PayPal Holdings, Inc. ($PYPL)

Okay. Sounds like we're live. Yes. Okay. We're ready. Okay. So thanks, everyone, for joining us, both in the room and online. Super excited to have Srini here from PayPal, who is the Chief Technology Officer. Reason being, I have a little bit of a tech background having run a fintech platform before in the Pay By Bank space. So I think we have some really -- I'm really looking forward to getting some good insight from you. And really, there's so much of PayPal's turnaround story that revolves around the technology. So this is going to be a super topical conversation that not only I think, provides us with good information for now, but good perspective on how the next couple of years are going to go as well. So welcome, Srini. Thank you very much.

Thank you for having me, Adam.

Yes, our pleasure. Our pleasure. So you started at PayPal in 2024.

Yes, 2 years ago.

2 years ago. Does it seem more like more than 2 years or...

Time flies by.

Lots of change going on. Obviously. And so let's start from a 50,000-foot perspective. As CTO, you walked into PayPal what were some of your first impressions about strengths and some potential opportunities?

I think the -- even before I walked into PayPal, when I was intervened for the role, I wanted to make it clear that like we had put all technology under one roof. That had never happened at PayPal. We needed to streamline, right? So that was an important piece and having done this in my previous company, it made lot of sense to really have everything from the infrastructure to the data, all the way to the product engineering to be streamlined. So that was a conscious decision PayPal made, and that was something that was important for this to succeed. So that was the first piece. Once I came in, I was kind of little bit seeing the multiple versions of identity risk and all these platforms, I felt like there were so many snowflakes within the company. So one of my key messages when I started was let's try to standardize. The standardize doesn't mean we need to depend on each other. We need to -- our terminology, which it all loosely decoupled highly aligned. So how do we standardize our systems? How do we build our platform capabilities that we can leverage across all of PayPal. So that's kind of the journey. That was my initial observation coming into PayPal.

Let's start -- let's dive in a little bit to those challenges, right? So I think the prevailing sentiment out there is that PayPal has a considerable amount of technical debt. Whether it's database integrity latency issues, prior acquisitions that weren't fully integrated, et cetera. Is this an accurate perception in your view? Or are there some clarity that you can offer on the internal tech deck, and then we can discuss the external.

Okay. So let me kind of split it into 2 pieces. The first one is what I call the standardization piece, which I mentioned, right? So our identity infrastructure was kind of fragmented or I would not say fragmented. We had multiple versions of identity. And what it meant was you onboard on one product and you try to onboard on a different product would create a lot of friction. You would have -- as if we had omnishare about the customer, and you have to do the same thing again. So that was the same case with the rest and compliance platforms and so forth. So you can collect that debt like we had multiple versions of it, which means we had to build it 3x or we had to maintain it 3x.

Sounds familiar, sounds to me like PTSD.

PTSD, right? So what we did was, so we said let's take the consumer side first, right? And I wanted to prioritize and do it in a consistent way. You look at PayPal, you look at Venmo, right? They are pretty much a similar wallet that has shared underlying infrastructure. So we decided that let's converge identity, and we started on converging identity, between Venmo and PayPal. That took us about 6 to 9 months. And as we finished it, the beautiful benefit out of that was MFA came in, Multifactor Authentication and it was enforced. PayPal at MFA already built in. Venmo did not. If we had not merged it, Venmo when would have to build this complete then again, you could call it dead tech because you're now building it twice. So we were able to just get it up and going really fast here. Now then we have now started doing the payments convergence, where we are bringing in the payment system between PayPal and Venmo to be the same rails. So the beauty of that would be Venmo would now get the better authorization rates. It can go internationally if it needs to go. It now give more credit products that are available on the payment platform. So that was kind of a tech debt that we felt. So I think that's been addressed that way. The other form of tech debt is what you talked about, which is Oracle or databases, right? So -- and that was important because what happens is databases create a center of gravity because it's very chatty, and it cannot be changed very fast for really growing high-speed innovation. So we are now embarking on a journey to completely go cloud native on our PayPal platform. So that's the second way of how we are addressing the technical debt right now. Braintree and Venmo, we have already made it cloud native and they are already up and running in the platform. So now we are now tackling PayPal right now.

