Financial Services Leaders Scale Production AI with Databricks

Summary

• Financial services leaders from Morgan Stanley, JPMorgan Chase, Mastercard, and RBC Capital Markets share how they scaled production AI revenue by more than 10x while managing regulatory complexity and data governance across multiple jurisdictions.
• The framework driving production AI success in financial services centers on three pillars: choice in models and cloud, context from proprietary data, and control through comprehensive governance using Databricks lakehouse architecture and Unity Catalog.
• Live demonstrations cover research workflows, financial modeling automation, and intelligent agents, including Mastercard's Virtual CFO Suite for small businesses and RBC's Aiden platform, all powered by Databricks.

Financial Services Leaders Scale Production AI with Databricks

The financial services industry is shifting from AI pilots to production deployments that deliver measurable returns. In this forum, leaders from Morgan Stanley, JPMorgan Chase, Mastercard, and RBC Capital Markets share how they scaled production AI revenue by more than 10x while managing regulatory complexity, data governance, and risk controls. Financial intelligence requires unified, secure, and governed data to safely move capital and satisfy regulators across multiple jurisdictions.
Discover how enterprise leaders unify data using Databricks lakehouse architecture and Unity Catalog governance, deploy agentic AI systems with risk and compliance controls, and balance rapid innovation with safety. Learn the framework driving production AI success: choice in models and cloud, context from proprietary data, and control through comprehensive governance. See live demonstrations of research workflows, financial modeling automation, and intelligent agents powered by Databricks.
🤝

Chapters

• 00:00 Financial Services Forum Opening
• 01:33 Erin Butler: Massive Industry Shift to Production AI
• 05:37 Three Key Questions: Choice, Context, and Control
• 08:53 Morgan Stanley: Enterprise AI Adoption at Scale
• 14:17 Global Regulatory Complexity and Data Governance
• 20:32 Democratizing Data Access with Natural Language
• 31:41 JP Morgan Chase: Data as Competitive Moat
• 39:01 Governance and Building Trust in Agentic Systems
• 47:35 Cost Management for Production AI Agents
• 53:08 Mastercard: Intelligence from Global Commerce
• 56:15 Large Table Models for Transaction Intelligence
• 59:13 Virtual CFO Suite for Small Business
• 01:06:53 RBC's Aiden Platform and AI Culture
• 01:09:31 Live Demo: AI-Powered Research Workflow
• 01:34:04 Financial Model Automation in Excel

FAQs

How did financial services companies scale production AI using Databricks?

Morgan Stanley, JPMorgan Chase, Mastercard, and RBC Capital Markets scaled production AI revenue by more than 10x by unifying data using Databricks lakehouse architecture and Unity Catalog governance. The key framework relies on three pillars: choice in models and cloud, context from proprietary data, and control through comprehensive governance that satisfies regulators across multiple jurisdictions.

What role does data governance play in financial services AI deployments?

Data governance is central to financial services AI adoption because organizations must satisfy regulators across multiple jurisdictions while deploying agentic AI systems with risk and compliance controls. This video explains how Databricks Unity Catalog provides the unified governance layer that enables rapid AI innovation without compromising regulatory requirements.

What did Mastercard build for small businesses using Databricks?

Mastercard developed a Virtual CFO Suite for small businesses, drawing on its intelligence from global commerce data and large table models for transaction intelligence. This solution is demonstrated in this video as an example of how financial services firms can translate enterprise AI capabilities into products that serve smaller customers.

What is RBC's Aiden platform?

RBC Capital Markets' Aiden platform is an AI-powered trading and analytics system presented in this video as an example of production AI adoption at scale within a major financial institution. RBC leaders also discuss how they built an AI culture alongside the technical platform to drive enterprise-wide adoption of Databricks-powered tools.

