Conference Transcript - DOX 2026-08-12

Speaker 2

Good morning, everybody. Tim Horan, the communications, digital infrastructure, and satellite analyst now here at Oppenheimer. My pleasure to be hosting Amdocs again, I think, for maybe the fifth year. And I've hosted probably Anthony, the group president of Amdocs, much, much more often than that, actually. Right. Anthony, I will always mess up your last name, but we're good with Anthony for now. Yeah. And the maximum head of IR is here also. You know, it's just really, I think, fascinating times in technology and networks and compute. And clearly in the economy, we just think we're on the precipice of incredible changes to the economy with this AI general purpose technology. And really, I think it puts Amdocs in an incredibly strong position to help the industry adopt AI, both from new services and also to digitize the businesses to help a lot more. But, you know, Anthony, before we get into all that, so I'm really looking forward to this conversation.

Speaker 1

Can you just describe a little bit more what you what you focus on day to day as group group president? yeah yeah yeah sure so you know we're obviously in the middle of uh middle we've had it seems seems like we're in the middle but you know three months ago we transitioned to shimmy as the ceo and we've really been focusing i would say on um just realigning our vision and our strategy in terms of kind of where we want to focus on how we want to drive the company for you know not the next quarter but the next year the next three years the next five years really grab grabbing all of this i would say convergence of technology from connectivity to ai to cloud acceleration to the edge to all all these technologies are kind of like on a convergence path which is great um and really delivering value right while at the same time also trying to balance the cost implications for our customers because obviously they're huge big customers as they scale so we're kind of right in the middle of these having very very interesting dialogues um with our customers while trying to set the roadmap for the company so you know it's a it's a lot of fun times uh i think uh we've been working literally seven days a week so uh i always say you know you need to be worried when you're bored and there's never a boring moment in amdocs and you got to work hard when you get new bosses yeah yeah and uh and good luck so can you give us uh any uh more uh color on the strategic review uh you know where you guys think you should be uh you know doing as a company competing and maybe you know just at a high level what other products do you need yourself sure sure so so as we kind of shared in our earnings you know our strategy very simplistically, to put it, is kind of these three pillars with a horizontal pillar at the bottom. And the first pillar is really what we call our agentic transformation pillar, which is really focusing on telecommunications, our BSS, OSS, really coming to the market with the next generation agentic platform, you know, keep doubling down on that. You know, we have a north star of we believe that you know eventually it is really going to be business to machines right meaning you this kind of massive layer of doing a lot of it and doing a lot of change requests and things like that is kind of going to disappear as software starts to kind of auto heal and evolve and be more agentic and be able to give you all of these capabilities and we're starting to push some of those boundaries and that is really pillar one just taking our customers to the next evolution of software, which is natively, agentically built. The second pillar is we believe that we have this DNA of going deep vertically, right? We are not a company that serves 17 horizontally, inch deep, mile wide. And we have some DNA in that, right? We build mission-critical software that needs to run 24-7. We work and build software that works in a highly regulated industry. We are super, super focused on things like PII and privacy. And, you know, obviously with quantum just around the corner with security, we're already working on that stuff. So there are some DNA and characteristics of Amdocs going very deep in these complex verticals that we think can be replicated to another vertical. And so even though we haven't shared specific information on that in terms of the vertical, we think that our next play is to go to another vertical that has similar characteristics and bring to bear kind of this core competency from the agentic side of toco to another vertical. And the third pillar is really something that I'm super, super excited about and i think there are a lot of very interesting emerging markets um not geographically but emerging markets from a from a kind of