So if you get that database integrity solved, that unleashes a lot of flexibility built on top of that. That's the foundation of the house.

That's the foundation of the house, right? And if you are familiar with Oracle databases, you have to tune it, you have to cash it. You have to build a lot of things. 20 years ago when PayPal started, that was cutting-edge technology. We are one of the best users of Oracle. So no doubt about the technology that we use but that's table stakes right now like if you go to any hyperscaler, go to AWS, Microsoft, it's out of the box. So I call it running on the shoulder of the giants. So let's just reuse the platform as a service and let's see the weight on top.

And how long do you feel like the transformation on the database layer will take.

I see it at least it takes a couple of years. The reason is there is stored migration, and then we are also changing the engines on the fly, while we are operating this huge payment network across multiple countries. So the data migration is going to be the one. It takes a while.

That's the tricky part. You can't mess that up.

That's the tricky part. You can't mess that up.

While you're flying the plane. That's -- I feel for you, we went through the same thing and I got an education on that and it ain't easy. It sounds pretty easy when you throw out some technologies, it ain't that easy. Okay. So that kind of summarize -- and then we have -- I'm assuming you have like a bunch of tech debt in terms of the code and how it's inefficient and in some cases rewrite a lot of code.

In some cases, yes, we used to run on C++. It was highly efficient at that point in time because it was a very highly performance system. But now as more and more people move to Java, right? The interesting piece and I'll talk about it when we talk about more AI stuff, we are able to take about 300 of those applications in less than 6 months. In the last 6 months, we have migrated 150 out into Java, complete AI rewrite. So no humans touching it all the way from C++ to Java and putting it in production. So that's been a boon for us, and we see that retired pretty soon.

Let's take a quick detour on that because that's important. I don't want to leave that line of thought. So you take the Java -- sorry, the C++ code and you -- I'm going to talk like a layman here, but you plug it into AI and it comes out with Java code, right? And then do humans look at it to verify...

So we do a review. And then like an AI also actually does the validation dusting. So it's not even just codes. It's a complete SDLC. So we have testings that we have -- that we test our C++ code. So we feed it to AI. It converts it into Java, we run it in the functional test. And if the functional test fails, it goes back to the review cycle that the LLM learns and does it. So it's pretty autonomous. I was pretty -- it was a need then that our team did, like took 2 months for them to build that skills on the engine. And then it just lit a factory.

That's years of condensed work in 2 months. That's important. Okay. So let's switch to external challenges and focus on the merchants. And I think your -- for those of you who don't know, Srini was at a big merchant in Arkansas called Walmart for a while. So wanted to get your view from the merchant side, and that's another reason why I think you're great for this role because you know what it's like and how hard it is for merchants historically with PayPal. So talk about what you're doing on the merchant side to make it easier to integrate with PayPal?

Yes. So our story, like I used to be at StubHub and eBay and PayPal were like -- were part of the same company. And I was the first, I think, merchant implement PayPal as a third-party merchant, right? So I've done integration then and I did this with Walmart. I think the reason I say that is PayPal has had like 3 or 4 versions of our APIs that we have created over that time. We have been kind of benevolent in that like we have never deprecated those APIs, which kind of is good in some ways, but it's bad in some ways because you're really having tech debt from a merchant perspective. You're having merchants on multiple integrations. And I think that was the biggest challenge that we faced. So we were trying to now get into this new checkout and new modernization as we talked about, we learned that. And it was very tough to get on the roadmaps of every merchant to go say, hey, you need to upgrade right now, right? So they have significant roadmaps and deliverables that they need. So I said, let's pivot how about irrespective of what the integration is, we give them the most modern experience. So we spent a lot of time thinking about it, and we've made it so seamless that irrespective of what the integration you are in, you will get the most modern checkout experience that we can give. So that alleviated a little bit pain from the merchant perspective and allowed us to move faster and give new benefits to the merchant.

How did you do that? Was it an embedded strategy in a sense where you're saying, here's one API to link to our system you're...

No, they were still using the old APIs. We were intercepting in the back end. So when it came to us, we took the pain on our side, and that's coming from Walmart customer first, you have to really look from a merchant-first perspective. Looking at it, we said, how do we alleviate the pain. So they will make the same API calls the way they make it, but we will intercept that on the back end. And be direct in the modern system.

