Amdocs Limited (DOX) Earnings Call Transcript & Summary

Hello, everyone, and welcome to today's webinar, unlocking the transformative power of generative AI. Ever since generative AI exploded into view earlier this year, investors have been asking what it all means for the communications industry and Amdocs' role within it. So to help with that, we're delighted to bring today's webinar, which is led by Amdocs' Chief Technology Officer and Head of Strategy, Anthony Goonetilleke. Over the next hour, Anthony and his expert team will offer their unique industry perspectives in respect to generative AI and the many opportunities it presents for service providers to grow revenue, increase efficiency and improve customer experience. We'll give you a glimpse into some of the generative AI use cases Amdocs is already exploring in collaboration with our customers and strategic partners. And we'll also address some of the ways in which Amdocs is embracing generative AI to accelerate productivity and efficiency internally. So with that, I'll hand it off to Anthony and his team for today's webinar, after which we'll come back for a live Q&A session conducted via the chat feature in your browser. Over to you, Anthony.

Good day, and thank you, Matt. Welcome to our webinar. This is a really exciting topic for me. I think many of you have heard so much about it. I don't think there is a day that really goes past where we are not talking about generative AI and the impact it really has in society. In a recent report that was put out by McKinsey, some of the estimations seem to be an economical impact of between $2 trillion to $4 trillion that generative AI would really have on our society. And this is really mind blowing if you think about it. I cannot remember personally any type of technology really add this impact. Going back to the days of when computers came to be and the Internet came to be. I mean this is really one of those cornerstones in terms of changing the way we think about how we live, how we work and how society is impacted. So with that, I'm really excited to share with you a little bit about what we are doing around generative AI and how we are bringing it to the industry and helping our industry accelerate it together with our partners and really embracing the technology, but also just making sure that we have the right guardrails around it so that we can execute knowing that we do have an obligation to make sure that we're compliance in terms of regulations, security, privacy. There is huge brand equity that our customers are responsible for. So of course, we need to protect all of these. But that being said, I think it's probably one of the most exciting times we've lived in. I'm sure you've been hearing in the last couple of weeks about ChatGPT-4 Turbo and all of their features that they're launching, all the multimodal features of interacting with vision, all the things around Elon Musk and Grok-1 as that rolls out, and of course, every company, including NVIDIA, and Amazon, Anthropic and the list goes on. And so how do you embrace all of these technologies, but how do you still provide a mission-critical, reliable, stable, privacy where -- observability where a platform to be able to innovate and to be able to generate use cases. And this is really what we're focused on. And so we were -- several months ago, we were very excited to introduce what we call our amAIz platform, which really brings all of these together and really combines all of the elements so that our customers can start to develop it very fast. So they are not developing the building blocks that they need to on every use case. If you think, for example, of our strategy, our strategy goes something like this. The first pillar we're focused on is infusing generative AI in all of our products. So if you think about our CES suite right out from the gate, we've infused generative AI, working with some of the leaders in the field like Microsoft, OpenAI, NVIDIA to make sure that generative AI is incorporated in everything we do. So every rollout that's coming out, starting from September this year incorporated generative AI into it. The second pillar is what we call our use case factories. So we are using our amAIz framework so that our customers can build up on top of it, our customers can come to the table with the big rocks, the problems that they really want to solve and try to figure out how to use generative AI and accelerate what they're trying to do. Now we know if you just take the U.S., for example, and you take the top 5 customers in the U.S., we're talking about 500,000 to 600,000 employees in each of these customers. So there are a lot of opportunities to accelerate work to accelerate efficiencies. And these are some of the things that we're focusing on. Now the interesting thing is it's not just about efficiencies by ensuring that you're contextually aware of how a customer works and addressing their needs, we also start to improve NPS, improve customer satisfaction and deliver what we call a seamless customer experience that we're all striving for that really is the north star of our industry. So we have a lot to talk about. I have an amazing team that's going to share some great examples with you, both from our product set and our amAIz framework. As a company, we're very, very excited about it. But right now, I'm going to throw to Avishai, who's going to talk about our framework. Avishai heads, the technology group and has been a key partner in making sure we bring this to market expediently, and Avishai over to you.