landscape perspective um where there are a lot of things that we're building uh that could be viable for so you know just we were talking just before this right i spent the day over here at nvidia and we spent a lot of time talking about the edge and talking about edge mesh grid and how that can work and how workloads can balance you know tokenization you know how do you kind of measure that you know balancing all of that together because that could be the next frontier as we start to go to all sorts of by the way it's not just that you need lower latency and better capabilities on the edge and you know you know There's a great opportunity for Telcos to create this mesh grid on the edge, but there is also available compute power, right? Like if you look at the need for the next three to five years, although there's this massive data center build out being done, they're still behind the curve in terms of the need for many of the enterprises. And here you have fully air conditioned, power supplied, fiber running, all of these edge points close to enterprises and consumers that can be utilized. But that also comes with a burden of how do you manage the workloads? How do you monetize it? How do you treat it differently? How do you treat security on it? How do you partition it? How do you virtualize it? So there's a very interesting stack of software complexities, and that's kind of some of the big items we've been sitting here talking with NVIDIA in the last day or two. So that third pillar is very, very interesting and exciting for us. And we're building some of this stuff for us, right? Like you take something like quantum security, like we're already, you know, been working on this for the last six months for our enterprise applications, right? They need to be PQC compliant and things like that. and then you just wake up and you go, hang on a second, like we're doing this and it's so valid for so many others. So this third pillar is really to look at, like, are there any of these opportunities? Now, you know, we're not going to wake up tomorrow and say, like, we're going to go into pharmaceuticals or anything like that. But are there some of these opportunities, like, on the adjacencies that could be very viable to take to market as well? That's kind of the third pillar. And the fourth, I would say the horizontal layer is really a focus on real internal transformation. And I think, you know, Shimi, you know, taking the CEO role, this is something he really embraced, soup to nuts, and said, hey, like, we need to be customer zero here, how do we transform completely internally to be fully agentic, right? And we are one of the, you know, bigger cursor users in the world, for example, as a stat to put out there. You know, the fight now is not how do we get our people, you know, coding on cursor and codex and things like that. The battle now is, you know, there's also this tokenization costs and things like that, that you need to measure and balance and make sure that there's an ROI. By the way, it's a completely separate discussion we can have for three hours. It's a global problem people are looking at and people need to grapple with because your cost is not just your workforce. Your cost is your work, labor, plus your token cost for each employee, which is an interesting economic equation for most companies. And I'm sure you've seen the tweets and things like that of companies like Uber just waking up and going, oh my gosh, like, so we are, and that's another good example of something that we've built internally, right? Like having these dashboards per employee, what are they using, routing to cheapest LLMs, all of these is internal infrastructure we've built. We're like, hang on a second, this is like super valid outside, right? So back to that kind of how all of these pillars connect, that's back to pillar number three. So that, in a nutshell, that's our strategy absolutely pillar one with you know doubling down on that we think there's an amazing opportunity amazing opportunity like you know that this LLA deal that we we signed and announced last quarter um it's an agentic transformation journey and and we are we are starting to see and I know you and I have had a lot of conversations about the many POCs we're doing we're starting to see these like our customers go okay you know it's not about can i launch an agent to production because yes you can launch an agent to production how do i transform my company how do i have a who are the two to three partners that i want to partner with that will take me down this journey and this is where we think there's a sweet spot for amdocs to really play and make an impact and really set the stage for the next you know five six well anthony i opportunity is obviously huge, right?