It's in back through the existing...

Existing messaging layer to get it done. So that will give you the new pay sheet It will not unlock new capabilities because you're still on old integration. But we have now decoupled that where you can now at least get the new pay sheet without having to worry about that.

And the customer experience on the merchant side is still the same, because the APIs are the same, right?

Same. So seamless...

Have you ever thought of like an embedded strategy where it's a single API, you create a hosted web page on their app or website all of the sausage making is on your side. They don't see anything except a single API integration.

That's the future, right? But even if I had to do that, I have to get them to change it. So first is you don't have to change it, we will make it happen. The second thing that we are now doing is we are using AI, we call it Merchant AI Integration Agent. So what we have done is we have basically created an agent similar to like Claude Bot and this agent, which basically the merchant can download to their site, and they can turn into their LLM of choice, right? Are -- we can connect to our LLM factory. What it does is it goes through the code, it actually changes the code, fixes it, upgrades it to the newer integration, validates it and make sure that it works completely and kind of certifies it, right? So the first few iterations when we tried out we tried out with merchants in Asia, right? It took 2 or 3 tries because it was reasoning itself out. But we built what is called a memory layer. So what happened was every single time, it saw something that it had to fix again, it put it under the memory. So what we are seeing now is as we go through more and more merchants, it's doing it in the first shot because it's now knows enough about it. So that alleviates us to move a lot of the long tail call it Ed, on tail. Ed, the top merchants you have to work with them, they're not going to trust like an automated agent, but torso and the tail we can now do this. And we are doing this to the SIs, Service Integrators to be able to go do the same. So that's the second part, which will allow us to migrate everybody faster. The third part, which is what you referenced, Adam, is how do we build a future-proof integration where you embed something and it will evolve itself rather than having to upgrade over the period of time. So we are now making it more future-proof it will never be completely future-proof, but...

Everything you want to do you change on your end and it's automatically updated to them. They don't have to touch it yes, that's what we did on our platform too and merchants really didn't understand it at first, and then they understood it and then it was smoother. But it still takes a little time to disconnect the old and implement the new couple of agile cycles for sure, right?

That's why with this agent now, it automatically does that, right? So it will take any version of your old integration and move it to the most modern one. So the way we have created the agent is not only now it's integrating and changing your integration, it can also certify. So the idea is to constantly monitor our merchants for the certification and the performance in the future because we can now use the same agent to look at their integration to see, hey, is this working very well. That has been our challenge because...

That goes halfway like you said, right? So they get the latest from your side, still the UX that they have today for consumers . So it's not optimal, but it's getting some of the way getting some of the ways.

But if you really look at the payment interface experience.

Yes. So let's say, take me into the room where you're talking to a merchant. And I am a merchant that likes PayPal and wants to continue to use PayPal and want the updated -- I want to be updated, right? How many sprints is it taking me to do?

So again, what I would say is use this agent, right? So irrespective of the merchant, I would say, use the agent. It will code assist, nowadays code assist and code development is getting there. So use a code assist, it will get you. It will go through. It is in your environment, so you don't have to worry about it, right? What we have seen is when it does it by itself, it does it in less than 2 hours, okay? It goes to the code, it knows the integration pattern, it does that seamlessly. Obviously, then they have merchant as their own functional testing, their production testing and what not to do, that's something that we don't have a control on. But this is what actually -- like we have done it on multiple code bases, right?

So let me just make sure I fully understand you. I'm a merchant an an older version, you say use the agent assist and you can get me on the latest version in a couple of hours, again, not changing my UX for the consumer. Yes. I want to change the UX.

That is a fairly involved complication because it depends on the merchant style because they have to change the checkout. That's much more proprietary their own stuff. It's not our stuff, right? What we are doing is we are seamlessly changing into this. We can provide more stuff with this integration as we can help you personalize, where we can tell you, hey, this person has an affinity to PayPal, if you want to present them better. You want to say smart message in with Buy Now Pay Later stuff. Those kind of things that comes out of the bots with the latest integration, that they can choose to surface, but they still have to do that change that -- I mean there is a clear line between that the merchant property and our property.

