Scaling Responsible AI: Autonomous Agents in Banking and Payments

Summary

Banking and payments are entering an agentic era where autonomous AI agents act as primary customers, collapsing traditional switching costs and making data foundations, regulatory licenses, and redesigned processes the key competitive advantages for banks.
FIS unified fragmented data systems with Anthropic-powered agents, Vantage Bank built the Hazel Network for tokenized deposits with compliance-driven observability, and Santander deployed agentic commerce for AI-initiated payments on Databricks.
Santander's Getnet subsidiary also implemented multi-agent systems for autonomous month-end close, demonstrating production-scale responsible AI with governance, lineage, and multi-agent orchestration on the Databricks Data and AI platform.

Scaling Responsible AI: Autonomous Agents in Banking and Payments

Banking and payments are transforming as AI agents become primary customers. When autonomous shopping agents optimize transactions in real time, traditional switching costs collapse. Banks and payments companies that own transaction data, regulatory licenses, and redesigned processes can win. Leading institutions including FIS, Acxiom, Vantage Bank, and Santander are accelerating innovation by combining data foundations with conversational AI to unlock new products and efficiency at scale.
Learn how to architect responsible AI for banking using Databricks. See how FIS unified fragmented data systems with Anthropic agents, how Vantage Bank built the Hazel Network for tokenized deposits with compliance-driven observability, and how Santander deployed agentic commerce for AI-initiated payments and multi-agent systems for autonomous month-end close. this video covers governance, lineage, multi-agent orchestration, and responsible AI at production scale.
🤝

Chapters

FAQs

What are the three waves of AI adoption described in this video for banking?

This video describes three waves: the prompting era, where AI assists individuals with efficiency tasks; the agentic era, where agents are deployed in service of workflows with humans still in the loop; and an autonomous era, where AI agents act as independent actors initiating transactions and managing complex financial processes.

How did Vantage Bank use Databricks to build the Hazel Network?

Vantage Bank built the Hazel Network for tokenized deposits on the Databricks Data and AI platform, with a compliance-driven observability layer to meet regulatory requirements. The Hazel Network uses a one-token, two-legal-identities architecture and includes a partnership with Custodia to enable a broader deposit network vision.

What is agentic commerce and how is Santander using it?

Agentic commerce refers to AI-initiated payments and transactions triggered autonomously by agents rather than human users, enabling commerce to happen in real time as agents optimize on behalf of their principals. Santander is deploying agentic commerce alongside multi-agent systems at its Getnet subsidiary for autonomous month-end close operations.

Why do banks have a structural advantage in the era of autonomous AI agents?

Banks own three critical ingredients that are difficult for competitors to replicate: transaction data that provides deep insight into customer behavior, regulatory licenses that govern permissible financial activities, and established business processes that can be redesigned around agentic workflows. This combination positions banks to lead the autonomous era if they build the right data and AI foundations.