Full transcript

[00:08] Hi everybody. Welcome. It's so good to see everyone's faces here. I've seen you all in the halls this year and we're so excited that you're here. This year is the largest financial services forum in all of data summit history. So please give
[00:24] yourselves a round of applause. Thank you. My name is Kim Hatton. I'm the global head of financial services here at Data Bricks. And we have a amazing show for you today. So, please make sure you stay until the end. Um,
[00:39] I've got some general slides here for your enjoyment. Um, and we also have a survey app. So, if you have time, please pull up the event app, leave your feedback. We are always iterating and trying to improve our content and speakers and all the things um that make
[00:54] these events so great. So, please take a moment as you can to fill out the survey. And so, without further ado, I wanted to welcome Erin Butler, who is the VP of financial services here at Data Bricks to the stage. She's going to come in and she's going to tell you all
[01:10] about kind of things that are happening in the space, where to look, you heard a lot of announcements, what's the direction, what's the way forward, and so please give my colleague a rousing round of applause, Aaron Butler.
[01:33] All right, packed house. Good afternoon everybody to the financial services forum. I love to see this room completely jam-packed. So we have a packed 90inute session for you guys and we're really going to try to bypass all the highle hype and get to the real juice so you can hear from real
[01:48] customers and what they're doing today. But before I bring our guests on stage, I want to step back and talk about the massive shift that we're watching play out in real time in our industry. So, let me take you back to June 2022. So, at that time, Data Bricks had less than
[02:04] 4,000 employees. We had 7,000 customers or so, and we hadn't even crossed a billion dollars in revenue. So, I was sitting out in the keynote. I had just started at data bricks about four to six weeks before and candidly I didn't know a ton about what we did. I knew that we
[02:21] were led by some of the smartest people in the world and that we were changing the future for data and AI. So as I'm watching uh mate gets on stage and this is the live keynote where he starts coding in real time in front of thousands of people and he was taking
[02:37] Delta Lake and open sourcing it on stage. Now, I'll admit at the time I thought, did I join the wrong company because I really didn't understand a lot about it, but that was the first and the last time that I had that feeling. But as a business leader sitting in the
[02:53] audience, I thought, how is this a product that can help me in my everyday? And I thought about you, our customers. How can this technology help transform what you do every day? And here we are. We've heard a lot about Genie over the last couple of days, but fast forward
[03:09] four years and I can't think of a better place to be. Today, we have more than 10,000 customers serving 20 uh 10,000 employees serving 20,000 customers and 75% of Fortune 500 companies run on data bricks today. Back then in 22, it felt
[03:27] like data and AI was still trapped in engineering lanes. And today I run my entire financial services business on data bricks. I work with natural language on my phone in the morning before Ranga Brisco our CRO calls me and asks me what's going on. I know in real
[03:44] time I'm trying to understand the trends of the business anomalies and it's giving me insights in real time that honestly I never thought was possible before. It's my forecasting assistant and it's giving me these faster insights in real time that matter.
[04:00] So, it's not just about how I run my business, it's how you're running your business. And the business is large and growing thanks to the people in this room. We're currently partnering with more than 2,000 financial services customers globally. And the market leaders on this screen. They're not just
[04:16] building pilots or never-ending uh purgatory of pilots, but they're utilizing data bricks to transform their data and get compounding competitive advantage. And the proof is that the logo is here. So in the midst of all this there's this
[04:32] structural change that we're seeing. So historically 95% of projects were intended to go to production. We had IT projects targeted each year and there were silos between the business leaders and the technology leaders. U traditional IT era has been completely
[04:49] transformed in this AI era. In the MIT said last year that 95% of AI projects do not make it into production. Now I know even in that time we've already started to feel that number come down but some of that's intentional exploration but it does mean that the
[05:05] structures of our organizations need to change and we need to think differently. We talk about having two lanes a fast lane and a slow lane. a a slow lane for this intentional testing iterative knowing that many of those projects are not going to make it to production and
[05:21] the fast lane of how we safely build successful projects that we know we can build to production and scale globally. What's interesting about this stat is that our data bricks customers are bucking this trend. So their production the production AI revenue since this MIT
[05:37] study for our data bricks customers has more than 10xed. So that begs the question, how how are these leaders unlocking the fast lane while everyone else still feels a little bit in that pilot purgatory? Well, these transformational leaders are asking three key questions. The first is about
[05:54] choice. How do we stay agile with no vendor lock in and have flexibility across model and cloud? The second is around context. How do we safely feed our important proprietary data into this to be able to see actual insights on our
[06:12] business? And the third question is around control. How do we scale safely but ensure that we can have that we can satisfy risk compliance and cost concerns? So let's apply these questions to this how this impact compound compounds ROI
[06:29] in AI. So, first we spend a lot of time talking about the model. Raise your hand if you feel like you spend a lot of time talking about the model. All right. I think I think there's more than I saw there. Um, but the truth is that's just
[06:45] the starting gates before you're able to do anything really meaningful. The space is changing so fast and you really can't build a foundational structure on the model alone because it really gives you the least sustainable competitive advantage of anything you're going to
[07:00] work with. So what do we need from there? We need institutional knowledge and context. That's where the special things start happening where you introduce your oil, your gold, your data, and the semantics of your business so they can understand and make sense of what's under the hood. But this is also
[07:16] where we see a lot of customers struggle and things don't get off the ground. An unstructured model with your data is not going to take you to the the production of view that we need to get to and that's where governance comes in and the control and it's not just about
[07:31] controlling the system but it's also how do we control costs and give the flexibility to scale and so this curve is where we focus. How do we innovate and deliver products so that we can help you bend this curve? And we can help you with this the vast amount of uncertainty
[07:48] that we see. We want to help you get out of uh get to success instead of that pilot purgatory. And it's all about how you focus on an open pathway that's going to give you the confidence you can scale into production and and worldwide. And this is the framework that the
[08:04] successful customers that we are seeing are executing to drive product innovation. So now this brings us to our agenda today. We have an absolute star-studded crew to coming up stage from Morgan Stanley to JP Morgan Chase Mastercard
[08:19] and followed by RBC Capital Markets with a presentation and demo. Many of these executives rarely speak in forums like this and we they have graciously offered to give us an exclusive look into how they're solving choice context um choice
[08:35] context and as we go forward. So with that, I want to thank everyone for joining us and let's get to the real show.
[08:53] We'd like to welcome to the stage Michael Pezy, global head of technology and operations, Morgan Stanley, and Dave Kiy, Chief Financial Officer at Data Bricks.
[09:43] All right. Um, sorry. Over there. Can't see you. By the way, is this organized appropriately? We've got Morgan Stanley over here. JP Morgan over here. Goldman over here. No. Like a wedding.
[09:59] You guys might think that we were sitting backstage waiting for the set to get done, but actually we were just really enjoying the rush walk up music and a a pretty ripping baseline. Something I learned about you and I'm going to read your bio here in a second. Bass player. Bass player. Yeah. Since third since
[10:14] third grade. Third grade. All right. How many axes? I have three right now. Really? Yeah. And what's the go-to? The go-to is a vintage Fender from the 70s. Early 70s. Really? A pbase. Yeah, a Pbase. That does everybody who knows what a Pbase is.
[10:31] All right. Yeah. All right. Precision base, right? Precision base. Yeah. Speaking of precision, let's talk about your bio. Sure. By the way, you guys clapped. I'm Dave Kiy. This is Michael Peasy. Okay. Just so we don't mix up.
[10:46] Okay. So, you have a pretty rare and unique background. Y uh both digital brokerage and global financial infrastructure. You currently head uh global technology and operations of Morgan Stanley. That's right. Okay. That's platforms,
[11:03] infrastructure, and operations backbone, stuff that we like to to do at Data Bricks. Um you're a member of both the operating committee and the management committee of the firm. Prior to Morgan Stanley, you were CEO of E Trade. You ended up at Morgan Stanley when
[11:19] when they acquired Erade. But interesting, I just found this out backstage. You were CFO before you were CEO. So which one? What do you like better? CEO.
[11:37] That's a good answer. I tell Ali that the loneliest job in the company is CEO, but the CFO is the least popular. Yeah, that's true. It's true. All right. So you you were leading E Trade through major retail investing transformation and then the acquisition happened. So thanks for
[11:54] being here. Yeah, of course. Great to be here. So let's let's talk in general terms like from a from a macro perspective like what are you focused on now and like what's changing the bank? Yeah. So you know just in terms of the focus coming from kind of the middle of last year into this year is really been
[12:12] scaling AI adoption at the firm doing that in a rigorous way in a well-ontrolled environment where we don't introduce new risks. We come out more effective than where we began. Get the efficiency but do it you know in a
[12:27] well-guarded secure way across the firm. That becomes the priority. We started you know middle of last year with some agentic experiments where we took sort of highly curated well ststructured data applied it to the latest reasoning models. We've become very good at eval
[12:44] become very good at at using AI to check itself and we were able to get real results and real end to-end workflow automation. Those experiments early on in PC's have now been scaled up across the firm. Um and every division everywhere in the firm has projects and
[13:00] we we institutionalized this by putting sort of transformation leadership in every in every area of the firm. prioritize projects, what we're delivering, and where we think it can make a difference. And um that's really been the main focus. Now, in the last
[13:16] couple months, that focus is has pivoted a little bit. Um as we know, with kind of the advent of these models and cyber capabilities, um we've been sort of thinking about how we adjust to a world where threat actors can now move at machine speed. Right.
[13:31] Right. And that has been a that has been a big change in sort of in our prioritization and set of what we're doing. Um but both of those are now really you know sort of the offense side and the defense side as we would say are now really the key priorities. And what
[13:46] that's what that's really led is you know really you know as we kind of look deeper at AI it becomes you know an imperative to modernize. it becomes an imperative to modernize your data structures, to modernize your platforms, um, and to really stay on that journey
[14:02] so that you can take full advantage of it from both sides of that of that point. Well, it it makes me I want to ask about Lakewatch in SIM because as we've described on stage here, it's and at RSA, it's not just the bad actors that
[14:17] are historic, but now it's the agents and therefore the volume. But let's let's table that for a second. You mentioned like it's highly regulated industry. Yes. So you mentioned many PC's projects. You got leaders across the firm. What's the
[14:33] step to get from testing ground to real operations? Yeah, I think you need to you need to build things in a rigorous way. You need to display your results and tests. You need to show the full capability. you need to run things in production
[14:49] parallel for extended periods of time to show that the agents can do the work and in many cases it's going to end up better, right? And you can see the results of that. Um that's really how and then really keeping that body of results, that body of data um to to show