Speaker 2

The telcos are still very, very, very inefficient, you know, in many ways. I mean, AI, it seems like made the order to massively improve the quality of the service, customer care, right? And frankly, their own cost structure. I mean, they can automate and agentize their entire business over time. And you guys are in an incredibly strong position to do so. You know, I guess the multi-billion dollar question is, you know, are you going to, can you execute or are you going to lose these customers to new competitors? Right. That's, I think that's a really, really important point. Like, you know, why do you retain the customers and, and what, what gives you confidence that you can execute?

Speaker 1

Yeah. I, you know, without being flippant, I would say like we are the adult in the room in a way. OK, and what I mean by that is you have a mission critical, in some cases with our big customers, 100 plus billion dollar business. And, you know, you are not going to wake up tomorrow and someone that, you know, great, amazing to see these amazing valuations, you know, the company and dollars getting a billion dollar valuation. You know, I also want that. Right. Right. But on the other hand, like, do you trust your entire business here to something like that? Right. And so if you can get both, meaning if you can get the responsible adult in the room who knows how to build mission critical software. Right. I'll give you just a very, very simple anecdote just to illustrate something like this. Right. So you have let's say you pick this great platform that has a great agent and he's amazing and does all the fantastic things. fantastic i'm going to get it um for whatever reason let's say you know claude's not available or open ai is not available or a zone is going down like i mean are you going to tell your customers oh like come back in three days um what happen if a model gets taken offline what happen if you start to scale it and suddenly realize that every one of your income inbound calls, I'm using, you know, 20,000 tokens for each call. I mean, there are a lot of complexities that needs to be managed in a complex organization. And, you know, we are thinking through every element of it because we understand the business. And that combined with knowing the workflows and the business processes and how you take a product from A, all the way to provisioning on the network, right? It's one thing to say, hey, like I can answer a customer care and do an upsell. It's another thing to know if you can provision it on the network and make sure it gets delivered in 24 hours. And so when we come to the table, we're taking all of this into account. We're taking your entire gamut of security into account. We're taking your regulatory concerns into account. So I think, you know, when we come to the table, it's not like, hey, like, look at this nice shiny little thing here. Like, you know, do you want to use AI? It's really deeply thinking through your business. And one thing we found, by the way, in this kind of new strategy that we're rolling out is we don't have 10,000 customers in telco. There are like 50, 70 customers that carry 75% of the revenue globally. And many of these customers need a very contextualized path versus, here's a box, take it, make sure it fits into your ecosystem. So I think all All of those with kind of the deep understanding of where our customers are at puts us in a very, very good position. Also, you know, like the relationship we have with NVIDIA and AWS, you know, it's not like we're just saying, hey, like Amdocs can do everything. We're bringing these strategic partners to the table. We're working very, very deeply with them. You know, it wasn't just me here at NVIDIA. It was our R&D teams that flew in here. Today, it's actually an R&D day. They're sitting together with the R&D teams. we are giving feedback on what works, what doesn't work on some of the Nemotron models, right? So it's like a to and fro. And so I think bringing all of these to the table, it's tough to find someone out there that does all of this for telecommunications.

Speaker 2

Yeah, I completely agree. You know, it was just partially, we get asked by investors, you know, why should you guys win in this, right? I mean, obviously you got to execute. You need a lot of AI skillsets that are very, very new. But it's like you're game to lose at this point, right, as long as you can execute. And, you know, I guess on that, I was just interviewing, I just had a fireside chat with Verizon's head of business unit right before this, Kyle Malady. And, you know, he basically said a bunch of the same things, you know, you're saying, what they need to develop. They need like an AI platform that will help customers manage their token utilization, token costs, model utilization, you know, which models to use, you know, integrated with data and everything you're talking about. Right. And they would love to have, for lack of a better word, like an AI platform in a box, you know, and it would seem like you have obviously a thousand customers around the world. they're not all going to develop this on their own. You seem like perfectly well positioned to develop exactly this product that they want. And I guess in that point, do you have a product like that? Yeah.

Speaker 1

And, and, and so one is the product, and we are continuing to push the boundaries on our kind of our BSS, OSS stack around AOS and making it fully agentic. The second is also the harness. Think about the harness, right? Like, and this becomes in a way, it's even like a bigger focus today. So you just spoke about all of the different models, right? I guarantee you, if you're a company like Verizon or Amdocs or whoever, and you allow your employees all the tools to do dev and stuff like that, everyone's going to default to Fable. You don't need to, okay? So how do you manage this? So now you have a LLM router that analyzes tasks and says, here are the tasks that can go to Sonet or something like that, or maybe even an open source model. Right. And then balance it across the organization. And we haven't even gotten to, you know, we're working on a bunch of kind of open weight models. Right. Because this is, by the way, this is the next big issue that's going to hit us. Right. When you're an enterprise, you know, and people, people mix the words open source and open weight and they're not.

Speaker 2

Yeah, I use open source just because I'm lazy, I guess, yeah.