Full transcript

[00:07] Today, we are going to be talking about scaling responsible AI in banking and payments. And so, we hope that you enjoy the programming. Um it's going to be a really good mix of a lot of conversations about what's hot and what's new from our go-to-market specialist. Um then we'll segue into a fireside chat with Vantage Bank,
[00:24] followed by a panel with the Santander Group, which I'm super excited to have because they have literally crisscrossed the globe to come here for us. Um and then we're going to close with a couple of goodies and takeaways for you. So, without further ado, I'd love to welcome my esteemed colleague Jennifer
[00:40] Miller to the stage. Um she is the go-to-market lead um for banking and payments. Give it up for Jennifer. Hello. Hi everybody. Awesome to be here with you. Do I have a clicker? Keep going. Okay, let's see. We're going
[00:55] here. Um Let's see. So, terrific to be here with you all. This is such a full circle moment for me cuz I literally sat in the chair in that back corner 1 year ago and I was making a massive life decision. I was running an almost $2 billion business, a skip level from the CEO at a
[01:11] top five bank, and I was deciding whether I wanted to leave that role and join Databricks in this capacity. And I did that for one reason. And the reason is is because I knew that this industry was about to undergo a massive transformation, and I knew Databricks
[01:28] would play a key role, and I wanted to have the opportunity to guide the industry through what is the next chapter of banking and payments. So, what what I'm going to do is talk to you about I've I've crisscrossed the globe, too, over the last 12 months. I've uh spent many hours with customers hearing
[01:43] about what they're doing, what they're thinking about, and I'm going to talk to y'all about what I think is the the future of banking and payments before we jump into some customer conversations. All right. So, first, the future of banking and payments is not agents or humans. It is human-led and agent-operated, and
[01:58] it comes to us in three waves. So, wave one is what we are all living and breathing every single day, the prompting era. We all use We all use AI agents to make our lives easier, more efficient, more effective, better, make the things that we do run faster and more smoothly. Where we're embarking on
[02:14] today is the uh is the agentic era, whereby we deploy agents in service of workflows, but agents are still very much in human-in-the-loop executing the workflow, making sure that those agents are doing the right thing before it actually executes on behalf of the need.
[02:31] Where we're headed towards, we are headed towards the autonomous era. And in the autonomous era, humans set goals and agents act. Like, I'll give you an example as a former banker, um the the the way that I think about this is uh in the future, we'll see a credit
[02:46] portfolio, a credit risk person, some some role like that within a bank set parameters for their credit portfolio. Here is my credit policy, here are my pricing restrictions, here is uh what I'm looking for in terms of yield, etc. And agents will autonomously underwrite
[03:03] loans on behalf of that, and agents will rebalance the credit portfolio on behalf of that, right? So, think about this world where agents are acting within the guardrails that are set by the humans, but they're acting autonomously in service in service of what that looks like.
[03:19] Now, I bet everyone in this room is either thinking about wave one or wave two, but what your customers are thinking about is wave three. Right? So, think about when Amazon launched the the uh point-and-click one-point purchase experience,
[03:35] that changed how every customer in the world thinks about shopping, and it changed their expectations on how they view shopping. AI is the new Amazon effect. Right? So, the minute customers start to experience autonomous experiences that are intelligent and predictive, and that
[03:52] work for them without having to act, your customers will start to expect that from the bank or payments company that you operate. Right? So, if this is the future if this is a future, what does that mean for the industry? Well, number one, agents become our new
[04:07] customers. Now, the good news and the bad news about agents is they don't sleep. So, agents will 24 hours a day be optimizing for their human. They will be looking for the lowest credit. They will be uh searching for the highest deposit. They will be routing
[04:24] transactions to the highest rewards cost, the highest rewards value at the lowest cost. Right? And because they can instantly deliver and shop, and they can automatically execute, switching costs will evaporate. Right? And that's the primary way that
[04:39] banks and payments companies have made money to this date. Uh so, you're like, "Well, jeez, that's that's like a little it's a little dark, right?" I think ultimately the reliance on lazy customers will come to an end for the industry, but
[04:55] I think the great news is is that banks and payments companies have all the ingredients, they own them today, to win in the next era. They own the action layer today. Every bank here has a regulatory license to move money, issue debt, take in deposits. Right?
[05:11] They own that action layer, that will become the new battleground of innovation. Uh they also own the transaction data. You all own the transaction data unless your customer self-permissions to give that to someone. That will become the weapon in which you will win in that new in that new battleground, in that action
[05:27] layer. Right? And then thirdly, there are a lot of processes at banks and payments companies, and those processes can be radically redesigned to take out costs, right? So, I really think about this next chapter for the industry, the AI revolution for this industry equals