[15:06] any regulatory authority um in terms of of terms of what what we're doing and why. You know, the rules and regulations are not really they're not really written yet. you know, we don't we don't have a a guide book like an FFIC book to say this is how we implement AI from a
[15:21] regulatory compliant way. We have to think in the mind of a regulator from a sort of a compliance and regulation type function and build accordingly so that we have a we have that standard ourselves and can display that to them upon review. So given the the demographic of the
[15:38] audience, I think everybody understands the complexity that you're dealing with. Uh but how does it differ by region in in particular like regulatory environment globally? Yeah. I mean if you think about I mean let's just talk a little bit about about data right in the US you know there's
[15:55] probably six regulators that can come in at any any point announce a data exam wanting to look at quality look how our standards how we report um and all of them depending on the business have the ability to place some sort of enforcement order. That's just the US. Now, let's look at globally, right?
[16:12] You've got European regulators, you've got the UK, you've got independent regulators in every Asian jurisdiction. Uh, add it up, you're getting a really large number. They all have their own rule books around data and how we have to handle data. Um, looking at that sort of, you know, a level deeper, those
[16:28] rules don't line up all the time. You know, what the ECB wants is not what the P wants. Definitions and standards can be different. Expectations can be different. Um, and we have to design a platform that works for all of them. Um that is really what it means to run the complexity of a you know a data in a global business
[16:43] right I mean it's I think it's clear for anybody who's been around data bricks if there's one word I would use to describe our product strategy it's unification put your data in the lake you like unify govern and distribute through unity catalog does that work well I in in the
[16:59] bank given there's so much regulation it does I mean you have to understand like banks have been built over generations of technology and data course if you look at sort of the transaction processing side you've got numerous numbers of databases that are running around those applications if you
[17:16] look on the analytics side banks because there we are information-based businesses we're always moving to the latest sort of analytical set of capabilities but you don't always consolidate or move things together the thing that had today is the value comes from bringing your data together into a
[17:33] high performance data store where we have used data bricks it has made a real difference you know in deposit processing in our bank cut the time down by 95%. We do a lot of work in um I'm sure everyone can appreciate the complexity of rag capital calculations
[17:49] what the extensive we have work we have to do in counterparty calculations um you know shorten those from really days to hours um in terms of the calculation capacity. So the performance of your data store the way you have it structured makes a real difference in
[18:05] business and allows you to meet all of these regulatory demands at scale. Yeah. Yeah. Well, you know, if you look at the innovation throughput that we've tried to deliver at this summit this week and for years, uh maybe
[18:20] comment on the pace of change like 20 years ago, 10 years ago, five you mentioned a few months ago like how how has the rate of change meaning how fast do you need to move to be able to keep up the technology and like support the environment?
[18:35] I've used the word machine speed before. you've got to move at machine speed. I mean the the environment has never evolved faster. Um if you pause on something for even a few months and come back to it, um the way you might address it is changing. Um that is a that's a
[18:53] that's a difficult environment even for technologists that are used to a very high pace of change in the industry. um the the degree of evolution, the degree of model changes, the degree of pivots between various types of technologies and what you're going to look at is
[19:08] constantly changing. Um and I think that that you have to just adjust to that level of change and you have to build yourself an architecture that allows you to support that level of change at the highest level. Right. It's got to be flexible, right? It's got to be flexible, right? Yeah. And open. Yep.
[19:25] I'm doing the shameless pitches for data bricks. So would you say that with the whole agentic by the way is that a that's a new word. I' never heard aentic maybe more than 24 months ago. Yeah, it's probably about right 24 36 months or so.
[19:40] Making up words tech. Yeah. So do you do you think that that pace of change is accelerating because of that or is it is it just the breadth of technology overall or is it the breadth of the amount of data the volume of data that you have to manage? Well, I think what you know what we've seen over kind of the last I would call it 24 months is
[19:59] the gradual winning over any skeptic around AI. You know, even our most seasoned engineers who initially said, you know, these coding tools will never write code as good as me. Um are now like okay, we need to we need to implement these at scale. Um and so that
[20:17] process has really driven and it continues to accelerate the pace of change. Um, everyone is becoming a change agent. Everyone wants to look the way to look at ways to use AI to drive productivity and to deliver results.
[20:32] That's driving incremental change. You could just look at the industry itself just in terms of just the pace at which models are coming out. The pace we see from from one groundbreaking model to the next next plateau shift, you know, sometimes isn't even three months, right? And that sets a whole new level
[20:48] of capabilities. And you've got to ask yourself when that capability level changes, do I have the right strategy? Can I go can I go one level up in sort of complexity of problem that I'm willing to take on? Am I taking on big enough challenges? Um, and I think that's part of the answer is to take on
[21:04] edge cases. Don't don't commit that everything has to be success when you start. Push the limit because by the time you're getting the project done, you will probably be able to get there. Okay. So you know we have a again a part of our development philosophy is choice
[21:20] I think it was it was repeated many times choice of cloud choice of model how important is that for you because you know not all models there's not one ring to rule them all y like how how do you think about that I guess I'd call it the flexibility
[21:36] to access your data across clouds across environments but also leverage these agentic tools the frontier models whether they're proprietary or open. Like how important is that? I think choice matters a lot. I mean, and I think you've seen it, right? I mean, we you've seen it live. The the the leading model is only in the lead
[21:54] for a short period before another one comes out from another competitor. It was only a short while ago where we thought open source was going to dominate. Then then one company and then yet another company. And so if you talk to the companies, you know, they're building the next generation. You understand like the next models will be
[22:11] one and a half trillion parameters, right? And so keeping your architecture open and flexible um is is really the way you should be designing things. Now sometimes if you if you take a little bit of lock in risk, which no one really wants to do, you may be able to get to
[22:27] market faster. That's the choices that you know people in my position have to make, my top engineers have to make. Do we tolerate a small amount of that to get a solution out quickly knowing that that risk and that that could be a risk that makes it unsuccessful in the very
[22:42] long run but in the short run we get to market quickly that's a balancing act and you got to think about that balance all the time in any project okay so something we announced AI gateway which I like and you might like but maybe for different reasons um if in
[22:59] your CEO hat you might like token maxing in the CFO hat. I like token valuing, right? And the the ability to pick which model is the right one for the right use case because they're, you know, they're not all made the same.
[23:14] Token maxing just leads to a ton of inefficient behavior. Um, I think that's coming out like it's it's not it's not the right sort of way to kind of how do you how do you address that? And you want to push the organization to be innovative. Well, what I've what I've what I've said is like when we look at a use case, you know, let's build to the
[23:32] best model and then let's bring the models down to the most efficient where the where there's really no deterioration in result. In many times it's the investment in data up front, not the model. You can run with a weaker model. Um in many times um a simpler model just gives you
[23:48] a quicker more direct answer from the underlying from the underlying data. Um and so it's and then it becomes about the auditing and checking right so you may have worse performance up front but if you can catch all the errors you don't need to you don't need to spend on the higher level of the model the
[24:04] frontier is not the solution to all problems um and I think we can see that right the latest generation of models are going to generate really large token bills yeah um and so we have to think about that how we run it now when you're putting it in end user tools hands it gets a little more complicated where
[24:20] people can choose the model and it's not an engineering problem, right? That's where it becomes a little more difficult because people default, oh, just give me the best, right? I think that requires a little bit of training, a little bit of coaching in terms of making sure people pick the right model for the job. Yeah. You said something in there that's
[24:35] I think resonates certainly for those of us at data bricks, which is we talk about context and context is really synonymous with the data. You don't go anywhere without your data state. Yeah. So how how do you think about the maturity of the firm's data like is how
[24:53] ready is it to be able to then leverage all the you know I I when I first came into this role I sat with Ali um this was probably the three or four four years ago and he we talked about an end vision of sort of you know how to structure data how to
[25:10] think about data in a firm that large and I think we shared a common end vision but I had a much more view of just incremental al improvement to get there, right? This is going to take time. You don't understand how big this is and how challenging this can be. Um, but I have to say he was right. I mean,
[25:26] the world has moved at a very fast pace and getting your data into a few high performance data stores as possible and off deep legacy. um getting the data quality to the level that it can support all of the use cases that you're trying
[25:42] to generate and building that sort of semantic layer across the firm are incredibly important to get the value out of AI. I think that's you know what we've been doing over the past two years is just been accelerating on that journey everywhere we can. You guys have been incredible partners, helped us on a
[25:59] number of cases. We got some of the we even got to the security levels that we need and now we can really start to ramp. Yeah, that's great. It's interesting and you know everybody in the room has their own set of requirements and I think outside of financial services people who look into
[26:17] financial services think there's a standard and everybody follows it but it's just not the case. Everybody has their own standards. So embracing those like that's that's you know there is a standard but I think we have to think is that standard is applied through different technologies
[26:32] and different architectures right so it gets it gets implemented differently in different companies and banks and so what works for one may not work for the other because how the how the original standard was met or how the problem was solved right and so that's where the uniformity really begins to break down
[26:49] as you kind of move into financial services okay so what we're seeing across the enterprise is and you know we're trying to enable this with Genie like you know natural language query of your data but we're seeing more and more demand
[27:04] outside of the traditional personas which are like engineering data scientist data analysts into marketing CFO CRO like all of the functions how does that play in financial services I just think it's it's another level of
[27:19] just the democratization of data across the organization you know, when I started my career, um, I had a sun station on my desk, right? 420 or uh, a 220 um, and I would write I would write code to pull data together
[27:35] because the PCs at the time weren't powerful enough at the Federal Reserve Board. Um, you think about how that when went to a generation of PCs where we could do various data work on the PC in Excel, VBA, Python became the standard.
[27:51] You didn't need these arcane statistical languages anymore. Data now if you just now now we're at a level of you can talk to your data. Yeah. You don't have to program in a language. You know machines are are you're you're you don't even see them. You don't know what's running it. You don't know the compute. Um it's a pretty
[28:08] amazing period of time to be in because it does change things across the infrastructure of the company like HR, like legal, like risk functions where that data has not been as available in past, you know, in past generations or systems as it could as it is now today