Speaker 1

Yeah, yeah, yeah. But it's two, by the way, two very, very important paradigms, right? But they're not necessarily exactly the same, right? You can have an open source model that's not open weight, right? And you can have a proprietary model that's open weight. Meaning, let's talk about open weights for a second, right? If you are using a model and you are an enterprise and you have all these non-deterministic software that's driving your business, you want to know where it's driving, how it's driving. You need the transparency. You need the observability. You are a regulated industry here, right? And so that's why open weights really, really matter in decision criteria. The open source element is much more driven, I would say, by the cost factor of tokenization or token economics, if you like. As you start to scale, and you know, our industry is, I mean, the volumes, I mean, you know this, Tim, but the volumes in our industry is like second to none. And this is only, you know, this is only increasing, right? And now you need to deal with not just employees coming in, but agents coming in, right? So if you thought you had 50 customers, tomorrow you may have 140 customers because 70 of them or 80 of them or 90 of them will be an agent coming in trying to negotiate a deal and connect your e-commerce server, right? And connecting via MCP and buying something and looking around. So like, who is thinking about that? We are, right? Like we are building stuff to have agent negotiations on the front end because this is just around the corner.

Speaker 2

It's already happening today. and and so you really and Anthony my question was I know you kind of you built like this AI platform internally to kind of manage all the AI assets right and and you're going to and the telcos need to do that internally but the telcos also want something for their enterprise customers right because the enterprise customers are going to need and they think based on everything you're saying you got to tie together networking clearly right with the AI compute data centers the actual GPUs, the models, the applications, the agents, it's all incredibly complicated, right? And then most importantly, you need privacy, security, data, on and on. And it seems like you guys, you're developing it internally, and you know how to deal with the telcos. I mean, do you think you can develop a product that the telcos can go to enterprises with and say, look, this will help manage all your AI assets?

Speaker 1

Yeah, this is where I think, you know, kind of connecting a little bit to Pillar 3, I think there is a great opportunity because what we're building is kind of agnostic in a way right I mean you can easily white label sell it to enterprises and there is no issue with some of these because the enterprise are grappling with exactly the same problem right you know talking about that model router for example right talking about you know security talking about harness governance all of these all of these things enterprises so I think it's a great opportunity that our customers can take some of these components white label it and just expose it to their customers and that's you know in kind of pillar one that's another i i would say now it requires two to tango right meaning it requires a customer to think bigger than just connectivity sure right so so like you're going to see future customers start to have bundles of you know we already have customers talking to us about this right bundles of like here's my phone here's my iPad, you can have 10,000 tokens you can use a month, for example. It could be across frontier models. It could be their own hosted open source model. You can get your own personal AI agents. So here's kind of where the industry, I believe, is going to morph to on the consumer and even on the enterprise side.

Speaker 2

That niche of SMB to mid-market, I think, is one of the biggest opportunities that telecommunications providers have right now because you know i hadn't heard of that until just now you know bundling tokens and it makes a lot of sense right anthony and you can bundle in many other products but you know the telcos had this they were in a good position 15 years ago in cloud right they they could have been the cloud providers for enterprise customers and you know they just didn't execute i mean i guess the question is can they execute on this AI, is there something different about AI that unlike cloud where they can, you know, actually potentially execute a bit better?

Speaker 1

Yeah. And, and that's the, you know, like the pessimistic view is, you know, you, they were first to cloud, right. They were, they were, you know, the whole OTT kind of thing that happened where, you know, the monetization really didn't flow as much under the telco side and things like that. But at the end of the day, I think the most important thing to understand is none of those technologies could have come to bear without the pipes, without the connectivity. So I think at least that element is there and that's not going to go away anytime soon. I think it is going to require some bravery from C-level within our customers, within CEOs, just going, hey, like here's where I want to go, right? Not just kind of stay with the status quo and say, Hey, I'm going to sell the next iPhone. Yes, you should sell the next iPhone. But like, what does that, what does that look like? And, you know, going back to the initial point where we started from, you know, this is where I'm also excited about the opportunity around the mesh grid around the edge, because you think of the number of pops that telcos have, right? You know, we're not talking hundreds, we're talking tens of thousands in a city. And it's not just about the proximity to the enterprises or the consumers. It's just about the fact that you have all of these kilowatts of power sitting in these things that are unused, that are AC, that are powered, that already has the space secured.