Thanks, Anthony. In order to achieve these huge values, service providers must tackle the loop complexities, ranging from warehousing and cleaning the data to choosing an LLM provider to putting in place data safety guardrails to identifying use cases and beyond. Back in June, we were one of the industry's first to market with our amAIz framework. Then, and now we believe this is the easiest way for service providers to harness the power of GenAI while simplifying the complexities that comes with its adoption. But this is just scratching the surface. Let's talk a bit about the magic under the hood and why we think amAIz is so important. Let me start with a few words about the complexity of GenAI in the enterprise context. For telcos, they can't just use ChatGPT API and connect it to their commerce care and network systems. They need to deal with specific complexities and semantics of the telecom industry. We'll talk more about it later. Let's start with the underlying technology. The generative AI foundation is an agnostic layer that allows our customers to plug in to the relevant LLM infrastructure based on their use case requirements. With growing numbers of LLM [ permutations ] we have the expertise for selecting and orchestrating together the optimal LLMs for each use case. We work with all of the major players, and we're constantly evaluating emerging technologies to ensure we are able to take advantage of the best possible tool for the job. Once we get done with the foundational technology, we can move up to the fun part for me anyway, amAIz itself. AmAIz consists of a robust trusted AI module, a scalable framework, a deep taxonomy forged for more than 40 years in the industry, in-depth knowledge and a set of use cases, kids or as I like to call them, use cases accelerators. Let's start with the trusted AI module. This is critical because it's one of the hardest to get right. We need to manage out the bias in data, ensure that the modules act the way we expect them to. Not to mention consumption, feedback and response time, we need to monitor. Let's move next to the framework layer itself. In the framework layer, amAIz leverages industry-leading open source tools like LangChain that allows users to create prompt templates, integrate, collect data, orchestrate and produce out. And now to the magic, the telco taxonomy. This really is our secret sauce. It helps to understand telecom context and improve accuracy with augmentation capabilities to connect, optimize and enhance domain verticalized outputs. With these capabilities, we can introduce faster, cheaper ways to deliver use cases. Last but not least, in amAIz are the use cases [ kits ]. These are predefined templates that packages capabilities to address common themes such as search, recommend, hopefully, that helps provide a little context for why this is so important and why we are so excited about the potential of amAIz to support GenAI adoption in our industry. Now that we've explained a little bit about how amAIz works and the challenges it helps our customers tackle, I want to try and bring it to life. I'm delighted to be joined by Mustafa Oyumi, who's going to walk us through an example from our first pillar, infusing our products with GenAI. So Mustafa, how is GenAI helping us better meet the needs of enterprise customers.

Thank you, Avishai. This is a very exciting area for me as I believe it is for our industry. Back in February when we announced our partnership with Microsoft, we explained that we're going to deliver the industry's best customer engagement platform. I am thrilled to update today that we've made great progress, right? So first of all, let me explain the customer engagement platform. It consists of all key business components that you require to run a telco business essentially, right? It starts with a catalog. It is connected to marketing, to sales, to configure price quote CPQ, to commerce, to ordering and finally, customer care. And what makes the customer care and platform really exciting and, I would say, unique for the industry is that all these capabilities that I mentioned are enabled by AI and GenAI. A survey that we did with our customers showed that almost 83% of sales personnel or individuals use the CPQ, configure price quote, but they spent only 22% of their time actually selling, imagine that. The remaining time is spent on manual mundane processes, right? So this is an area where we thought we could make a massive impact, right, by injecting GenAI into the whole quoting process. A typical enterprise quoting process or ordering process usually takes weeks to months, right? For example, if you're offering an enterprise customer, a combination of fiber services or along with mobile services, along with cloud services, you can imagine that, that could be a very complicated bundle of services and products that you're selling through customer. As you're configuring each component that could be quite a manual process that could take literally days to configure every single component of that particular offer. What we have done is to inject and to leverage GenAI in a very massive way in our CPQ to streamline that process. We are literally taking that complicated product and automating the whole quoting process and reducing the time to take -- generate a quote from literally 6 weeks to hours, for instance, right? So this is an example of how we can streamline the entire quoting process and in fact, make the sales cycle dramatically shorter, right? I hope this explains how we are leveraging generative AI to streamline the whole sales process, right, to generate new revenue streams. Here, I'd like to ask Rima Khoury to come and tell us about how we can accelerate revenues using GenAI.