That's a bigger lift. Yes. Okay. So what you're saying is we can get you to the updated version with the best identity and the best decisioning and all that kind of stuff and very -- but the UX would be more in the future, okay.

The UX of the merchant. The checkout will be our UX. As soon as you click buy with PayPal, it will be pretty much a single click experience. It will go to the pay sheet and then go back. That part is going to be very streamlined.

Important distinction. So the consumer UX would be the most updated on what you're talking about with the agent, that's important, okay, I misread you there.

So there is the merchant checkout that they have to make the changes. But you click on pay with PayPal or buy with PayPal, that experience that we used to go through, which was like 4, 5 pages long, now it's a single page. One click if we have already seen you. Okay.

That's huge. That's a huge change.

That's a huge change. And we are seeing conversion benefits out of that and merchants, and that's kind of the value proposition that we are -- I would tell the merchant when I go there.

Okay. That's a big change Yes, just leave it at that. Steve's nodding his head for those of you not in the room. Steve is very happy with that answer. Okay. We touched on AI for a second with the agent, but we got to delve into this a little bit more, right. So talk about how you've leveraged it in the prior 12 months, how you see it evolving over the next 12 to 24 months, and then we can get into maybe think about it from a cost reduction perspective, product generation perspective and how you think it can help you drive growth, right? Because I think those are the 3 sites we need to tackle.

So let me start with I'm very proud of the PayPal technology team, right? We were one of the first teams to add up MCP. I'll start with that, right? MCP, as you know, is Model Context Protocol that basically allows to connect to any tools. So if you think about LLM as a brain, it can resend and it can do a lot of things, but then it needs tools to do it. Tools as in I need to connect and update this record. I need to do this API call. So that's what MCP protocol was used for. The reason we were the first to jump in. We were the first remote MCP server, where we put our invoicing out to the public. So that like you could go into Claude and say, hey, invoice this plumber for me for $30 for this. And it will know how to call the PayPal API and like do the invoice and send it out. So that's how we started on actually implementing MCP. So then what I said was this is so powerful. So we set an enterprise standard. We are going to MCP everything internally, whether it's Salesforce, whether it's Workday, whether it's all our tools, that will be an MCP server in front of it. So that the brain, the LLMs can actually do the actions. So that was our first foray. By doing that, people could do workloads that they could not do before, right? So that's how we started on the journey. Then I talked about the C++ migration. We are planning on running checkout across multiple countries. We wanted to run buy now pay later into multiple countries. These are what I call compressible problems. These are rinse and repeat problems. There are some changes that you have to do, but it's predominantly rinse and repeat. So we just said that state code assist and just run through it. That's why we are able to go to 154 countries and check out pretty quickly. That is what I call code assist, where we basically use the code, assist technology to make it happen. Now we are getting to this layer where we can do Jira to code. Basically, our product manager basically use a Jira story and it automatically writes the code. We were able to do it for small task. But recently, what my team has done is we have taken the payments on the checkout infrastructure, and we have created our knowledge layer. It is not just reading core. We have had architects, put knowledge into it to really make it to understand the full flow. And it took us like 3 weeks. Now, you -- with that knowledge layer, you can pretty much implement feature, triage your problem, ask any kind of questions, and it will answer to you that a human being cannot do. So that I see as the future. And so when we go on the cloud modernization journey, I'm expecting the team to basically say, you know this, convert it to Google, span up and move to the dual interface, and we'll give you the cloud blueprint and will be done. So that's kind of our journey right now. We are at AI STLC like our code assist PR ratio is growing 50% month-over-month, right?

Code Assist is growing 50% month over month, right? Can I ask you where it is on the spectrum from 0 to 100 or is it...

So I would say the number of PRs that we are doing right now, it will be around 2,000-plus PRs a week. AI-assisted PRs, no human involved in it. right? So that's -- if you really look at it, like that's a significant amount of -- right? And so that's kind of how we are really thinking about this.

Okay. You got to the edge of my understanding of all that stuff. Let's pull it back a little bit into what some investors -- is most important to investors, right? So a big debate with investors is whether PayPal can reaccelerate branded growth and checkout. What are the most important technical levers behind that? We just talked about stuff, you got to get on the most recent API. You got to make sure you're on the best UX and all that kind of stuff. But like talk to us about how you -- what's your plan to reaccelerate branded from a technical perspective.