Speaker 2

Yeah, no, that's exactly what Verizon was talking about. I mean, everything you and I are talking about, we just had this entire conversation. And they were also talking about maybe selling or releasing gpus as a service maybe they own it maybe the enterprise owns it but you know with with cloud the utilize one of the reasons i think enterprises outsource compute to cloud was the utilization internally of cpus is fairly relatively low i mean if you're looking at ai and you would know better than me because you're spending a ton of money on ai yourself but you know you're running those things 24 7 i think almost right because um and so why would you share that compute with somebody else when you were there's no there's no economic rationale behind that you know so and and and also enterprises they want to control their data right they want to control their intellectual property you know privacy and security is absolutely paramount everything we're doing here so i mean i think there is a whole lot of reasons why the telcos can be very successful here and why you guys can you know enable them to be successful but you know to your point though you need some brave leadership right and and they they got to really step up to the plate i mean verizon now has a ceo that came from silicon valley right so and he's already you know had his career you know in many ways so in a lot of ways he's doing a lot of brave things because you know it's uh he doesn't have a lot of personal downside and it's yeah look yeah it's definitely an opportunity i the only thing i would say just to the initial comment you made was you know i think there is a bigger play than selling gpu as a service right gpu as a service at the end

Speaker 1

of the day, is going to be a, I don't want to say commodity because we're not there yet, but some type of commodity play, it's a cost plus exercise. But now if you can start to go northbound and start to provide services, inferencing services, all sorts of things above it, that's really where the margin play and the opportunity is. It's not just saying, hey, I have a GPU, do you want to buy it? No, you need the whole platform of everything we're talking about. The GPU as a service is one piece. Yeah, it's one piece. So that's why I think, you know, if you look at all of the SMBs and the mid-market companies around the place, these guys don't have a seat at the table with the NVIDIAs, right? I mean, they are the ones that are needing to, you know, there is absolutely no doubt that every company out there in some shape or form is either embracing AI or wants to know how to do it. And so I think, you know, as a telco, you are the first to the table, right? When a business opens, right? Think of the order. If I'm a SMP, I go and start a business tomorrow. Think of the order of what I would do, right? I would register my LLC. I would find a physical premise. I would connect the electricity and I would get connectivity. Like that's literally the order of what I would do. and then every all the other 10 000 things i need to do okay so so i think you're you're the first at the table so when you are the first at the table you have the most amazing opportunity to go hey like have you thought about this you know like we have this harness very secured privacy and you'll be coming to the table with the most amazing name in the industry because you know, people may say everything about telcos and write bad comments on social media or whatever, but they trust the telco at the end of the day. They're like, I know you're not going to wake up in the morning and decide to screw me, right? Like, this is not the brand that telcos have, necessarily, right?

Speaker 2

That's a great, great point. And I think just maybe you can switch gears a little bit. You know, a great deal with Liberty Latin America. They gave a number out, and I don't follow the company, you know, savings, but can you put, you know, not just focused on them, but you know, if I was to entirely adopt AI, what kind of cost savings should a typical telco be able to see? Do you think, I mean, all in, if I just went completely agentic, you know, and I, and I'm not holding you to this number, but I mean, yeah.

Speaker 1

Yeah. Look, I think there are, there are two streams you need to look at in parallel the first stream is you know if i was going to wake up tomorrow and create a completely new telco i would do it ai native from the ground up and and in a way that's easier to do because you don't have a legacy of systems and things that you need to bring along the journey right but most of our customers on track two meaning they are an existing company they have a business all of that stuff so the ones that can push the needle in terms of savings are coming horizontally and going look let me completely reimagine how i'm going to do care let me completely reimagine how i'm going to do um service in terms of truck rollouts and and you know those layers or even if you look at the network layers horizontally if you start to kind of slice those out and you, you, we are going from the perspective of selling, you know, if I go even three, four years and I look at the customer discussions that were like, Hey, we need like this functionality, this feature, we need charging to do this, things like that. The discussions we're having with customer now, customers now is let's look at your business workflow end to end versus, you know, bits and bytes and feature functionality. Let's look at your, let's map out your customer workflow end to end. How do you now change this customer workflow? We have one customer, not sure if we released the name, but we had one customer that came to us with a problem on the network side. And, you know, like we were able to save in that particular space, we were able to save like 60% of their annual cost, right? And this is something that they did and they have to do on a daily basis, right? And this was taken straight out of their annual operational cost.