Thanks, Mustafa. Excited to take us through this. As you all know, our applications need to meet the needs of a wide variety of stakeholders from end user customers to partners. Amdocs Vindicia Retain product help enterprises and SMEs solve failed payments and improved customer churn rates. New business customers often want to know how much routine can help them, while existing customers often want to understand if a new feature can improve their subscriber LTV. To that end, we use generative AI to simplify access to information, such as allowing new customers to query our knowledge base to understand product capabilities and benefits. But what we think is more interesting is the potential for new customers to interact with generative AI to quickly analyze the expected benefit based on their transaction data. In order to make the assessment impactful, the tool provides guidelines about the key fields, period of time and more. However, Generative AI helps simplify the process such that the potential customer does it need to adhere to a very strict template when submitting their data. So in this example, a customer provides a redacted sample of their subscriber transaction data to the upload prompt based on the provided guidelines. Generative AI performs the submission analysis and summarizes the data just to ensure that there's no surprises. From here, the generative AI performs a predictive analysis, and the analysis show that based on the data model, the customer might have expected to recover 13% of the failed transaction at an estimated value of $1 million per month. Of course, how we do that is part of our secret sauce. But generative AI ensures that our potential customers understand the benefit of the platform. By asking how they can upload transactions Generative AI leverages the available documentation from the Knowledge Center to provide guidance to the consumer on how to submit their transactions. Because we have the context for the business and the subscriber insight, Generative AI is able to quickly perform the assessment of all the captured transactions to date. The magic, of course, is in the context of the customers and the predictive power of the model to understand the likely impact to their business. And that's exactly how we accelerate revenue growth and customer onboarding. Avishai, back to you.

Continuing our discussion on the first strategic prong infusing our products with the GenAI, I want to touch another example that's a pivotal system for many service providers, and that's the catalog. In a lot of ways, we think about the catalog as the brain of most CSPs offers and promotions. And within GenAI can make that brain even smarter. Let me briefly explain why we see the combination of GenAI and the catalog is unique and differentiating factor. If you are an Amdocs catalog user, you already have a great drag and drop business-friendly UI but we are leveraging amAIz to inject GenAI across the offer creation process. Instead of manually searching and comparing existing offers to introduce a more compelling offer, the user can simply use the catalog copilot and carry a conversation to obtain the most applicable offer to the market with their commended pricing and targeting. I want to shift gears now to the second prong of our strategy, creating a GenAI foundation. I already explained the role of amAIz in Telecom. Now I want to try and bring to the reality with a working example. If you -- as a user today, what to ask ChatGPT what to do is to save on your wireless bill, I would get a response that looks something like this. This response explains the steps needed to be done to provide an accurate answer based on the user's usage, the plans they are registered and the many more data points. You likely are not looking for this. We are looking instead for this. A much more context aware, personalized response to the same questions based on everything you as a service provider know about me. This may seem easy, but let me show you why it's not as simple as it's out. With amAIz, we pretrained the GenAI module to be telco verticalized and as such, understand customers' context. This enables to analyze large amounts of telco data and provide an accurate and trusted answer in regards to the user's bills, for example. At each stage of the response to that question, the system needs to understand telco business semantics, specific customer context and independencies in order to come to the right answer. This is where the magic of the telco taxonomy shines. The library of predefined use cases kits is critical to the ability to quickly and efficiently solve these challenges and because of our work with service providers around the world, we're able to proactively identify high-demand use cases to preempt the development of these capabilities. I hope that it helps illustrate why GenAI foundation is so important. Anthony, back to you.

Okay. Thank you so much, guys. That was great. As you can see, we are so excited about what we're doing in some of the demonstrations that you saw really bring to life some of these generative AI capabilities that we're so excited to bring to our customers and to our industry at the end of the day. One of the other pillars that I really want to talk to you about today was our partnerships. And these are critical around the foundation we're building of amAIz. Microsoft, NVIDIA, who we've just announced as well are both very key strategic partners, together with our broader ecosystem of other partners that we will announce in the near future. Of course, we're using various open source elements, and we're bringing these together in a mission-critical infrastructure that really delivers value at the end of the day. But now I want to throw away to Chris from NVIDIA to talk a little bit about our partnership together.

At NVIDIA, we have a long history of reshaping computing. Today, we're at the forefront of an AI computing revolution, specifically with generative AI. And as a leader in this space, it only makes sense for us to partner with a leader in communications and media to help accelerate and scale generative AI across this vertical. Working with Amdocs in the massive $1.7 trillion telecom and media industry, we're able to tackle some of the world's largest generative AI opportunities. We recognized early on that enterprises would need custom models to better fit their industry and needs. And at the same time, we believe the true value of generative AI rests in the ability to quickly and safely adopt the technology. Amdocs shares our same philosophy. And so we're thrilled that they're one of the first customers to use our AI foundry service, which brings together NVIDIA's unique enterprise-focused generative AI acceleration assets. The AI foundry service helps Amdocs optimize the large language models for its amAIz framework, translating it to faster generative AI adoption across the telecom industry. We are confident that working together Amdocs and Nvidia will be able to create market-leading generative AI capabilities that drive efficiency, scalability and new business opportunities.