[28:25] or as as it should be and it makes a real difference in the way you operate the business. It's it's interesting as you were describing that technology. I I threw out the 420 220 just so I could date myself with you. Like that was the way back machine. Um but I left my phone backstage. I have a
[28:43] Genie app on my phone. Yeah. And the phone is more powerful than I think a Sun220 and I can do query real-time query in natural language of the data that's inside of data bricks governed that I have access to. And you can talk to it. Yeah. And I think that's one of the most
[28:58] exciting things. But we've done some initial work at voice um with our wealth advisors and we could see an immediate uptick in productivity when they can inter interact with the AI through speech. Yeah. Yeah. Excellent. Okay. Let's we talked about the wayback machine. What about
[29:13] the future? Like where do you see this playing out inside of the bank say you know three years, five years from now? Yeah. Yeah, I mean I think if if you look at kind of today's use cases, you can start to see how that scales and builds in the future, right? You know, probably one of the largest areas that
[29:30] we are focused on broadly as a pattern is document processing, right? So if you think about a bank, there's lots of functions in a bank where it was someone's job to take information from one file, put it in another. You know, a customer record submission for an AML KYC type operation. uh a mortgage loan
[29:48] underwriting where a customer is submitting documents. Um that's already AI that's creating that that final document with a human decision maker and in many cases that human decision makers getting now a recommendation from the underlying AI. Um we're only at the
[30:04] beginning when you look at sort of the scaling of processes. Um when you think about service, right, in terms of customer service, customers now will demand high quality AI interaction. They don't want to wait for a call. They don't want to be put on hold. They want the answer through any means they want, whether it's chat, call in, wherever
[30:19] they want to be able to get the answer to their problem or their question. Um, it becomes almost not just a way to save, but a way to deliver a better customer experience. Um, and I think, you know, we think about that service channel broadly, not just in for my old business, Erade, but in sort of how we inter interact
[30:36] information to our adviserss, what we're doing in level one support, how we're how we're servicing our employees. These are big functions within a firm of our size and that matters across all of that. Um development is changing dramatically. The way we can develop software is is already changing. That is
[30:53] going to continue to change. I mean implementing the Gentic coding tools at scale when you have 21,000 people who can commit code is a is a real challenge. But you start to already start to see where profound differences have been made. So when you get to the
[31:08] future, you really see an AI enabled firm, you know, a native firm where all of the sort of underlying processes that can be run with AI are and sort of firm where humans are working on strategic problems. Um they're working on relationship issues. Let's understand
[31:25] our business is a relationship business. It's about being the trusted adviser. Um and if we can empower those trusted adviserss to scale and do more business, then we're just going to grow and continue to grow. Okay. So last prediction in the future, we've talked a lot about how folks
[31:41] inside the firm will talk to their data. When does that reach your customers, the the individuals, the consumers, the the folks that leverage the firm externally? We're already starting to look at ways where customers can interact through voice. Um, you know, we like I said with
[31:59] today we're we're launching products where adviserss can speak to to what they want. You know, in terms of just queuing up information, pulling information about a client, asking for an asset allocation model or something for for that particular client. The next step is opening that to the client.
[32:15] Yeah. Um, and so really we'll use this work to leverage to get to that. Excellent. So, it's fun to look forward into the future, do predictions. This has been a great chat. I'm sure one prediction folks would like to know is like when will data bricks go public, but we're not talking about that today.
[32:31] So, I'll uh I'll leave it with that that we don't have that prediction engine. So, really appreciate it, Michael. Pleasure. Thank you. Yeah, thank you. Thank you, Dave and Michael. I'd like to welcome our next guests to
[32:48] the stage. Terresa Heisenry, chief data and analytics officer at JP Morgan Chase and Arcelon Davakoli Sharaji from SVP field engineering at Databrooks. Please give them a hand.
[33:19] All right. Thank you for coming, Teresa. Thanks for having me. They got us nice and cozy on these chairs. Absolutely. Um, well, look, I think JPMC doesn't really need much of a introduction in front of this group. If it does, they probably are in the wrong room at this point. Um, but maybe we start. You guys
[33:36] have been super public about your, you know, commitment to AI and what you want to do with it. So maybe help us understand where you where you kind of are in this journey and how it's going. Absolutely. So um you've heard a lot of uh rhetoric about the tone from the top and I think uh we certainly epitomize
[33:53] that. So our CEO is has always been very forwardleaning when it comes to AI and I think when you look at the journey that we've been on, it really is a journey over more than a decade at this point. So when we think about machine learning
[34:09] techniques and things that we've been using for ages in our business around fraud and pricing and marketing and if you think about like the size of our consumer business and and our institutional side that has real value very quantifiable and we continue to see that just grow and compound over time.
[34:27] Um so I stepped into this role for those who don't know I have a business background so not from the technology side uh from the investment banking side and as I've been in this role for three years I think we've lived another few cycles. So first it was about this idea
[34:44] now that you have the capability to give this powerful technology to every business user in the company. So the first thing we wanted to do is safely enable large language models, make sure that our data was protected, but really put it in people's hands. And we've done
[34:59] that. And we have of our 300,000 people in the company, twothirds of them actively using large language models every single day in their day-to-day work. and we've enhanced it continuously to make sure that in addition to the models, you have the data that's
[35:15] relevant to your function and tools that are relevant to your function if you're finance or legal or or what have you. And so I think the biggest thing that that's done for us is really drive this cultural shift in the organization like AI is now part of the way people do their jobs. Um we've definitely also
[35:33] focused on where are in addition to the ground up like where are the big top- down initiatives that are going to really matter to the company and um you know obvious choices even when you just look at the distribution of people in the company um at 50,000 engineers and
[35:49] technologists like enabling those people to use AI and be more efficient is a big is a big boost as well as call centers and all of those things. So we've we've gone at it top down and bottom up. Um, I think we're about to enter or we have entered the next wave which is now we
[36:05] we're into these much longer running autonomous agents and and that's coming with a whole set of other opportunities and also considerations that go with that. Fair enough. Well, look, you you talked about the pace of innovation there in some point. I mean, you guys have been
[36:21] on this multi-year journey. How do I adopt it? How do I go? Uh, AI has been evolving a lot itself, right? whether it's uh you know kind of as you think about models and feels like every week there's a new model that comes out there's a new third party capability so I'm curious how from a JPMC perspective
[36:39] because you guys obviously are this interesting balance of wanting to harness innovation and you know as having worked with you guys for a while you guys clearly have a couple of regulatory concerns and security concerns as you go through things as well. So how do you balance that to basically adopt that technology while
[36:56] continuing uh to think through what you know how to do it in a way that makes sense for JP? Yeah, I mean I think we always want to be on the cutting edge of leveraging this technology in ways that benefit our customers and our our company, but we are a highly regulated uh organization
[37:12] as you've rightly pointed out. So, we have to always do that in a way that respects the trust that our customers place in us and and really be thoughtful about the way that we go about it. Um, as we've looked at the evolution of the ecosystem, I think one of the words that
[37:27] you've heard over and over over the last couple of days is choice. Yeah. And I think that that's been one of the guiding principles that we've had from the beginning. um we want to make sure we we just are too large an organization and we're too embedded in the ecosystem
[37:43] of of the financial services world to have any over reliance on a single provider. So we want to make sure that we're both leveraging the best capabilities that are out there and working with the best uh uh providers in the ecosystem but doing it in a way that still maintains
[38:00] that optionality and flexibility. when I think in my role of what you know I'm responsible for basically enabling the firm to give them the tools to give them the governance and the standards that let everybody go quickly and then within each of our lines of business there are
[38:15] embedded teams that I really want focused on particular business problems and opportunities in that space. I think the biggest disservice that I can do is if I choose something that locks you in Yeah. too quickly. Like that's a you don't want
[38:30] oneway doors when you have technology that's moving at this pace. Yeah. And I mean, you know, building on that, we we've seen the similar thing like a lot of folks are trying to figure out exactly what you said. How do you harness innovation? But as we've seen now, it feels like every week there's a new model that's better or there's a new
[38:46] model now a model has been taken away and how do you basically balance that as an organization? Um but we talked a little bit about models and everybody's trying to figure out like what is a moat in this area right like how do you harness AI and everybody has access to
[39:01] the same uh you know kind of models and you hear people now talk about okay an AI transformation is a data transformation it's having access to really really clean and unique data maybe that's going to be the moat so I'm curious from a JPMC perspective you guys
[39:17] started on the data transformation way before the transformation so what does that been like like how do you make you know as an organization it like the data to be AI ready and then harness it to power many of the you know kind of use cases you just talked about well I think you're sitting on the stage
[39:32] probably the biggest convert in JP Morgan so when I used to be the CEO of a large business and the data people would come you'd say like I don't understand this but I I'm 100% converted um because without data we are not going to realize
[39:48] the ambitions that we have for the things that we can do with AI. Um I think when you look at this um you know where the advantages are JP Morgan services 87 million consumers in the United States like we do car loans and your deposits and your credit cards and
[40:04] we have all the Fortune 500 companies. We move uh trillions of dollars of cash every single day. So we see the payments flows. If you can harness that information to create much better experiences for your customers and to
[40:19] create insights across the company, you think the consumer side you're charging something and over here in payments you've got the merchants. You know, if you can join that up, you can create much richer experiences. Um and so this idea that you have to really focus on
[40:36] data as as a really valuable asset and that's for us been a a long journey that uh you've certainly been a part of. Um, I think that part of it is a culture change of really like elevating data and the people that are looking after it
[40:51] within the company and we've done a lot of that. But as we've gone through this journey, I think we're getting to the point now where it's it's very real and the work that we've done to create data products working with yourselves like when you can create the data products and now you have tools like Genie
[41:08] where our finance folks can exercise queries and talk to their data and get insights that otherwise they would have to wait on or would take months or hours. um it's a much better experience and I think that that
[41:24] flywheel has really started to turn in the company when people see what's possible when you have data that's high quality and well curated that it just cuts the time and just drives the insights in ways that I think are really