Speaker 2

So- I believe it. And you guys had a big push two years ago, like trying to automate the contact centers a little bit more. You know, have you had much progress there or the telco seeing results there?

Speaker 1

Yeah, I think the contact center space, I think is jumping in leaps, right? So I think, yes, you will have, you know, potentially a contact center and things like that. But I think what we are starting to see now is the customer care agent and most of the inquiries coming in, which is like 40 odd percent billing related right i mean you can you can take not a haircut like you can completely slice this by you know a magnitude because most of these agentically very contextually driving high nps can be solved purely in an agentic manner and so you've implemented that and had a lot of success it sounds like yeah well we're starting to see now customers are doing it by so they're taking instead of going hey we're doing all of care they're like hey like let's do it on a like we have one customer that started on a whatsapp channel so on the whatsapp channel all of the inbounds and this is a this is a customer in europe all of the inbounds coming in a whatsapp channel and you know in europe they use whatsapp a lot so that's a very uh important channel for them. There were only 3% that needed human intervention. So 97% were being solved without any intervention or escalation. And that's that to me, like, I went back to my team. I'm like, hey, just double check this. Like, is this right?

Speaker 2

You know, and we'd be clear about that. Sorry. So said another way, they were able to cut like 97% of their live human handling of problems.

Speaker 1

Yeah, and having it solved by just people chatting to an agent in the back end connected to WhatsApp, connected back to the system.

Speaker 2

And were those humans running those chats or agents running those chats?

Speaker 1

So it was a combination. It was a hybrid. So there was like originally on WhatsApp, they had like a standard chat bot that were doing it. but when the chatbot was doing it like 60 70 percent of the calls were like i need to talk to a person like give me a human uh like this is not the answer that i wanted versus now you're getting 96 97 percent of those completely handled and people happy and moving on oh but but it's a hybrid agent and and uh live person yeah yeah so the light no but but the live person is only three percent of all inbounds oh okay got it got it so what you're saying is the agent has improved dramatically with these chats yeah so 97 are being handled completely by an agent versus three percent are like oh i still need to talk to a person or 60 and then and what was the big change is it just ai has gotten so much better yeah it's it's just inferencing training um knowing kind of the path, you know, I mean, this is just completely, like, think of the original WhatsApp launch that they had done was just like a standard AI, not even gen AI, I think it was just a standard chatbot. And that was, you know, fielding 30 to 40%, something like that, still, you know, decent metrics. And now you've just accelerated that to fielding 97% using an agentic agent.

Speaker 2

So Anthony, I've gotten through about four of the 30 questions I wanted to ask you. So maybe we can just be brief in this. I mean, are you seeing the hyperscalers as competitors or are they getting more aggressive? Have you lost any customers to them or any business to them?

Speaker 1

No, you know, they still continue to be partners of ours. You know, obviously the the big three we still work very very closely i was just in seattle three weeks ago with aws had a great session with them um so no i i mean we we don't really see them as a as a direct competitor just it's just because we go so deep right yeah and how about the you know the systems integrators the accentures of the world or maybe even like the service now sales forces the do you see them as competitors or are they getting more aggressive no i i wouldn't say so Well, I would say, you know, the kind of the competitive moat, I would say, is more around the internal IT space going, hey, we have this tool set, maybe we can build it. And they may try that path internally. And then they'll go, oh, you know, maybe, and we've had several customers try to do stuff and they go, oh, you know, maybe we can accelerate even faster if we bring Amdocs with us and we will still build some stuff. But so I would say.

Speaker 2

Well, that was my next question. And, yeah, I mean, for them to build this in-house would be a total disaster, right? So just my words, it makes like perfect sense.

Speaker 1

But it doesn't stop them trying, right?