So let's start with this. It's all about consumer experience. It has to be simple, right? When it is simple, it makes it a lot cleaner. So the first and the foremost trend is once you get into the funnel, our friction rate, our authorization like lagging friction rate, what we call, is at the lowest ever, right? So what we've done with that is with the MFA, passkeys, right? We're rolling out passkeys. We are getting millions and millions of passkeys generated, which what it means is for the customer, they don't have to really use a password. They don't have to input anything. If they use their biometrics, it signs you in, you're ultimately launching, right? That was one of our biggest friction points right? So we have removed that friction point, we get there. And then we also -- we historically had a billing agreement product where you can vault your payment with PayPal. When you do that, that's a 0 click you're not even going to PayPal. You're just like checkout, and it automatically is doing that. So we are now with the conversion improvements that we are doing in the flow of sell it in the instrument and so forth, now billing agreements are doing now, which then gets you vaulted and moves faster. So that has been the one biggest win that we've gotten. The second thing that we are doing from a technical perspective is, we are making, remember like I told you, we are not waiting for the merchant integration. We are upgrading everybody. That was with the consumer mindset because the consumer, irrespective of which merchant they are checking out on they're going to see the same experience. Familiarity is actually good because it makes you checkout faster. So those are 2 things. It is after the fact that you clicked on PayPal, but then now let's go into technically out and we influence merchants to adopt PayPal better. We call it presentment, like we want PayPal to be the choice that was shown. We have something called payment-ready APIs that we say, hey, merchant, this person because we know from the device and the sign-in from past merchants is a PayPal user, has a propensity to use PayPal. So what that helps the merchant is when you present PayPal as an option, their conversion rate is much higher there because now they know safe credentials are there, this payment is something that will flow through seamlessly and sort of a credit card, which might decline, like a lot of those things. So this is what I call presentment taken to the next level with technology. And last but not the least, is we have inverted the payment experience to be, you don't checkout and you pay we call it one-click storefront at that we launched with Meta. So in this scenario, a merchant is created an advertisement, it says by with PayPal. You click on it, you just pay and the orders dropped to the merchant, right? In that scenario, you're taking away the presentment. You're taking away the conversion, no choices, right? The merchant sees absolute high conversion. So those are the 4 ways. From a technology perspective, we are empowering the merchants to a better conversion, better customer reach and being able to deliver on their demand generation.

At what point do you think merchants collectively will say, PayPal was a pain in the a** in the past to use, conversion rates were low, and now PayPal is actually really good. When do you think we see that narrative emerging in aggregate. You're always going to have merchants that say you're terrible, but they're going to say that about anybody. But in general, when do we see that narrative emerge collectively?

I would slightly rather than -- I'll tell you a condition which creates that scenario. Right now, the merchants are facing so much fragmentation on the demand generation. They have social search, which has historically been there. They have seen social, they have seen Agentic. What is happening is in all of these scenarios, they have to revamp their system to be able to checkout and pay what we are hoping with PayPal similar to what we did 25 years ago is one connection too many with this checkout, I'm able to present you in any of these services and we're taking the friction out. That, to me, is the higher value, and we have agented services covered now, we have social services covered now, and we can improve their own website experience. I think that combined -- what I call is compounding benefits when they see that, that's -- and that's where -- when some merchants started seeing that, they are now moving more aggressively towards PayPal.

Then they start promoting you moving you up the fold.

Yes. And it becomes a flywheel now because the consumer now wants PayPal, and we are now going back to the same flywheel strategy.

So it sounds like you're laser-focused on conversion rates.

On demand. Generating demand as well. From your user base and also making it seamless for the merchants to be on any surface, whether it's LLM or social, right? So now like if you're integrated with our PayPal Checkout, you could go and put an ad on Meta, and you can have a single click purchase. You don't even have to checkout on the site.