[41:42] starting to make a big impact in the company and and when you have agents like it's just going to take off from there. So I agree and you so you mentioned agents, you mentioned how great it is to get things in people's hands. Uh but as you mentioned early on, you know, JPMC is kind of just like a
[41:58] slightly regulated uh you know bank. And so and innovation is great, but now as especially as you start moving and you talk about agents, you go from agents that assist to ones that start to kind of automate and do work on your behalf. There's a whole bunch of open questions,
[42:15] right, around how do you think about identity? How do you think about entitlements? How do you think about access control? I mean um you talked about choice is a big one we've heard and you know we'll get to cost but it's like controls is also a big one. So like how does an organization like JPMC you
[42:30] talked about a bit but like think about AI governance so that you can hand things into people's hands. You said twothirds of your organization uses um you know LLM in a big way. How do you make sure that they're doing it in a secure way in a governed way? Yeah. Again it's an evolution. I think
[42:47] when we first started rolling out large language models to the whole population and you go from having a handful of like highly trained data scientists who I don't think had the most delightful experience by the way in the company of of traversing all of our controls but
[43:02] when you try to do that at scale across the company it's a very different story. So we've had to really evolve and adapt. I think the world we're in now from an agentic standpoint is another level. Like this is a new paradigm. It's not like governing software or technology
[43:18] and it's not like how we manage our employees where you hire somebody, they have a specific job, they have supervision, they understand, we teach them how JP Morgan works and how we do things. Like you've got a little bit of this hybrid going on, right? So as and
[43:34] you have no rule book like the rules haven't been written yet. So what we're trying to really do here is what are those principles that are going to be enduring like what are the things that we we know have to be true um as we move forward to help set those standards. So
[43:52] fully agree you have to understand that the agent has its own identity and that that persists as the agent operates. You have to be able to separate that. You have to have the agent operating in, you know, within some kind of parameters, right? It what makes them powerful and
[44:09] great is that they can, you know, take actions and do a lot of things, but in the same token, you want to make sure that the things that they're accessing are specific to what they're trying to do and that those entitlements are properly permissioned. So, we're having to continue to evolve the thinking
[44:25] around this and it's also governance in real time. There is no such thing as kind of after the fact or check it before it goes in like these these agents are dynamic. They're not deterministic. And so you need to make sure that as you're doing the governance
[44:41] that it's you know this is the definition of shift left. It has to be embedded for the scale that we're trying to operate at. Which makes sense. And look maybe building upon that uh you said it yourself earlier. I think your CEO is
[44:56] nobody would describe him as shy or basically like about talking about what he wants to see AI do. And then in our conversations, I think the one thing I've always struck with been struck with is that you're always like AI for the sake of AI is you end up with what this
[45:11] token maxing phenomenon, right? But it's like what I really care about is transforming how the business operates with AI. Like what is the productivity? What is the output that we get with it? um and especially size and scale of JP like that's not trivial. I mean
[45:26] technology is a part of it but there's other things that come with it. So how do you think about transforming the operations to the AI fueled at a place like JPMC? It is by far the biggest um challenge of of really getting this to where it needs to be. Um I think you know when people
[45:43] ask me about the AI strategy of JP Morgan and I I say there is no such thing. there are business strategies that AI enables and and and as we work with our businesses um the technology is so amazing that it's very easy to get um
[46:00] focused on the tool itself like when you sit down and you see like what what can be done like that creates a lot of excitement. So we want all of our leaders to really understand like what is the frontier like what is the cutting edge of what's possible and we are that
[46:16] that is about putting pe the tools in people's hands and making sure they understand that. But the really important thing like that's kind of our job like I'm here to enable you. I don't want you to have to worry about the tools or how things interoperate or work together. What I want you to do is focus
[46:32] on what are the things in your business that matter. So I always ask our business leaders, what are the three metrics in your business that if you change them in a meaningful way, if you were really audacious in the way that you think about the goals, like what would those things be? It could be clock
[46:48] speed, it could be assets under management, it could be how many clients you can cover and then what is it that's holding you back from doing that and how can these tools help you? Like that is the biggest cultural change to drive. It's it's very hard to reimagine the way
[47:04] you do things or what's possible and and that's kind of the journey that we're on. I think that's where the real value unlock is going to come from which so you've mentioned a couple of things now for keys to success. Um I think you mentioned data and context being important to make it. You
[47:19] mentioned security and control. You said choice which is important. Um one thing we haven't touched upon as much is you know kind of cost. You hear all these people talk about overruns. So, is that one that you guys just had a handle on from the early place or it's been more like let a thousand flowers bloom or in the middle like how do you think about
[47:35] costs and budgets for something like this? Yeah, I I think when you're doing um so so when we rolled out large language models to most of the people in the company um it it it is actually spawning innovation and we really did not go to great lengths to try to create an ROI.
[47:52] It's just we're considering it table stakes. This is the way everybody's going to operate. We need these tools to be effective and we understand it's creating hours of productivity per week for people and that's a great thing. Now you get into the next wave where you've
[48:08] got these longunning agents that are doing a lot of inference. They're they're creating a lot of tokens and I think it's not about giving people a tool or an assistant that helps them. It's about giving them a complete workforce. Right? So think about it like you don't let people go out and hire 12
[48:25] people and say I don't really know what they're going to do but when they get here we'll figure it out right you you want to make sure that you understand like what's the objective that you're trying to achieve right so I think it's a little bit of a mindset shift like this is not Q&A this is like you can
[48:41] drive some real cost and that's not necessarily a bad thing yeah as long as you understand the outcome on the other side and that you're measuring that you know I think it also is shifting And that's a lot of things that you've spoken about over the last few days here. It's shifting around, you know, your
[48:57] choice of models. Yeah. How are you optimizing? How are you, you know, not using, you know, you have to use the right model for the right job. You don't just default to the most powerful models. Where do you want your compute to be? How do you balance all of that? So, I think there's so much more
[49:12] optimization. Like you talked about gateways and things. I think there's so much more optimization. Yeah. that is to come as we continue to kind of move forward in this in this wave that we're in. Always an exciting time. Always exciting. Um so maybe one or two last things for
[49:29] you here. So one uh you know I think JPMC always has a reputation for kind of being at the bleeding edge of financial uh paving the path and you kind of I know you'd say hey we have a lot more to go on the journey but you guys have you know gone a fair way. Uh maybe what are
[49:46] some of the big, you know, I would say lessons you've learned that you're like, look, for the folks in the room who who are going along that same journey, maybe a year or two behind you that you'd say be if I had to do it over again, maybe here's a couple of the key things that I think are really important in to kind of invest or do as you go through it. Yeah. Uh I mean, I think we've covered
[50:02] some of them already, but I I cannot emphasize enough the importance of data. Like if you and it's hard work and and if you don't do that work upfront, it's going to slow you down. So if there's something I wish we could have even accelerated more, um I think that that is the critical
[50:18] piece. Understand what is valuable about your data, curate it, get people to own it, create the right infrastructure, and just that will just create so much speed for you. Um it's it's amazing. Um you know, I think the governance is another thing you really have to think and be
[50:35] very intentional about how to make it scale. like it it's a different challenge now of you know just when you think about the scope and the scale of of how this has to work that is a very intentional decision. So getting those two things right the rest will follow
[50:53] like and and they're not the sexy things that's not things that everybody wants to focus on but those are so foundational and fundamental that you know that that will expedite your journey by leaps and bounds. Fair. It's funny that you say that they're not the
[51:09] exciting things, you know, at the summit here. I've had to do a couple press interviews like what is important and I'm like governance and I'm like data and the lady's like yawn and I'm like I don't know what to tell you. Like it is boring but it is what you have to do at this point. Um so I I won't ask you the question
[51:26] that everybody always asks which is like what will the world look like in five years because I'm like nobody knows what the world will look like in five years but maybe a simpler one. you kind of had you've spent all of this time setting this foundation getting it ready and you guys have some of the most interesting kind of problems and opportunities to
[51:41] apply AI to if I just say if we're sitting back here in a year from now and I said hey like what has changed what have you done what are some of the big things you're like here's what I'll be eager to see us uh make a lot of progress on yeah I I've spent my entire career at JP Morgan and I think this is the most exciting time I've seen um because of
[52:00] what it just allows to be possible possible. So, the clock speed of things that you can do, the like there's so many great ideas and the great ideas always are limited by budget and how long it's going to take or how much of a
[52:15] development effort it's going to be. You're just unconstrained at this point. And I'm so eager to see like what that unconstrained innovation looks like. Yeah. I just think that it's going to be amazing. I also think that it takes, you
[52:31] know, it it frees people to do the things they like to do. People like to be innovative. They like to work with customers. They like to come up with new solutions and solve problems. And I just think that we're all going to have so much additional time for the
[52:46] things that we probably all enjoy doing much more that I I think it's going to be amazing. Sounds great. Well, I'm sure I could ask a million more questions, but they're going to play the Oscar music and pull us off. So Teresa, thank you again for the partnership. Thank you for the partnership. We appreciate it. Thank you. Awesome. Thank you all.
[53:08] Thank you Teresa and Arcelon. Love to welcome to the stage Rahul Desh Pande, EVP of global head of research and development at Mastercard and Jennifer Miller, global head of banking and payments go to market at datab bricks.
[53:32] All right. Hello everybody. Glad to see so many faces. Rahul, thank you for joining us. No, thanks. Thanks u thanks for your hospitality. Excited for the conversation. Um before we jump into questions, I got to brag about Mastercard here for a little bit. Mastercard processes 175 billion trans
[53:49] that's with a B billion transactions per year across 200 countries. Incredible scale. But that's actually not the headline. The headline is what you are doing with that raw data and how you're turning that into innovation. And that innovation led to Mastercard just being
[54:06] named the top future company uh right next to amazing tech giants like Nvidia and Microsoft. So huge congratulations. No, thank you. That's that's amazing. That's absolutely incredible. Absolutely. I love that. Um All right, let's jump in. So, you're focused on turning that
[54:21] immense scale into innovation. Can you talk about why now is such a critical time for that and how data bicks is helping to enable that? Yeah, so um Mastercard has been business for a while. Um our core mission has been to connect the uh connect the
[54:38] consumers with the businesses and the governments. And the way we do it is through simple transactions and transactions that are simple, smart and safe, right? U now we do a lot of these transactions as you imagine and so there's a lot of data that is coming in. Um so that data then turns into