Speaker 2

Yeah, they're going to try. But I guess just even in that regard, I mean, how do you get your own AI people and skill set? How do you keep those people to be able to build the tools that you kind of need right now, right? I mean, they're in demand.

Speaker 1

Yeah, it's a tough one. You know, we have a very strategic geographically kind of spread company, right? Right. So the U.S., Israel, Cyprus, Philippines, India. So we are not always coming to like one market and it's not the same market. And I think that really helps us a lot. And, you know, people people come along the mission. Right. You join Amdocs, you know, you're on a mission to save the world. I mean, you're signing up for that. So so I think I think there's a lot of different factors that kind of play into this. um there's no there's no magic at the end of the day right i mean it's a it's a tough it's a tough skill set tough market out there and by the way the training that you get the experience that you get is bar none in mdocs right like with default you're coming in you're not working with ar you're not working at all right from marketing to legal to hr i tell you one funny thing i just got an email this morning from one of the HR leads going, hey, we built this amazing system that does all of this stuff in HR. Do you think that there's a marketing, you know, can we like take this to market? Like that's the mentality in the company. We have nothing to do with building HR systems, right? But every unit within the company, like this is the focus and this is the mind. And this is really, you know, Shami as the new CEO, this is really what he's driving through the entire organization.

Speaker 2

Well, that's really, it's really, really exciting stuff. Well, and I guess in that regard, I mean, do you think yourselves it's a major productivity improvement? I mean, is it something that can move the needle in margins? Yeah.

Speaker 1

Yeah, yeah. I think, you know, I think definitely, I think on the dev side, I think the one thing we are cautious to do is, you know, we need to have complete transparency in what we build, what it does, because when it goes out to the field it goes into mission critical systems so this is not just a you know vibe coding angry birds and put it on the app store right like so so every step that we do we are making sure there's human in the loop in terms of you know the the quality the code checks but definitely you know this pace is accelerating fast we think you know here's the exciting part right like we think productivity wise what we could have done don't just think of it from hey like you can do A and now you're going to have much more margins. Maybe we can deliver A, B, C, and D to the customers. So, you know, look out for new areas. We might go into, you know, expanding kind of our addressable market, giving them more value because, you know, our customers also have access to these tools, right? So they are expecting more. They're expecting more productivity. You know, it's not like Amdocs is the only one that has this.

Speaker 2

Yeah, no, that's for sure. And can you talk about your, just on the compute side, I'm kind of asking everybody, it seems like there's a mad dash to get more GPU compute, you know, prices are going up. And, you know, it's very, very hard for companies like yourself to control your token costs, right? Can you get the GPUs that you need to run your business? And can you talk a little bit about, you know, has it hurt margins? Has it hurt expenses?

Speaker 1

Yeah, look, I mean, we have great relationships directly with NVIDIA, you know, with every level from the top all the way down. So I think we are good from that space. We were actually talking about, you know, looking at going much more hybrid in terms of, you know, Frontier versus on-prem and reutilizing some of the NVIDIA GPUs and things like that. So we haven't hit that bottleneck from scarcity that we don't have access to the GPUs yet. But I think from the cost perspective, we need to keep an eye out on our employee base, obviously, because it's very, very easy for, you know, Anthony to turn up and, you know, suddenly consume, you know, seven different AI apps and just drive this cost number. So we are putting, obviously, controls in place, metrics in place. We can now monitor every employee and their usage. I actually saw a dashboard yesterday. I was like, ah, I wonder what this person's doing. So you can probably got a call from Shimmy and got a call from me going, hey, what are you working on? It better be good. Yeah, yeah. You want to drive AI.

Speaker 2

On the other hand, you also want to make sure there's an ROI to whatever they're doing and they're not building Angry Birds. well i know i asked you a lot of questions that matt didn't appreciate but you did very good answering them all and uh but we're out of time anthony and you know i could talk to you for another hour and uh you know you know where to find me yeah i'd love to go see you in texas in person yeah yeah love to we can do a podcast on all things ai we can matt thank you for lending anthony to me thanks everybody have a good day thanks a lot thank you guys bye