[05:44] the innovation revolution for banking. We are about to see the next chapter of reimagination for this industry and that requires the right foundation and the right set of tools to do that safely and smartly for your organization. Now, what's really cool is we're already
[06:00] seeing this at Databricks, right? The leaders are already well under their way. So, if you look at the data here in in 2024 versus 2026, we saw a 600% increase in the use cases that are running on Databricks that are focused on driving growth
[06:16] for that organization. So, growth is underway. The AI revolution and innovation is underway in banking and payments. And speaking of that, I'm going to invite two of my friends up here that are leading the charge on innovation for their firm. Both of these two firms serve the banking and payments
[06:33] industry. They see where this industry is headed and both of them are innovating as we speak and I want them to share their story. So, please welcome Himal and Greg. Himal leads corporate strategy and AI for FIS and Greg and Greg from Axiom leads their partner
[06:49] ecosystem and growth. So, Yes. So, Himal, I'm going to start with you. So, you your company operates like in the heart of banking and payments. You serve the banking and payments industry,
[07:04] thousands of customers across the globe. And you recently announced a really big partnership, bold move with Anthropic where you are building agents that are industry-specific, which is super cool. Can you talk about that strategy, how you're thinking about
[07:20] it and how Databricks plays plays a role in helping to make that successful for you and your customers? Yes, absolutely. So, a little bit more color on um FIS. So, as Jennifer correctly said, we are critical infrastructure that serves over 10,000 banks globally and we sit across the entire money life cycle. So,
[07:36] what we mean by that is where the money is at rest in a deposit account, in motion on some sort of money movement mechanism, or at work in some sort of investment account, trading account. FIS is critical infrastructure, uh supports that for all banks and
[07:51] financial institutions over the world. So, we've been around for 60 years. And 42 acquisitions later, we were facing a world-class problem, which needed a world-class solution. So, disparate systems, payments data here, wealth data somewhere else, you know, no
[08:08] single view of the customer, different latency. Uh so, we've been on this journey with Databricks for a couple of years now. And, you know, as the saying goes, steady steady surely and then suddenly, uh we saw a lot of traction from using the the great platform that Databricks
[08:25] underpins at FIS. And we were able to get customer data hydrated into the environment um at the account level, um governed with all the right access and permissions and everything like that, which really unlocked the ability for us to serve our customers with AI
[08:42] earlier this year. So, we announced this at our um you know, our annual uh you know, client event earlier this year. And since then, in the last 8 weeks alone, we've had over 50 customers commit resources to this project um where they're going to get, you know, a
[08:58] Databricks instance uh from FIS. They'll be able to see their own data back in a way that is consistent in how they want to see it with all the tools and bells and whistles. And really that underpins uh the agent that we're launching with Anthropic. So, without that, I don't know how that would have been possible.
[09:14] It's really the context layer. So, you know, all of our data in that system, which is the agent, is going to read and uh use as as as as a mechanism to deliver the deliver the end outcome. And FIS builds all the harnesses around it. So, it's been an incredible partnership. We are a very happy customer, and the
[09:30] use cases are growing by the day. And we have a wonderful pipeline together. Uh so, So it's been amazing to see all of this come together in the last few months. I love that. And I love that you see the problem, you see what's on the horizon, and you're attacking it with your customers. Congratulations, it's incredible, amazing. Um Greg, pivoting
[09:47] to you, the autonomous era promises a change in how we think about marketing. I was at a bank for a long time, and we would run campaigns that would take 12 weeks to ship, and we would wait 90 days to see the results, right? I think that that is changing, and Acxiom is on the cusp of that. Tell us how you're
[10:03] supporting the marketing today. Jennifer, excited to be here. Uh yeah, we're going to talk about marketing here. So, marketing being this function that is important to one, identify a a new client, acquire a new client, and then also to
[10:19] retain a new customer. So, who here thinks that that's not an important function? Anyone? Anyone? And then, raise your hand if you know who your CMO is. And I'm that's that's Chief Marketing Officer, just so everyone knows.
[10:37] But that's an important um it's an important uh uh function within an enterprise. All of your bank All of your All of you have marketing that's critical. Now, Acxiom, we've been this marketing partner for over 40 years, actually about 50 years.
[10:53] So, we're up there with FI Not 60, but and many of the large banks, like nine of the top 10 banks, trust Acxiom to run their marketing today. So, what we've done though is we realized like the future is going to what we've taken Acxiom, all the things that we've
[11:09] learned over these decades, and now have embedded that into Databricks. So, all of our core capabilities are now in Databricks for you and everyone here to have access to. So, we're excited to announce this new capability. It's part of the customer lake that was discussed