Okay. Great. Let's switch gears a little bit. You've had a broad array of assets. There's a big debate on whether they're really linked and it's hard to separate them. And some people say, yes, they share services, but it's really not hard if you were to sell one piece or whatever. I won't mention that it does matter. But if you were to sell one component of the overall it could still survive on its own. So I've heard both sides. I've heard it's really hard to separate and I've heard it's really easy to separate, and you just have some kind of shared services agreement in your vibe. How do you feel the businesses are connected from a technical perspective.

So that's why I went back to my original statement, right? Standardized loosely coupled, highly aligned, right? There are a lot of capabilities. And the way I see technology strategy is not to think about like, okay, how do we get the best value, overall? Why decoupling the system so that like they can operate independently. So what we have done is we have been very laser focused on like there are SITs capabilities that are fundamental to everything that we do, right? Identity payments, risk servicing, right, credit and be BNPL. When you have the SIT systems, irrespective of what is the merchant or the customer offering, you have to build a best-in-class, so we are building that. We are standardizing so that all properties leverage the same platform underneath. They are decoupled, they are not tightly dependent on each other. They're abstracted out. So that's the best way to really build for a full future, while trying to keep the independence effect because I want each team to move faster. But that doesn't mean they have to do the same thing multiple times. So that's why I -- from day 1, the mantra is being loosely decouple highly aligned.

Loosely decoupled, I think, is the term just to take forward, right? So this is my interpretation of what you're saying. You let me know if you think it's right. it sounds like you're trying to solve for optionality and efficiency at the same time. Build the capabilities where it's more efficient, where they can all -- the different components can leverage things that they need, they'll do it once for all. But if there is optionality in the future and you need to separate, they can survive, you can separate it out. It would require connectivity and multiyear service agreements and stuff like that to draw on the services. But it can be done. Is that the right way to think about it?

I would say not only optionality, it's also the speed of innovation, right? It's also economics. One, when you see all the payments going through the same layer, you are learning from each other, right? I think if I see an instrument in Braintree per se, when I'm processing it with the same payment system, then I see that on the PayPal side. I have a better understanding of it rather than like -- so I would not just call it just optionality. It's a better way of doing business, better economics, better speed of innovation, while preserving the optionality.

Makes a lot of sense. And I think that's pretty consistent with what Enrique has said at different points. Okay. Let's talk about Venmo Yes, just looking at the time here. We've got a boogie. We're not at the speed round phase yet, but we're getting there. Venmo has incredible consumer engagement. Monetization has taken a little bit longer maybe than some we're hoping for. From a product and technology standpoint, what needs to happen for Venmo to become more meaningful commerce and financial services engine.

So we talked about identity, I will not repeat it. But what I wanted to say is like that basically gave the oxygen for Venmo last year to do the user-facing innovation stuff, right? They launched in Venmo Stash loyalty platform. They launched Venmo Debit Cards and Credit Cards and so forth, right? If they were focused on blocking and tackling the identity do not have moved fast. So that was the first on lot that we got out of the identity win. We are now converging payments. As soon as we converge payments, now everything that we did with PayPal Checkout, we are going to take the same set checkout subsystem and now make Venmo provide that functionality. You might say, what does that get you...

Talk about what do you mean by being able to pay with Venmo anywhere where PayPal is accepted kind of -- Venmo Debit Cards or whatever.

So there are 2 pieces, pay with Venmo is a product by itself. Right now, it's using stored value and credit cards cuts, right? But they don't have a buy now pay later offering. Now when the payment systems are turned this year, right, what happens is now Venmo will be able to offer buy now pay later instantly. That is the demography that really is very attractive to BNPL. So the fact that we can light up the BNPL pretty quickly becomes a meaningful point. The other point is Venmo debit card penetration still ways to go, it's about 8% of our total Venmo. And that came from our common financial services, right? So now we are able to issue a debit card like we do for PayPal debit card. So that's where as they leverage the same underlying stuff, they can now do credit, they can do buy now pay later, and they can do the pay with Venmo and move faster.

Common functionality and Venmo.