[54:54] intelligence and then that intelligence is where we provide different services uh in decisioning in security products um and in personalization. And so that value that it generates kind of drives this flywheel, right? and it it then creates more transactions and the the
[55:10] flywheel continues. So it's a very virtuous cycle that we have built over the last decade or so. So the last few years obviously there's a lot of hu you know three different shifts if I may say so that are happening. One is obviously technology and AI and data and platforms that exploding data bricks is obviously
[55:27] one of the leading ones. Um we also have the change in the expectations of what customers are looking for. they are looking for now real time insights um actionable insights and then finally because of all these changes the trust is becoming a bigger issue as such. So
[55:44] there's a heightened uh need for governance and control and transparency. So um that's the world that we are living in and so our response has been to u you know to start with the foundation and build a new modern foundation data foundation that is based
[55:59] on the data bricks uh lakehouse architecture and so the what we're doing is we're bringing together data AI and analytics together uh to enable a faster development cycles to build new products in the same uh things that we're doing
[56:15] and but then with the uh with the trust and control by design. So having the control layer, you know, using Unity catalog for example, the goal is to then provide that trust underlying trust, but then also go faster into the market. Amazing. Amazing. That sounds like it
[56:31] sets the foundation for new model model development. You you've talked about building new foundation models that are very specific to commerce. Can you talk about what that work unlocks for you all? Yeah, so it's very interesting. you know the um so the transformer neutral neural networks is what is used for LMS and
[56:49] it's used typically for the unstructured data uh and you know the way it does it it it gets the next word in the lang natural language and it predicts that word based on a token uh but then these um these transform neural networks are
[57:05] also good for time series u data so which is the structured data which is where the transactions come in so we we are trying to use the uh that part of transformer models to create the uh large table models. So based on the transactions that we already have. So we
[57:21] are fitting billions of transactions uh anomized data to these transaction models. And so what's that unlocks or what is we are starting to see it unlock is um you know so if you're traditionally we are AI native company
[57:36] we have doing AI for the last 20 plus years. uh but the way you do AI uh is you look at the business problem then you build a business you know a particular AI model um you add more features to it and then more features you add the more uh more detailed the
[57:52] model is more accurate the model is and so now there's a tons of models so what LTM allows us to do is to have one model that can solve for multiple different problems that's what we're hoping for and we are starting to see some early good results uh especially
[58:09] in the um in the false false positives where a transaction is you know is is considered to be a fraudulent when it's not. An example uh and these are the extreme examples where you know once in a lifetime event where you know you use a car typically for groceries and gas
[58:26] and then one fine day you go and and buy a engagement ring. Now that's a once in a lifetime event. It's anomaly right? Hopefully it's once in a lifetime. But the um you know but the goal is again to to find these uh these are these are obviously priceless events right and we
[58:42] want to make sure that those are not declined uh because it's a bad experience. So this is where these LTMs are being uh we we seeing them to be uh productive in actually decline you know removing some of these false declines.
[58:58] So it's a journey that we are on. Uh but it's a very interesting space. That's for sure. Yeah. Amazing. I love the the little pop in there on priceless. Nicely done. Nicely done. Um so it's awesome that you have all this intelligence, right? But I presume you and your team think about
[59:13] how do we take this intelligence and build new tools and business value and unlock that for your end customers. And you recently announced the virtual CC seuite product for small business. And I love that because small business makes up 90% of the total companies in the world and 50% of the global GDP. Um but
[59:31] they also operate with a very lean team. So can you talk a bit about your virtual seuite product? What problems are you solving uh for your customers and why did you start with the CFO? Yeah, that's a great question. I mean first of all I mean that that sector is
[59:49] very personal to me. My parents are a small business owner. So, I've seen some of these problems firsthand where, you know, you're wearing different hats, right? You have probably managing the same amount of complexity in terms of the business itself, but then there's no one else to do it. So, you're you're
[01:00:06] being your own CMO, uh, marketing, you're being your chief financial officer, you're being operational officer. You're wearing these different hats and you have to be good at all those and most of the businesses are not right. At the same time, you know, we have lot of data that is available to
[01:00:21] them, right? Right. So there's payment data, there's operational data there. Uh you know we have the cash flows are available but then since you you don't have time as well as the expertise um so that's where the AI comes in and so that's what
[01:00:36] we are looking at. uh in fact and so why we're going into finance uh CFO agent first of all is about 83% of the um small businesses fail in the first five years uh simply because of the um they
[01:00:51] are not able to handle the cash flow that's a huge statistic so and it's very simple for you to actually do that you know so um if you have the data that is associated with for example right if you have an invoice that is coming in the invoice has terms like 830 which says
[01:01:08] well it's supposed to be paid in 30 days. Um so why do you want to pay it now right? Uh is there a way for you to then negotiate the term to be well 60 days or 90 days um and there is another one like 210 which is well if you pay that invoice which is due in 30 days in
[01:01:24] two days uh sorry in 10 days you get 2% off. So simply by looking at some of these nuances about the invoices that are coming in and then looking at your operations and seeing what's going on around your business, uh you can probably make more meaningful decisions, right? And that's huge. So that's the,
[01:01:40] you know, and that's where the AI is really really helpful. Um and so again, it's a huge um there's a huge need for that um as the world is making more and more complex. So um you know, this is this is going to be cool. Yeah, I love it. Um, and I agree like I I think the opportunity to serve the
[01:01:57] small business segment is vast and I'm really excited to see the impact that that has on on that community. Let's pivot to the issuing banks. And as a former banker myself, I know all too well that turning insights into action can be difficult for banks. The amount of data they have is is vast. But
[01:02:13] transitioning that into insights that you can act on oftentimes takes a long time. uh can you talk about how you are problem solving that for for issuing banks? Yeah. So the so the problem that we are trying to solve is about you know how
[01:02:30] you look at u an insight that is coming in from a static dashboard for example your um your car decline rate uh has gone up right so that's a huge problem uh so today what most of the banks do is look at you know a static dashboard that
[01:02:46] gets updated once once in a month or once in a week and then then they to figure out what actually happened there right um to figure out and then then based on that they had to do an action in some other systems and then finally look at what you know did that action
[01:03:02] actually help in in lowering that uh decline rate and so there are different systems that have been used and this is where we have we're launching the product called performance pulse uh which is uh which which which lets the customers continuously monitor the
[01:03:19] performance um of your portfolio and then uh go from the insight which is there to um to the action to then measurement and then continue that loop right in in one u uh one system and so
[01:03:35] we using AI there obviously um that's where AI is really helpful it's a natural language so we are using multiple models uh we create our own models um and we are using uh lake base I guess in in this so that we can give those insights right away u so that is
[01:03:51] one of the coolest uses of link base that we can find. Uh we are grateful for database for that. Yeah. Amazing. And I it's an important use case because even a small amount of change in credit card spend can have a a massive impact for bank. So that's
[01:04:08] terrific. All right. So you spend a lot of time in data and AI. What is your advice for those in the room as you think about leaders versus followers? What are the leaders doing uh right now to prepare for this future? Yeah, I mean so there are changes that are happening
[01:04:24] right? So one is technology is evolving rapidly uh the consumers are expecting more um and then at the same time the trust is also important right keep the trust and so trust is the currency for innovation as it says and it's very true
[01:04:39] right and so the way we're doing it is by setting up the foundation and that's what you would do I I heard uh from the other speakers also you start with building a data foundation um then you put an analytics layer on it you put a control and governance layer on top of it and then you start building the
[01:04:56] products. You don't start building the products and then figure out some of these things. So I mean obviously that would be the advice is to start with the uh with the foundation and start building on top of it. Uh if the foundation is weak, the trust goes away
[01:05:11] and then innovation goes away right and that will be the worst thing that can happen in this world of AI which there's so much opportunities. Um at the same time we just we got to be careful about how we using some of these tools. Yeah. Amazing. Yeah. Amazing. That was
[01:05:26] such a master class. In seven minutes you covered three incredible innovations that Mastercard is delivering to your end customers. Congratulations on such an incredible body of work. You're literally building an intelligence layer for global commerce. Uh which is which
[01:05:42] is incredible. and you're not just predicting what that future is going to look like, but you're actually building it and you're leading the charge. So, huge congratulations. Thank you. Thank you so much for being here and sharing your incredible insights with the group. No, and thank you for data bricks being the great partner. Um, you have been
[01:05:58] always been and we don't consider them to be a vendor. We consider them to be a partner for a lot of reasons including the joint development that we have done. Uh, starting with Mosaic now with Lakebase and you know and the journey continues. So, amazing. Thank you. Yeah. Amazing. Thank you.
[01:06:20] Kevin. All right. Thank you so much, Jennifer and Raul. I'd love to welcome to the stage our final speaker of today Bobby Gruber from RBC Capital Markets. He is the global head of AI and digital innovation.
[01:06:53] Good afternoon everyone. Great to be here. Energizing data bricks conference. Truly a privilege to be here on stage. Uh thank everyone from data bricks for having us. Uh normally I'm slowing down the pace of how I speak but I committed to Kim to get this uh done in five minutes. So maybe I'll go at my normal
[01:07:08] pace. Uh happy to go through our RBC journey in an accelerated version and then really get into being able to demo an AI capability uh that Mike Tran our head of digital take us through. So RBC bank uh been investing in AI for more
[01:07:24] than a decade. Dave Mai the CEO of the bank had the vision to hire Dr. Fini Fiod. She stood up Borealis AI. This is a decade ago. Hired a hundred machine learning experts, AI scientists and researchers. And that was the beginning
[01:07:40] of our AI journey. The beginning of the journey for capital markets was 2016. Uh we co-invented with Borealis a platform called Aiden and Aiden executed optim optimized execution for our clients. That was the beginning of our journey
[01:07:55] there. Fast forward to today, our AI team in the business is accountable for building AI capabilities with the business, driving revenue growth, accelerating AI execution, and moving AI
[01:08:11] capabilities onto the income statement. So the ROI, and we heard Arcelum uh talk yesterday a lot about ROI. Today we're going to talk about how we're doing it and why data bricks and why data bricks is such a critical partner. Datab Brick's core thesis of democratizing data and AI aligns exactly
[01:08:28] with RBC's focus of leaving nobody behind on this AI journey. We chose data bricks because they help us safely accelerate growth and accelerate execution. If we think about what data bricks offers, what no other platform offers is
[01:08:44] a unified data layer to ingest unstructured and structured data into an AI ready foundation. So a data layer that moves with speed and precision data bricks delta streaming allows us to take latency in news and market data down to