[11:26] this morning from Ali. More More on that tomorrow. Um but as you think think this, you think about marketing, we want to help all of you get to this layer of agentic marketing, really to transform marketing um and help you put this agentic capability that sits on top of
[11:42] all of your customer data, and that's going to be critical for you to get to this stage of what we would what clients really want to get to, what you want to get to is trigger-based marketing in the future. So, a couple other things that are are critical here as you think about
[11:59] marketing and you think about the ability to measure marketing. Right? Think about measurement and the ability to measure marketing. So, anyone that knows anything about marketing, there's there's a lot of things that happen with your customers in the paid media ecosystem, and there's a lot that
[12:15] happens in your owned media, like on your website. You have a very difficult challenge to measure that today, to really know what is happening with all of your marketing dollars. So, we're able to unlock that insight now by having all by having Axiom embedded into Databricks, you're
[12:32] able to then use this to create this layer that allows you to put this measurement genie layer on top of it. So, we're very excited about that. Um data collaboration is another new uh area that's happening as we speak. You want to take advantage of your partner
[12:47] data. We've embedded Axiom into the data collaboration capabilities of Databricks, so excited about that. And lastly, we've looked at use cases that are really important, and we focused on two, uh one around acquisition and one around
[13:03] retention. So, on the acquisition side, we've got uh a new app that we've built that we call, and this is we're going to introduce a new app called Deposits IQ. I'm going to let this run, but we've taken the ability to look at how you can acquire clients that you
[13:21] want to drive deposit accounts, and we've built this capability into Databricks. And this is available. So, right? So, if you were interested in this, we can talk to you about it. We have a booth here. But, this this is really just showcases a lot of the capabilities that we can
[13:37] help unlock. Like I said at the beginning, by having the Axiom identity and data capabilities built into Databricks. And we're excited to be here and excited that we have this opportunity to work more closely with you, Jennifer. Thank you. Awesome. All right, thank you guys.
[13:54] What I love about what we just heard, is these guys and their firms aren't waiting. They're not waiting for that autonomous era to come, right? They're acting now. They're building new solutions now to ready the industry for that, which is awesome. Um all right, so
[14:09] I wouldn't be a good good go-to-market leader if I didn't leave y'all with a call to action. Here's what I would ask you to do. When you leave, take these three questions back to your firms. They're really important. And getting a yes to every one of these questions will enable you all to win in that autonomous era. So, number one, does my AI
[14:27] understand my data and my company? We heard from Ali this morning the importance of context. This is mission-critical. Without context, your AI is dumb. And no one wants to spend money on dumb AI, right? Make sure that your AI is reasoning over context- rich
[14:42] enterprise data. Two, can I manage the cost and risks of my data and AI? We all work within regulated industries. It's really important that you can look your regulators in the eye and give them end-to-end lineage and give them confidence that your agents are doing the right thing for your customers.
[14:58] Databricks helps to do that. You need that control. And third, how much freedom do I have to change course as the industry evolves? The ecosystem is moving every single day, almost hourly. Make sure that your firm is protecting your future and allowing you
[15:14] guys to have the ultimate flexibility to bring in new models, bring in new capabilities as the industry evolves. I'm super pumped about where the industry is headed. I I'm really thrilled about being on this journey with each one of you and let's go build the next chapter of banking and
[15:29] payments. Should we? Thank you all. Right now, next I'd love to welcome to the stage Jason Ferrante. He's our senior director of banking and payments.
[15:44] So he works across all types of accounts and really sees kind of everything and we've got Vantage Bank Joel Castaneda who's going to come and share a little bit more of perspectives of what it's like working with Data Bricks and some of the new technologies that we have.
[16:00] Welcome them to the stage.
[16:16] Perfect. We're in business. Awesome. Well, Joel really excited to have you here. Happy to be. Thanks for coming to talk with us. So Joel's the chief risk officer at Vantage Bank and Joel as I mentioned we're really excited to have you share your story and kind of understand what's allowing you guys at Vantage to really
[16:32] out innovate a lot of your larger peers within the space. So if you could start with kind of educating everyone on first who's who is Vantage Bank and how has your thought process on technology evolved throughout the years as well. absolutely. So thank you for allowing me
[16:47] to be here and talk about Vantage and what we're building and and why we're so innovative. But Vantage is a Texas based community bank. We're about 5 billion in assets. And when I say community bank that means we're embedded in our communities. We know our customer we have boots on the ground, but it's really managed by the data that we have
[17:02] to and that's part of the innovation that we'll talk about here. Awesome. That's great. And yeah, being a 100% family and employee owned bank I thought that was really interesting. So like how when you look at the next the next stage and we were prepping for this we spoke at kind of how you were thinking about
