that resides underneath of it. Too often we've seen that clients struggle to interact with a data set because context is missing. And if they do eventually get there, it's a one-time effort with a subset of data that leaves them in a position of of status quo which is suboptimal for a long period of time.

I won't drain these but it unlocks a lot. lot. Unified data management from a product point of view, a core product point of view but also imagine across all products. It's not uncommon for Nasdaq to be selling our core products to a client and they take multiple different products. The unified intelligence layer threads that together. It also provides an amazing platform or foundation for agentic enablement. So whether it be partners of Nasdaq creating digital workers to to go on top of this platform, whether it be the client themselves or whether it be a Nasdaq agent that sits on top. We're able to provide the interfaces, MCP servers, APIs, the contextual layer that we talked about earlier. All of that providing client choice. Some clients want to build their own. Some clients have other workflows that they want to plug into this. Others want a partner to help them. Others might prefer to procure from Nasdaq. Those choices Those choices we lean into. We want our clients to have choice. And governance and controls, so important. Absolutely critical. Do I trust what is in my production system? And that is something we expect to earn over time and we've already earned with many clients. I mentioned earlier that we have around 500 clients using Nasdaq Agility workers or agents today. So, it'll start with a human in the loop and it will progress from there. But that's okay because we have the framework that will support all of that. So, another visual and sort of demonstration of what this is. It's quite literally an intelligence layer that we've extended across all of our products. And that will now build deeper and deeper capabilities into it. Agility enablement being one of those. But you also have the ability to have cross product insights across the Nasdaq products. You have the ability to have collective intelligence at larger scale. Today in our surveillance product and our Verafin product, we already have collective intelligence. So, we're pushing data back to clients to tell them where they are in their peer group for certain things. But we have the ability to extend that. And then I mentioned earlier that we can bring data into this platform that isn't native to Nasdaq applications. Again, I'll underline how important that

is with an example. with an example. You can imagine an energetic workflow that is helping someone to make a collateral call from the Calypso application. Wonderful. It goes out. It pushes something into the world, right? That that it thinks is great. But you need the Swift messaging and the Swift data of what happened in the real world to come back in to understand if there was value in what just happened. To build that trust and then maintain that trust. To train the model, to train the agent. So, we're very keyed in to understanding that we're not everything. And that we need to be flexible. So, we we we've built um extensive capabilities with security controls, etc. to bring data into the platform. And then I mentioned earlier the the contextual elements of semantic layer and data models, also being able to extend them beyond our applications across that bring your own data.

So, I mentioned earlier the journey we've been on and it really does go back 20 years. I mean, like many of you, we started with folks at Nasdaq that were curious about the cloud 20 years ago using their own credit cards to gain access and experiment. And that resulted in our first data platform with AWS. Something we still have today and actually evolved into this. But we're well on the way through phase one of developing that app scale across Nasdaq to now phase two, which is extending the modern version of that across all of our products within fintech. And then we're really working through phase three right now as well, which is expanding the breadth of that platform. So, I mentioned earlier we don't know the future. We don't pretend to know everything, but we do generally know where we need to be for our clients. And the architectural choices that we make are always rooted in ensuring we have some level of optionality with that. And AWS is a great partner in that, both with the R&D that they put into their core product, but also following the trends and and being ready for the next big thing. One example, a small example, is it's you know, probably once a week that someone would engage myself talking about quantum computing or quantum networking and what we need to do to prepare for that. Now, I don't think any of us at Nasdaq quite know where that's headed or what will be impacted, but we're already thinking about are there one-way doors that we're going down that we might regret regret and are there two-way doors that we should be moving down?

So, really in summary, and I hope this came across clearly, came across clearly, we're looking to unify all of our financial technology products across one data layer. That'll create a strategic fit both for us, but also for our clients, giving them optionality and capability. And we think now is the right time to extend our product to the agentic enablement layer that I've mentioned. Data quality, the semantic layer, data models, data being organized and cataloged, those are fundamental must-haves and it's not easy to do. It's not perfect. But if you don't have that, you you just can't go to the next step at scale. So, we recognize that and we continue to invest in that. And the road ahead is interesting, right? I I'm terribly excited about collective intelligence as an example. You know, what can we do in a world where you submit data, it's anonymized, and then you get back peer information? You know, am I in the top 10% of settlement breaks for a certain country or a certain market segment? Am I in the bottom 10%? Then I probably need to do something about it. What one of my workflows is breaking and why? And where does it break for other people and not not me or for me? And then the client and partner enablement um is also exciting. Opening up our platform, allowing people to connect and innovate, deploy capital to create value for our clients, but using our infrastructure and guardrails to do so.