[01:09:00] 200 milliseconds uh or even faster. Then there's a governance the unity catalog the security the discovery the audability all one place. So the data layer governance one place allows us to accelerate execution. we take a step
[01:09:16] back and put it in practice and our flagship AI innovation that Mike Tran is going to go through this AI capability and demo it live be really the exciting part but if we take a step back March of 2023 no one talking about agentic AI at that period of time our capital market
[01:09:31] CEO Derek Nelder reviewed the Gen AI technology we had been learning about it for actually six months a year before that the next layer after all the reinforcement learning AI work we had done and Derek you supporter of capital markets research knows that business
[01:09:46] really really well. Looked at Gen AI, looked at the vision, said what do you need? Said watch our back when we're pushing boundaries and go for it. So that was the start of the journey in applying Gen AI to research. We set up our office of AI, one front door for all
[01:10:03] things AI across all out 8,000 people in capital markets. The problem we're solving is that our analysts cover 1,700 companies. Now 100 analysts cover 1,700 companies. If you think about on average an analyst will spend 60 minutes on an
[01:10:20] earnings release per company. So 1,700 companies 60 minutes. It's 102,000 minutes. It's actually 141 12-h hour work days. We co-build this solution called data quick takes with Michael
[01:10:36] take us through uh with data bricks and we've taken 1002,000 minutes down to minutes or or or hours and not only did we co-build it but we actually locked in with the CTO team from data bricks with the talent that work directly with our
[01:10:51] AI engineers some of them are here Santi Barl and Andre Motti and but we locked in and we worked together to build a client and commercial solution together so big success that inspired us and to build out our ADOM platform for all 8,000 people in capital markets. Again,
[01:11:07] we heard amazing talk from Arcelon yesterday and talking about the foundation, the adoption, the ROI and kind of aligning that to our journey. So, our platform for 8,000 people is called Aiden. It's a beautiful intuitive obvious interface that we co-designed with Coher. We have a exclusive
[01:11:23] arrangement with them in Canada. The capability has an agent creation process. capital markets people created up to almost 10,000 agents. Now it's powered by the most cutting edge models like Anthropic. It's real-time data with
[01:11:38] Ravepack and if you listen to the ROI our people have asked 4.7 million questions and we save we quantify everything in the capital markets quantify and measure everything but we save 1.7 million hours and you can uh align that back to client activity and driving client and commercial outcomes
[01:11:54] and it always comes back uh to the to those business outcomes. On top of that, we build AI capabilities. We're building a front-toback AI enabled investment bank. If we think about accelerating cycle times, we created a pitchbook creator, our engineering team. It's a
[01:12:10] really, really hard engineering feat. And we're accelerating cycle times there from 14 days to five. Client meeting prep, accelerated cycle time, 7 to two days. 50% time spent reduction in financial modeling. All of this then is applied back to covering clients and
[01:12:26] driving more commercial outcomes. And this is where data brea again is uh invaluable to us. Quickly where are we going? People may know our CEO committed to the marketplace that will drive and deliver a billion dollars of AI
[01:12:41] generated value to our shareholders. Uh how are we going to do that? We are very focused on making the technology invisible. What does that mean? It means bringing AI to where people work. Aiden
[01:12:57] in Excel, Aiden in PowerPoint, Aiden in Outlook. We heard Sarah Frier yesterday at one of the sessions uh talk about her automated email process or the time sequence. Well, we're building our own AI thought partners. And having your own AI thought partner and bringing AI to
[01:13:13] where you are will allow ultimately our vision is that it allows the adoption to actually become business as usual. You don't do built-ins built-ons. You actually bring it right into the system into the tools that our sales and
[01:13:28] traders use for example all the decision making process. So it's not kind of you going to AI AI comes to you. are doing our job and using AI, we're making that line disappear. So, and again, the AI
[01:13:43] ready foundation for data bricks makes all this uh achievable. So, very big ambition. Lastly, so hopefully I meet meet the five minutes um before we get to the really exciting part. I want to touch on culture in this AI world and at
[01:13:59] RBC culture is very important to us. We believe we have a very unique culture. Some of our observations and I'm sure everyone here feels it in this AI world. We're in race conditions. Time to client, speed of innovation, speed of the releases, just the releases from
[01:14:15] data bricks alone. Talking to Braden like, okay, how can we now apply this and do more and scale up and scale out. So the pace is incredible. But more than AI, we're solving a change and transformation problem. And you don't solve it with software, you solve it with culture. And we always say our team
[01:14:32] that culture trumps technology and the humanness the cultures that instill the humility the courage uh the curiosity in people will be the cultures uh will be the firms that win and driving adoption
[01:14:48] and education AI adoption education those firms will win. We heard it yesterday again from Arceline loved his talk if all our notes could send it around our firm but we heard it yesterday. We stood up an AI champion community. One of the best things we've done culturally. It's 750 people across 8,000. We meet weekly. Every single week
[01:15:05] we're getting 5 to 600 people attend. One week is education. We're blessed to have Borealis that can educate us. Then we have evidence of use case. And then we have innovative strategy partners like data bricks come in and Braden and Nasser and the whole team came in and they talked to this community, but it's
[01:15:21] recorded and it goes to 8,000 people, but it really goes to 100,000 people across the bank. So you have the data data bricks leadership telling us what's the art of the possible grounded in practical execution to deliver for our own clients.
[01:15:36] And yet data is our differentiator. Data is our moat. Data is our competitive edge and our partnership with data bricks protects it. Just in closing before introducing Mike,
[01:15:52] we heard Alli Godsy last year say that he believes that AI can help elevate every person on the planet. And at RBC, we believe that. We're living that as our collective ambition and we want to leave no community behind as well. So on that exciting and uplifting note, really
[01:16:09] excited to introduce Michael Tran, head of digital product for capital markets to really get into the fun uh walking through the AI capability and AI demo. Mike. All right. Good luck. You're gonna crush it. You got this.
[01:16:33] On Bobby's comments at RBC, we started our generative AI journey in capital markets research. Some of our peers started in wealth management. Others started in sales and trading. Some banks started in investment banking. Again, it was a very
[01:16:48] calculated, very strategic decision by our CEO alongside Bobby to start our generative AI journey in capital markets research. Why do we do this? The reason is this. It's quite simple. We are laser
[01:17:04] focused. We are obsessed with key measurables, KPIs. Now, when you think about the research analyst persona, we know exactly how many research reports you've written over the course of the past month. We know exactly how many clients you've
[01:17:21] talked to on a week over week, quarter over-arter basis. We know exactly how many companies you're covering at any point in time. At the same time, we can see all the broker votes. We can see your IIA rankings. We can benchmark
[01:17:36] against your peers on the street. When you think about this slide, what we show here is our vision. It's our strategy that's showcasing the digital transformation of our global research department. We like to call this the
[01:17:52] nucleus slide. When you look at this slide, the nucleus is Aiden, our foundational models. It sits smack dab right in the middle of the page. When you think about Aiden, it powers everything. It sits at the center of everything that we do across capital
[01:18:08] markets as it pertains to artificial intelligence. When you look at the rings around the nucleus, this is a mosaic of AI agents all working together to elevate our global research team to be able to compete at the highest levels by leveraging AI.
[01:18:27] Now, when you think about when you think about Aiden research in a sentence, what does Aiden research stand for? Scaling analyst productivity without diminishing quality. Scaling analyst productivity without diminishing quality just rolls off the tongue. But we live
[01:18:43] and we breathe this every single day. So let me break this down into two components. First, scaling productivity. What does this mean for us? Our northstar within our capital markets research division globally is to
[01:18:59] increase the number of companies that we cover. Scale ticker coverage. You may say, well, why are you so laser focused on scaling ticker coverage? Again, the answer is simple. It's client relevance. The more companies you cover on a global basis,
[01:19:14] the more market intelligence you have, the more that investors will trade with you, the more that corporates will bank with you. Simple as that. Becoming more relevant, bringing better outcomes to our clients. So let me take you into the tent and walk you through our journey a little
[01:19:31] bit in terms of how we've been doing in terms of scaling productivity. If I think back to 2023 when we started our gen Genai journey in capital markets, the average research analyst at RBC covered 15 companies.
[01:19:48] Fast forward to today, that same research analyst now covers 17 and a half companies. We have an AI enabled blueprint to take that research analyst to covering 22 companies. Now to benchmark this with
[01:20:04] our peers just to give everybody an idea of what that means on average when you look at our bold bracket peers the average analyst at those at those banks cover somewhere between 16 to 17 companies. So as we go from 15 to
[01:20:20] current state at 17 and a half ultimately to 22 companies by definition our ambition is ultimately to make us the most efficient global research department on Wall Street. Scaling productivity now without
[01:20:37] diminishing quality is just as important. And there's many ways to think about measuring the quality of a research product. But what I'll share here is what we're really proud of. So look, we're in a room of of financial services professionals. So I don't need
[01:20:53] to explain what Euromoney is to everybody in this room, but Euromoney is the Grammys for bankers. Euromoney is the Oscars for bankers. And we're really proud that when Euromoney announced their awards a few months ago, Euromoney
[01:21:08] named RBC the best investment bank in North America for research. And when our CEO Derek Nelner accepted the award, he attributed the number one ranking to the digital transformation of our global research department. Scaling productivity without diminishing
[01:21:26] quality. We're really proud of that. So what we are showing you here today is is Aiden research. Let us let me take you inside the tent and the best way is really to show you a demo of what we've been building. Now what we're going to take you through
[01:21:41] today, we're going to take you through a persona of Walter Sprackckland. Walter Sprackcklin is our star industrials analyst. He covers a rail companies. He covers the transport companies for us. Walter Sprackcklin is also the head of Canadian equity research. Now you ask
[01:21:58] any research analyst and the biggest pain point is always earning season. So when you think about 15, 17, 20 names, you have these companies all reporting, all reporting earnings four times a year, all within a very short span of several days. This is the most
[01:22:13] competitive time of year. We're looking to get insights to clients faster than the competition. At the same time, we're trying to work with our our clients, our investors because at that point in time during earnings, it's the fog of war. This is when we need to link arms and
[01:22:29] help our clients the most. So, what we're going to do is walk you through four distinct different modules as it pertains to earning season. The first one, as Bobby described, quick takes. What you're going to see with this demo is an acceleration of insights to the
[01:22:45] client inbox faster than the competition. So, let's take you inside Walter's framework. We'll take you into the user interface of QuickTakes. What we're going to do is we're going to take you inside the QuickTake module. Walter Sprackcklin again covers the rail
[01:23:02] companies. So we're going to walk you through when a rail company, let's call it Union Pacific, big company, reports earnings and we're going to walk you through the day in the life of how this ultimately works. Now what you see here is Walter's Walter's coverage universe.