[17:19] the evolution of technology and specifically when you look at blockchain. And there's kind of a defining shift when you were looking at, okay, this went from an interesting technology to actually a strategic threat to the overall business. Could you speak to a little bit on what drove that? Sure. So, I'll I'll start with the
[17:34] family-owned part because we don't have quarterly earnings to meet, right? We can invest for the long term. We can have long-term strategic decisions. And it was about 6 years ago when we were sitting around a table with our CEO, our CFO, our operations, our chief business
[17:49] architect and saying, okay, what what's next as far as innovation where we need to be to serve our customers. What I didn't mention early on is that about 30% of our customers are foreign nationals. So, we have a very large revenue center allocated to you foreign exchange. It's really, really important
[18:04] to our our line of business and serving our customers. So, with that said, we thought, okay, next 5 6 years we're seeing a lot about digital assets, about blockchain, what's going on with stablecoins. It was just emerging. We need to be in that space or at least adjacent to that space
[18:20] understanding that that could potentially disrupt our foreign exchange line of business, which I said is substantial to us. So, that really started the origin of, hey, we can think long term, we can invest long term. What do we need to do in relation to blockchain and payments and that started
[18:36] the ideation on what now is our Hazel network. So, Hazel network is a network of deposit banks that we can trade deposits between customers on Ethereum. So, it's a public permissionless blockchain and we've been
[18:52] promoting that. The interesting thing is that we have a single token. So, Hazel token is can become a stablecoin. One token, two legal entity treatments. So, really interesting what we're building. Yeah, absolutely. And if you could help just educate the audience the difference
[19:09] on why kind of one token with two different identities is so important when you look at kind of the risk that stablecoin kind of drove for Vantage Bank versus the tokenized deposits and what the traditional banks were looking at. Yeah, so I'm I'm glad you brought up risk because one of the biggest risks that we see is is compliance. And the
[19:25] reason that we chose Databricks and a unified data layer is we have to comply and as a chief risk officer sitting up here, I need to give assurance to Jennifer talked about it, regulators, right? That we have a compliance program. And so we need the data to do that. So that's really the foundation of what we built with Hazel. But the
[19:41] interesting thing is one token is in the bank deposit network which we're currently building, it is an FDIC insured deposit, just like any other deposit in a bank except it's on the blockchain. Outside that same token, once it leaves
[19:57] the network, becomes a stablecoin. So it can go anywhere. Ultimately, we want it to come back into the deposit network because as a community bank, those deposits, as y'all know, fund our lending. And that lending funds economic prosperity in communities. So really
[20:12] absolutely critical and important. Yeah, and when we were prepping, one stat that really stood out to me, I can't remember if it was one of the documents you provided, I believe it was though, 88% of dollars that would leave a bank to go to stablecoin wouldn't come back. So that's why they should be so important. Yeah. You know, it's possible. Exactly
[20:27] what we're trying to prevent is deposits leaving into stablecoin and never coming back into the banking system. And again, that's that's our fuel. Deposits are our fuel to go lend out in the community. That's absolutely critical mission of what we do. Got it. So would it be fair to say that really governance and security is what made you feel comfortable with kind of
[20:44] moving from that wait-and-see period to where you're okay, we have to really evolve on how we're looking Yeah, absolutely. Go going back to to data, governance is absolutely important over that data. Again, why we chose a unified data platform in Databricks. We know that everything that we're doing on chain and and from our ledger is flowing
[21:02] into Databricks. We can see it. It's it's visible. It's governance clean data. So that was absolutely critical. My job is to question things and ultimately people think I say no, right, as a chief risk officer, but I saw the strategic threat. It was too big. We had the data infrastructure to build on top
[21:18] of it and we were really excited when we had this opportunity to build with our partner. Awesome. And then something else that came up is when you're bringing Hazel to market is your partnership with Custodia. Yep. Um, so I believe that your CEO Jeff was ranked number one most innovative person in
[21:34] finance for American Bank and then Custodia's CEO Caitlin was ranked number two. So, could you speak to first why the partnership is so important on what you're delivering to the market? Yeah, so early on that partnership became really important. We were like I mentioned in the beginning a traditional commercial bank and Custodia Bank is a
[21:51] digital asset bank chartered out of Wyoming. So, we have a banking license, they have a banking license, but they're very different. They were built for digital assets, we were built for fiat money movement. Together very like-minded CEOs came up with this idea and our our CEO Jeff said, "Tear down walls. We don't want Hazel network to
[22:08] become the next Zelle where everyone pays a tribute, it's expensive to get in, your customers don't pay for it." So, we wanted Hazel to be accretive to banks and the way we do that is each bank in the network shares on the yield on treasuries that are held at Custodia