The third thing that we did, which is we launched we call PayPal World, we had made PayPal, Venmo interoperable. You could would actually send money from Venmo to PayPal and PayPal to Venmo, seamlessly, you just have to put a phone number, and we can look it up, right? By doing that, we basically now made Venmo to be able to send money to anybody in PayPal, internationally, right? That has been a huge offering where people are now leveraging that very much. Why I'm bringing this interoperability is with this interoperability when we get the payments and the pay sheet integration done, with the interoperability, any merchant who's already offering PayPal, we can actually do pay with Venmo seamlessly, right? When you click on the PayPal button, any wallet that is PayPal World wallet will automatically show. It would be UPI, it would be WeChat, it could be Venmo. So now what we are doing is we are basically plugging the Venmo into the PayPal flywheel. So any merchant, you don't have to integrate separately for Venmo.

When that was first presented I was like, wow, that's a really interesting idea. But the UX was kind of -- if I'm a consumer in a foreign country that doesn't know PayPal, right, and all of a sudden like checkout with PayPal. My first reaction is, I don't have PayPal I'm not going to use that. Is it more going to be whatever the local scheme is powered by PayPal.

So there are 2 pieces to it. The first piece is the local country will promote it. That's one piece they see PayPal you will be accepted, that will be number one. Number two, remember, the payment ready API that I said, when you -- and the merchant starts using the payment-ready, right, with the new integration. Now even the PayPal button can be rendered in the local thing because we already know this person has a UPI account or a WeChat account. So we are trying to be backward compatible, while we are trying to -- so that's where you might have seen it as you clicked on PayPal before I see the wallet. And with payment ready, we can actually move it one step ahead in the game, right. And I want to -- it's a challenge for a lot of merchants because they do not know where people are coming from. They usually have a master of all the payment methods, you've seen that like APMs, multiple APMs, that is usually very detrimental to the conversion. So if somebody can tell you the propensity to buy on this payment method is the best, it actually will help the merchant.

Okay. Let's hit Braintree for a quick sec and I got a summary question for you. Braintree is strategically important for some, some people think it is not, difference of opinion, fair. How do you think of Braintree's role in the technology stack and the opportunity to leverage Braintree, which is a merchant solution with the PayPal and Venmo Consumer Solutions.

So for me, first, it's the foot in the door, right? If you have a payment service, everybody needs a payment service partner.

Braintree is a foot in the door.

Foot in the door for merchant? So whether you're doing branded processing or unbranded processing or any APMs that you want I want merchant to adopt Braintree. So by doing that, number one, we have an integration in which we can now upsell branded PayPal, Venmo and so forth with a single click of button to enable those payment methods in the merchant side. That's number one. Number two, with the volume of transactions we do, payment is all about auth rate. Payment is all about like how well we do economy of scale, right, whether it's routing, whether it is knowing what is the low trust provider to go to. Knowing everything about the back-end interchange rates, right? Those knowledge that we deliver with more and more volume it helps us on the economics. So it helps us on the risk. It helps us in the fraud. It also the economics, right? So that's why debt tech is very important because it gives us a holistic view of the consumer. And then I can go on about like the data play, which we see is even more valuable.

Yes. Okay. In the last few minutes we have, I wanted to ask you, the investment community is so focused on growth rates and margins and that kind of stuff. Some people who have a technical background totally followed what you said. Fewer would probably say that was a very basic conversation because not many are -- and some will say, it was over my head. So from your perspective, I will be able to say this conversation and what you've set out as your vision, it's showing up in the numbers. How will we as an investment community know that what you're doing is impacting results? Where will we see it?

I think you will see it when the merchant I talked about the multiple surfaces on which they want to be treated. You would see when the merchant now says, my catalog is available for demand generation on Agentic surfaces, on social surfaces...

As it relates to PayPal.

As it relates to PayPal or Venmo either way, right? And it's a buy with Venmo or buy with PayPal, and it's a single click checkout. That's when you know. But like we basically are helping the merchant drive demand, which is what matters most for them.

Yes. So we'll see it in revenue, we'll see it in margins. And I'm assuming maybe at that point when things are at an advanced stage, maybe Steve and his crew are going to offer some different KPIs about showing where the progress is. Steve is nodding his head for those of you online. So okay. The last thing I want to ask about is the company saying there's $1.5 billion of cost savings. Some of that will be reallocated. Not asking you to quantify how much is going to be reallocated at Steve and the management team's job. But the way you see it, the efficiencies on the cost saves, where do you want to reinvest what your -- do you see the IT budget or what you supervise the spend actually going up and realization of much, much more efficiencies from that spend? Or do you see it as, no, we are a driver of cost reduction and efficiencies across the board, while still facilitating growth.