[01:23:19] The key idea here again for this tool quicktakes we're optimizing for speed to the client inbox so that we can bring insights to clients faster so that they can make smarter higher con higher conviction investment decisions.
[01:23:36] Now when you think about a company like Union Pacific there's 25 bank analysts that cover this rail company. The reason why we're optimizing for speed is once the press release comes out, if you're not the first, if you're not the second,
[01:23:52] if you're not the third note research publication into the client inbox and instead you're the 13th or you're the 23rd, you're just simply not relevant. So again, what we're doing here is we're optimizing for speed. So let's click on
[01:24:09] the the Union Pacific module here. What you'll see load up here is a user interface. You see the time stamp. Union Pacific is about to report earnings at 7 a.m. in the morning. What Walter needs to do as he goes into
[01:24:25] this user interface is quite simple. It's only it's only one action. The table at the bottom, this is where Walter would input his estimates for earnings per share, cash flow per share, all the other operational metrics that is important for him to uh communicate
[01:24:42] to investors. At the same time, what's he what he's going to also do is input street consensus estimates. And as such, this is what's telling Aiden, our model, to understand if Union Pacific is beating or missing earnings.
[01:24:57] Now, envision the scenario where the press release just launched by Union Pacific at 7 a.m. Within 200 milliseconds, what Aiden has done is it's gone in ripped all the content off the press release and it's
[01:25:14] holding it on our cloud platform. And as Walter presses the generate button, what you're going to see is momentarily you're going to get a client ready note that pops up on the right hand side of the screen.
[01:25:30] Now, let's not be fooled. This is not just a simple summary of the press release. What we did with Aiden was we trained it on the voice and stylistics of Walter Spracklin. Walter's worked at the bank for over 20 years. He's written hundreds
[01:25:47] of research reports. He's written many, many on Union Pacific. And what Ada knows is Walter's voice, his stylistics, how he annunciates his words, his vocabulary,
[01:26:03] all the KPIs that Walter's all always cared about shows up in his note. So from there, Walter is able to copy, paste into our publication software, edit, put any sort of spin he wants on it, but it's client ready.
[01:26:19] And with that, this authoring process typically would take a research analyst 45, 60, 90, sometimes 120 minutes. We truncated hours of work into minutes. And this is how we're winning the earnings war to deliver insights to
[01:26:35] clients faster than ever before. Now, if we zoom out for a second, when we think about building AI product, particularly one that's focused on speed, what are we measuring as we're building this? We're laser
[01:26:51] focused on three different KPIs. Speed, accuracy, and relevancy. Number one, speed. You saw how fast this is turning hours of work into minutes. Number two is accuracy. When we talk about accuracy, I mean, frankly speaking,
[01:27:06] humbly speaking, as we were building this, the first turn of quick takes, we had accuracy at 80%. Before we rolled it out, we never rolled it out at that point. 80% nobody would use it. 90 95% nobody would use it. We
[01:27:22] held ourselves to a really high bar. What is that bar? What we effectively outlined was if you're authoring something like this, it needs to be higher accuracy than a human fat finger. We defined it as 99%. And over the since rolling this out two
[01:27:38] years ago, we're really happy to report that accuracy levels 99.7%. So we really hold ourselves to that regard. Speed, accuracy, the last one is relevancy. And this is why I highlighted the training on the voice and stylistics of
[01:27:55] the research analysts. Those were the three KPIs. Now, if we move on to the second module as part of the earnings process is after the press release comes out, an hour or two later, we have um a
[01:28:11] management conference call, an earnings call. And this is where um management teams, so the CFO, CEO, often the COO get on a call with all the research analysts that cover the stock from a bank perspective. So in this case, Union
[01:28:26] Pacific would have 20 25 analysts sitting on that call. What we've engineered is what we call earnings agent where you're able to auto extract key intelligence, key insights from the earnings call and go deeper than ever
[01:28:43] before. The way earnings calls were done in the past was the research analyst would be on the phone uh and listening to the listening to dialogue. As we take you inside this module, earnings agent, very similar user interface. We'll click
[01:29:00] on Union Pacific and what you'll see on the right hand side is a live streaming transcript of the earnings call that takes place. Now, if you remember, we've timestamped everything. As Walter got his node out
[01:29:16] to clients faster than the competition, what happened was this unlocked Walter to be able to start calling clients faster than than his peers. As such, he's able to steal more mind share. And as we steal more mind share from our
[01:29:31] competitors, we turn this into stealing more market share from our competitors. As such, for pure demonstration purposes, Walter is 5 minutes late for this earnings call. Remember, every other analyst is on the phone. We have a live streaming transcript with little to
[01:29:48] no latency. He joins at 8:35, 5 minutes late. We did this purposely for a demo to show you that we have a catch me up feature that shows up on the lefth hand side of the page. Now everything that you've missed that's
[01:30:03] key is summarized for you in real time and this refreshes for you. Now what's also really important is if you look at the top left hand corner, you also get a push notification. Now how is this different?
[01:30:19] Again, Walter's worked at the bank for 20 years. We've taken the history of Walter's research. We've plugged it into the back end of Aiden and it knows exactly what Walter's price target is for Union Pacific. It knows what his biases are. It knows
[01:30:34] the three reasons why he's bullish on this company. And as such, as he's late, the agent is able to go through all his historical research, recognize what his views are, and be able to push ideas
[01:30:52] directly to Walter. And as such you can see a framework where Aiden will tell him based on the based on the the most recent operational metrics. This is actually Union Pacific is
[01:31:08] actually guiding at a pace that's x% faster in terms of delivery versus what your current thesis is. You may want to think about upgrading the stock and the reverse as well. So that's number one. We have a persona that understands
[01:31:24] Walter because of all his historical publish published research. The second where it cross references is we can go into the prompt bar and we can have the Aiden cross reference between the current transcript and the transcript from last quarter. So we can
[01:31:41] put in here look at what the CEO is saying in terms of guidance. tell me how that's changed and also how is the sentiment how has his tone changed since last quarter. So this is where we're able to have Aiden
[01:31:57] look across different transcripts and give you more qualitative details as well as a quantitative as well. Now with that you're able to decipher deeper details faster than the competition
[01:32:13] and as such you're able to communicate those uh those details to clients uh at a deeper level. Now if we take you into the next module and we'll go through this one quickly. The third thing that typically happens after um after a press release comes out
[01:32:29] and then the earnings call is what we call an analyst call back. So, the management team sets time with each analyst and they'll have a one-on-one call with you. And the reason they do that is so that you can go deeper on any clarification.
[01:32:45] You can get a refresh on some of the numbers that they mentioned that you you may have missed. This is where we introduce Aiden noteaker for you. a virtual notetaker that sits on your calls, Microsoft Teams, WebEx, Zoom, and
[01:33:00] as such, it's able to alleviate the need for you to take detailed notes. Instead, you're able to focus deeper on the conversation with the CEO, perhaps challenging the CEO and clarifying some of those numbers. So, with this, what
[01:33:16] you see is you get a pop-up on the right hand side taking you into the virtual call. You have Walter Sprackcklin on the call. You have the CEO, you have the CFO, and you have Aiden noteaker as well. At the conclusion of this call, you immediately
[01:33:32] momentarily get sent to you an email that highlights the key summaries, key takeaways, attendees on the call, who said what, key deliverables. This sits in your email, but this also
[01:33:47] is fed into Aiden, so further in the future, you can always go back and prompt against it. Now for the last module that we'll take you into and as I spent 20 years as a research analyst I can tell you that updating financial models is the most
[01:34:04] ownorous part of any research analyst's uh workflow. During earnings, what a research analyst would do is they would spend hours scouring through PDFs, whether it's a press release, whether it's a 10K, 10 Q,
[01:34:19] 8K, uh, corporate presentations, looking for single numbers so that you can manually punch that, hardcode it into your financial model. Some of these financial models are 100 lines long,
[01:34:35] some of them are 200 lines long. What we've engineered is we've been able to take hours of work and streamline this into minutes. So why don't we take you into Walter Franklin's uh financial model for uh Union Pacific that's of
[01:34:50] course scrubbed for for demonstration purposes. Now what you see here is column K is entirely blank. Everything to the left of column K are the his historicals. To the right are his are his forecasts. This is where align with what Bobby said
[01:35:06] about making um AI invisible. We're meeting you where you work directly in Excel. So from here what we do is click on the AEN icon on the top right hand corner. Here's where we can just prompt a natural language update the latest
[01:35:22] quarter with real time data. Hit go. And what you'll see is it's thinking here. This is where it's pulling data from the 10K, 10Q, 8K, CAP IQ corporate presentations. And then within moments, what you're going to see is that entire row has just been
[01:35:39] populated. Your financial model, which used to take you hours to complete, is now done for you. Now what we know what we believe to be absolutely critical as we're
[01:35:54] building AI product in order to drive adoption we need to build trust with the tool as we hover over any single cell here what you have is full audit trails full sourcing and as you can see from here as
[01:36:09] we hover over a cell it shows that it took this number from the 8K. If you wanted to go deeper, you could press on that blue button and it would take you directly to the PDF so you know exactly where that number came from. Now, the last module that I'll show you
[01:36:25] exceptionally quickly also in Excel is we're going to make we're going to challenge Excel a little bit more. And what we're going to say is imagine a scenario where you just got off the call with the CEO of Union Pacific, your one-on-one call, or during the earnings call, they said, "We're going into
[01:36:41] acquisition mode. We're looking at targets over the next several quarters. Here is where you can make Aiden4 XL an ideation partner. In this scenario, what we're doing here is we are saying, let's build a merger model. Let's have Union Pacific buy a
[01:36:59] fictitious company called Railco, another rail company. Let's have Union Pacific pay a 25% premium to yesterday's close. At the same time, let's make it a 50% stock, 50% equity deal.
[01:37:15] Lastly, let's make this uh price all the data as of yesterday's close. And as we click generate, I'll just pause for a moment and we can all see the numbers magically populate on screen
[01:37:36] as it builds this out. What you're seeing is it's going through an accretion dilution analysis and you can run multiple iterations of this very quickly because historically it would take research analysts hours to do something like this and what you're seeing is we again have t taken hours
[01:37:53] into minutes. Now, as we begin to wrap up, what I'll leave you with is what we just showed you is the Aiden research workflow. I'm going to finish this conversation exactly where I started in
[01:38:10] the sense that three years ago at the with the strategic vision from our CEO Derek Nelner as well as Bobby, we started our AI journey in capital markets research. In the years since, we have subsequently
[01:38:27] built out the capabilities across our investment bank, global investment banking, global corporate banking, we're going sales, trading, operations, our ambitious goal is to be the digital investment bank of the future. And while Derek and Bobby
[01:38:44] made big bets on us several years ago in terms of building out Aiden research, this is just simply what we do at RBC. We make big bets on AI. We make big bets on our people. We make big bets on our culture. And finally,
[01:39:01] alongside data bricks, we make big bets on our strategic partners. And together, this is how we're able to dig deeper, bring deeper insights, datadriven intelligence, so that we can help our
[01:39:16] clients make smarter, higher conviction investment decisions faster than ever before. Thank you everybody.

Learn more about the Databricks Data and AI platform.