[22:23] against the stable coins. Got it. Super interesting. Thank you. Perfect. All right. So, when we look at what's next, what's what's the next evolution? Yeah, so two things. Number one, building the network. We really need that network effect to make this work. So, getting banks in our network and we
[22:38] had to make it easy for banks. The second thing is and tying to making it easy for banks is we're going to be building our risk and compliance apps on top of Data Bricks. We already have two prototypes. We built one prototype related to analyzing a bank's financial health, taking, you know, public call
[22:54] report UBPR data and turning that into a scoring system. We built that, all that data flows into Data Bricks and we have a way to evaluate banks that may join the network. We want our network to be a network of healthy banks that are in it for the long term. So, that's scoring. The next one is building our
[23:09] liquidity management engine and our AML platform on top of Databricks using Databricks apps. So, network building and app building is the next evolution. Great. Well, awesome. Joel, really appreciate the time and sharing how you're evolving with the Hazel network and how you leverage Databricks to do
[23:25] so. Happy to be here. Thank you. I'd love to welcome on stage Jennifer Miller who's going to introduce our guests for us and we'll get this show on the road.
[23:41] Come on up, friends. Let me see if I can. Yeah, thrilled about this one cuz it's super fun to the juxtaposition of hearing from a community bank that's just killing it and doing amazing innovative things. And now a big global bank that's doing the same. So, y'all thank you so much for being here.
[23:56] Excited for the conversation. I'll stay if that's okay. All right. Um, let's jump in. Shall we? All right. So, y'all are part of Santander. You each bring a very different angle to this whole AI conversation with your seats within the firm. So, can you just start
[24:12] with introducing yourself first and talk briefly about your role and how you fit into the Santander organization? Excellent. So, very happy to be here. Thank you, Jennifer. I'm flying from Spain. So, so excited to be in San Francisco
[24:27] where innovation happens. I'm the CFO and head of a strategy of Santander Payment Solutions which I will explain in a second and I'm also double hatting with a CFO role at Getnet. So, when I travel AI and looking forward, I have the dual view
[24:43] around what we can do for our business as well as what we can do to do better in finance. Amazing. Hello everyone. Everyone, I'm Leila Curachi. I'm leading the CTO organization in Brazil and along with the business
[25:00] we are developing a lot of value out of the data that we hold from our customers and making the best to deliver services and a better journey for them.
[25:17] And I'm Bruno Peronetti. I'm the global head of finance transformation at Santander Payment Solutions. Uh my goal is actually to make the finance professionals uh stay close to the business and business partner with the different parts of the business and being closer to the customers and not uh
[25:33] super close to the spreadsheets as they usually do. Bankers love their spreadsheets. All right, um so awesome. So most people know who Santander is, but Christine, can you talk a bit about Santander Payment Solutions and Getnet? How that fits into the organization, why payments
[25:49] is an important part of your strategy, and how you're thinking about AI as evolving that. Excellent, thank you. So So I would like to start saying so Santander is 170 years old. Um and AI is actually a game-changer for us
[26:04] as well. Uh we have uh the this recently presented our investor day with ambitious target and AI will be an enabler. We have been working for the last few years on creating the data foundations and governance that is key for AI innovation.
[26:20] We anticipate about 1 billion euro of combined new revenue and cost savings over next three years on the back of AI. Um Within Santander we have five global business divisions, being payments one
[26:36] of them. Um payments is one of the five that is going to be most subject to disruption. I think you mentioned before, Jennifer, that actually the customer is shifting from the human to the agent and that's what we're focused on. Um at Santander Payment
[26:52] Solutions we have we process 30 billion transactions every year. That's growing double digit. We are present on issuing, acquiring, and payment processing uh, across the US, uh, LatAm, and Europe. Um, so when I think about our strategy
[27:08] in payments, uh, we have one piece, which is the agentic commerce, that we can elaborate. Uh, clearly we see, uh, huge change, uh, for the merchants and what we're focused on. How we can help merchants to accept those transactions that are going to be initiated by agents. Last week we actually, uh,
[27:25] published we're the first company in Latin America to accept an AI-initiated, an agent-initiated payment. Woo. Um, yes, thank you. Um, um, but but AI goes beyond that. What means AI in payments? AI in payments means
[27:42] customer 360, connecting all the different information we have around transaction, contracts, and so forth for our customers to do the best. It's about, um, smart routing of our transactions to increase approval rates. It's about new product development on the back of software. So payment
[27:58] companies are software companies in the end, so we can go much faster on software development. And it's also about how we do better in our operating model. And this is where finance comes into place and we can do much better as a company. That's amazing. And by the way, isn't it so cool to have a CFO talking about AI?
[28:14] Like, give it up. That's amazing. Thank you. Love it. So good. I love it. Love it. Um, all right, Leila, as you think about, uh, just data and AI within Santander from the CDO's perspective, what has to be right from a governance