So there are -- I would answer it in 2 parts, right? We want to be the best architecture. I'm not wanting to be evolving architecture. I want to have the best platform out there 3 years from now. So that's the goal. So the idea is if we modernize and if you cut down the tech debt, we should have significant savings, but I want to reinvest in actually delivering the best platform out there, right? And we do not know where the AI will take it. But what I would say is we would reinvest to be the best-in-class technology platform out there. So that's something that I would, as a CTO would push for how do we be the leading platform play in there. I do not want to sacrifice on auth rates. I do not want to sacrifice on risk, fraud. I want us to be the best identity provider out there. I want us to be the best checkout instrument out there. I think with the amount of automation, cloud and tech debt and all that kind of stuff, I think we can reinvest and we can be more efficient is how I see it, but it will be a 2- to 3-year cycle. And I think there's definitely efficiencies there.

Yes. I always used to press my CIO. This has got to go faster in his response and knew what he was going to say every time, it's like you can't put 9 women in a room and have a baby in a month. Things take time to build, which was his version of saying, we're moving as fast as we can, but shut up. We think we're doing our job. How do you -- you're getting pressure, I'm sure, to go faster and faster and faster. Do you feel like there is -- what's kind of like the vibe inside, like, look, you got a big job to clean up and build and do everything all at once, while you're running a multinational platform that's processing hundreds of millions of transactions all the time, right? So how do you balance speed versus doing it right?

I have a technology, which I call [ S and ], okay? So we have to balance, right? And the thing is out have said the same thing like your CIO said, you can't do that. But now with software factories and agents, it's becoming a little bit more of a reality that we can do four things parallel. So I think what you're going to see is significant velocity improvements in the coding innovation piece of stuff, where I would be very cautious and we'll be thoughtful is on the migration piece, right, as we move from this Oracle to the cloud, it's going to take a year, 2 years to really migrate, okay? We do not want -- we can't rush that. So -- and that is an important part. So we are now balancing the art of the possible, which is how I push my team. There are things -- we do not rush. We plan it and we execute it systematically. But there are other things we should just be innovating faster if it's a user innovation, if it's a front-end innovation, right? Like and Venmo, like, for example, the user innovation, 50% of it is auto coded by right now. How can we innovate faster on the user experience is how we're pushing right now?

This is the last question because we're down in 5 seconds. But what is the leverage ratio for AI agents versus human programmers? Let's say, a human program cost $200,000. If you allocated that same $200,000 to agents and using the best tokens or whatever. How much -- what's the ratio there? Or $200,000 in tokens, you get 10 programmers, you get 20?

So I would say it slightly differently. It depends on the work that you're doing. Yes. Okay. If you are a vibe coding, it's not maybe the most efficient form. If you're lifting and shifting complex applications, it's worth its pound like maybe 10x more 20x more than what it is. So I think what is happening is people are using it a class and diverse nature of it and your target seeing people talk and match out right now. So without seeing value, right? So what we have been trying to -- at least I'm trying to push is been a judicious use of the tokens, right? We are focusing on the big lift of shift, which is why I'm doing AI STLC, where hey, if we can fully be autonomous, that's worth while spending the tokens on, and if we can understand the knowledge layer and all that kind of stuff. Everything else is just a speed to the market right now because of the trust a tokens. Summer, when the economy goes down, are we start putting AI machines at the desktop of our developer, then maybe we don't have to pay the tokens now, it's all free. It's all your open source LLMs, and then it's CapEx. So it's an evolving story, Adam. I would say right now, it's a hybrid. But if you -- for modernization or major platform rewrite, you can't find a better solve. It's not about cost, it's about efficiency, accuracy, speed and consistency.

That's great. Okay. We are a minute over, but thank you very much. This is a great conversation. And again, I think it sets the stage for some more narrative coming out of you. So Steve, all my questions coming up on earnings calls are going to be about Srini.

Good. On it.

Thank you.

Thanks, Srini.

Nice meeting you.