[28:29] lineage standpoint to deploy responsible AI at a bank? Uh, first of all, we have to have the basics done. That's why we started to work on the data transformation program about 4 years ago in Brazil, so that we
[28:44] could, uh, enable the capabilities and all the data products that would enable Christian and Bruno here to work with AI solutions to accelerate growth at our markets, right? So, what we did was to
[29:02] provide context. We organized all of our data. As you can imagine, we are old. So, we have a lot of legacy systems. Bringing all together it's not an easy task. Uh, so that you have in mind, we moved
[29:18] about 16 petabytes petabytes of data uh, to the cloud. Uh, and doing this it's not an easy job because we have to get the right context. We are very regulated. So, we have to have all catalog and we have the
[29:34] costs to take into consideration to do it efficiently and to serve the business at the best interest of our customers. Amazing. 157-year-old bank and you're fully on the cloud. That's remark- I mean, that's remarkable. That's incredible.
[29:49] Congratulations. All right, so Christian, you talked about making using your finance function as a catalyst for driving an AI agenda within Santander. Can you talk a bit about your strategy and how you're thinking about AI, automating workflows that are labor more labor intensive, and driving more
[30:05] efficiency within your organization? Excellent. Yes. I think we can move on. Uh, so so we let me first introduce a bit about Getnet. So, um, Getnet is the number one payment company in Latam. It's
[30:20] uh, processing 260 billion dollar transactions. Uh, it's present in across 10 different countries with payment licenses in Europe and Latam. Um, and more than 1 million customers from the small SMB up to the large global
[30:37] accounts, okay? Getnet is moving, is changing, is transforming from what is the local leader in the countries where it operates into a global payment leader. This means a huge change in the operating model and
[30:53] we in finance wanted to be on the same place. We in finance wanted to be at the forefront of that transformation. Um the challenge that we face is think about every day, every month reconciling different accounting standards,
[31:10] different reporting processes, uh different currencies all up into one global financial result of the company. Um and that uh that needs to completely change. Uh I want the finance team to be closer
[31:25] to the business strategy, to be closer to the execution, but that doesn't happen today. What ha- what what used to happen is teams they spend a lot of time on manual processes. I see Excel files moving around, 200 megabytes heavy Excel files that then take a few minutes to download
[31:41] and to open in the computer and then to go through the different lines of the Excel file uh to check whether the number is right or wrong. Okay? That uh it's the reality combined with different legacy systems in every country with different ERPs that make this huge
[31:57] complexity around uh change in the operating model. And that's what we're doing. We're changing the operating model. How are doing that? We are envisioning to have a system, a multi-agent system that will reconcile data is reconciling
[32:13] data uh and validating uh in an autonomous way. Uh it's uh helping uh to escalate the uh pro- the scope of the exceptions and have always a human supervisor on top. It's allowing uh the system to the
[32:29] finance teams to interact directly with the data and the results that I see and it's creating a common ground for all these countries and the global team to interact. So, that's what we're doing and it's going to change the way we operate. Amazing. So, accelerated month-end close, no longer waiting 15 to 20 days
[32:45] after the month has ended to know how your previous month performed. Incredible. Um Bruno, can you talk a little bit about Christian set this ambition, this is what I'm aspiring to do. Can you talk about the work? How did you actually tackle that? Well, um when Christian actually came to me and
[33:01] said, "Look, I want this to be automated and um we need to work as one finance team." I was like, "Okay, oh my goodness, let me see where is the first place that I should tackle?" And looking at the activities from the finance department,
[33:17] one of the most tedious and time-consuming activities is the month-end close because every single finance professional is actually involved in the month-end close, from accounts payable to accounts receivable to accounting to FP&A. That means a lot of hours, right? So, we
[33:34] were like, "How can we do that?" So, we thought of a platform that could put all the data together and make sure that uh we would be able to access the data super easy. And then, that's when Databricks actually came into the picture. So, uh
[33:49] we connected all of our data that are in databases, like the data from the ERPs, the planning systems, uh everything related to the transactional uh systems into Databricks, put there in a bronze layer,
[34:04] and then we said, "Look, we don't want to take away the opportunity for people to actually load uh Excels and other files uh into there." So, we created a front end where people could just drag and drop those files and
[34:20] uh have that information in there as well. And that's where the magic starts. We created analyst agents that analyze all of that data and make sure that we bring that data from bronze to gold so we can consume that and we can
[34:36] accelerate everything that has to do with finance, right? Uh that was amazing. We already saw that we could reconcile the data, we could understand uh the variances uh that uh the the data was super well structured
[34:52] and was at the tip of our fingers super fast. But what was really tedious in the month-end close was not organizing the data. It was actually coordinating the whole thing, talking to all the departments, making sure that all the information got there on time. So, what
[35:09] did we do? We created an orchestrator agent that manages the whole month-end close. This orchestrator orchestrator agent, what it does, it talks to the analyst agent and has in a Delta table everything that has to do with the
[35:25] month-end close, controls, policies, all of the information that we need when it comes to the strategy of the business. And then it pops up prescriptively the information that we actually need for the business. And this helps us quite a lot because uh the month-end close, it's
[35:42] not just shortened, but it actually gives insights to the executives at the company. So, having said that, what did we use? Is this magic? No, it's a Delta table well structured. We used agent bricks, we used uh
[35:58] we we used for example Databricks apps. So, all of the tools that you guys already know uh that are available in Databricks. Amazing. So, when you started using Databricks, like what meaningfully changed for you? Well, once we had the data in, what we saw it is that we accelerated uh the way
[36:17] that we were uh doing finance, right? Uh the cool thing is when we started adding the data, it took a while because we had to put a dictionary. Imagine that this that something there is old as Christian and Layla said. So, we are a franchise
[36:32] of companies. You can imagine that we don't have the same ERP in every single country. We don't have the same systems and databases in the same in the in every single country. So, we had to be agnostic. So, we made sure that uh
[36:49] we went flexible and once the data was there, well, we were able to build this multi-agent system in only 5 weeks. 5 weeks. We knew where the data was. We knew how to do it. So, 5 weeks.
[37:06] Uh this accelerates the way we do finance and the way that we manage finance, right? That's amazing. That's amazing. Now, uh Christian, Bruno's team built something incredible and amazing, um but your team had to trust it, right? And what I when I talk to customers, the
[37:22] biggest thing that uh they're grappling with is like, "How do I get my end business user to trust my agent, my work, my automated workflow?" Can you talk a little bit about how you how you uh got your team to trust what was built? Um the the the this is a people management
[37:39] process, um and this is a very good question. Um Before uh the the the person who led transformation was reporting to the person who led most of the finance activities.
[37:54] And it's very challenging for two reasons. One, they have a very heavy business as usual agenda, uh and they don't spend much time looking, you know, 2 years down the road.
[38:10] Um second, what I realized after making the change and bringing Bruno on board directly working with me, is there is a lot of resistance in the uh financial roles. So, we're spending a lot of time. I need to spend a lot of
[38:26] time to sponsor and commit to the change. Uh and and to to get the both sides of the finance team work together. The finance users, let me use this word, who run the finance core activities at P&A, accounting, control, and so forth.
[38:41] And the transformation team, which Bruno leads, okay? We have supported that with a program what we call a finance forward. Which means we're investing in our employees. So, so we bet on our employees. That finance forward program we recently
[38:57] launched, it's 50% data and AI, 25% uh business, and 25% finance. And our team members need all of it. They need to get better on data and AI for sure. But also they need to get better at the business and at the finance function. So, that's what we're doing.
[39:13] Yeah. Amazing. Thank you. Thank you. All right, Leila, back around to you for our last question. Uh so, hearing them describe what they're doing within the finance organization of of Getnet, like how does this align with your your data and AI agenda back in Santander, Brazil? So, uh when we started all the data
[39:29] transformation process, I'm very happy to hear what Christian is saying because uh we really believe to that to scale AI inside the company, we needed to be a business lab federation so that we could actually make the transformation that it
[39:46] not only takes technology to do that. We have to have people on board. We have to have the senior leadership believe in that the investment that they are doing it's worth and we can make it and we can prove it. So, that's how we did it. So,
[40:02] in a considering the size of Santander, Brazil in a very short period, we have built all 35 data domains that we have that translates exactly what the business is. So, no matter how we change the organization,
[40:19] we are always going to be able to deliver the data products that they need, trusted according to the regulations, and we will continue to do so so that we can accelerate more and more the AI agenda with the proper guardrails that we need so that we can
[40:37] scale and also, as Christian mentioned, doing a lot of investment in our people so that they are trained. That was one of the struggles that we had in the very beginning because everybody is used to working on the legacy systems and they
[40:54] are now it's working here, so I don't want to move. And nowadays, we can say that we have about 4,000 users. We tripled the size of users that we have on cloud and they are fully on board to
[41:09] do this journey along with the business. Amazing. Amazing. What I love about these sets of use cases is the world is moving too fast to wait for a 15-minute day close and to wait 2 weeks for an analysis to make a strategic decision. And you guys are tackling that problem. Congratulations. Thank you all so much
[41:25] for being here. I hope you all have a great rest of summit.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.