Thanks, guys. That was great. And as you can tell, partnerships to us are critical in this journey as the ecosystem expands and as we try to deliver value day after day to our customers. And just to recap our session today, you've seen us talk about the pillar strategy. So we have the first pillar, which is really our CES pillar incorporating all of this generative AI functionality into our broad set of products that we deliver to our customers on a daily basis. The second pillar is our use case library, where we are working with our customers to use our amAIz framework and to be able to deliver them value through our use cases. And of course, you heard from our partners on where we're focused on really leveraging that partnership together with Amdocs and our customers to bring value to our industry. And last but not least, is what we do around our data strategy. Let's face it. Generative AI is only as good as a data that it ingests. So having the data ready, having your privacy and the authorization features, normalizing this data becomes super important to a really good generative AI strategy. And this is also an area where we're very focused on. So with that, I'm going to open up the line for some questions.

Well, that was great, Anthony. Thanks very much for that. And for sure, let's take some Q&A with yourself and also Avishai. Welcome, Avishai, Head of our Tech group, who's going to join us for this part of this session. Just to help unpack things a bit. I've got some questions of my own, but just as a reminder, for anyone listening out there on the webinar, if you would like to ask a question, you can definitely do that using the chat feature on the browser. So while we're waiting for some questions to come in, let me get the ball rolling a little bit, Anthony, I'll start with you. Look, I always think it's good for an investor to have some perspective on where we've been in the past and how we've gotten to where we are now. The speed of which Gen AI has exploded onto the scene this year as being phenomenal. Obviously, it's an ongoing journey. But talk to us a bit about why Amdocs is being able to move so quickly in this area to establish this leading position as the telco industry is expert in generative AI. Is it DNA? Is it [ hysterical ] and what is it that you would sort of put your finger on to explain how we've been able to create this position that we have?

Yes. I think it's a factor of a few different things. I think if you look at kind of what we bring to the table, we've always said we're a product led, right? So we invest a huge amount of R&D annually to make sure we have the best cloud native, 5G-ready, intelligence-ready products. So it's not like we woke up one day and said, "Hey, GenAI is here like, let's try and do something we said" no, we've worked with ML, AI, neural networks, we've worked with some partners in the last history of our product suite to incorporate these. So a couple of years ago, when we started to see kind of GenAI come into the foray, we said, okay, like how can we grasp this? We also know we sit on a huge amount of data, right? Our systems, our core BSS, which sits on the foundation of data that our customers have. So joining these together was just this kind of, I would say, like a natural pivot for us. And of course, the strategic partnerships with Microsoft NVIDIA just really help us accelerate this. So to us, it was like a no-brainer, right? Like it was just something you have to do. It wasn't an optionality. It wasn't it nice to have. And on the other hand, if you look at our customers, -- they have huge employee bases. They have huge amounts of data. They are mission-critical systems. So there's definitely a problem to be solved or a job to be done there. And so I think the 2 intersect very nicely between us and them.

Yes, it's remarkable how this sort of comes back to the product-led services driven unique business model that we have and our ability to constantly refine what we're doing. We talk about our early position in GenAI, but what does the competitive GenAI landscape actually look like to us? And what would you say is our moat around that and key points of differentiation?

Yes, sure. We are not out there trying to build the next foundational model or the next large language model, right? Like we use a cross-section of proprietary and open source large language models. And so I think our moat around is really the verticalization in the telco space. So the deep -- like Avishai spoke about, the deep breadth of telco taxonomy that we're injecting it -- the trusted AI components, right? So the explainability, the observability putting the moat -- the guardrails around the information, non-biased, which is a huge, huge issue when you're dealing with millions of subscribers, right? We're not dealing with [ 10 ]. We're talking about on any given day, Amdocs touches 2.8 billion, 2.9 billion consumers worldwide. So all of these things focused on the telco vertical is super, super important. And I saw like a question here from George that popped up. Maybe it's a nice segue into it a little bit. Yes, George, like I think there are some low-hanging fruit. But when you say lower-hanging fruit, I'll phrase it a little bit differently. When we go to our customers and say, "Hey, like we've got this amAIz framework. Here's what we can do with it. The discussion I have with them is let's discuss where your biggest kind of " the impact we can have on the cost center in the shortest amount of time." And these can be in call centers. This can be in call deflection. This can be in service delivery like truck rollout. And we have many discussions at the moment going on in parallel. And we have a use case library of -- we call golden use cases, and we tweak them a little bit for each of our customers. And we believe that we set a time period of, I call it, the 90-day window, right, like the 90-day window to deliver results in a POC. And then if that works, then we expand to be broader scale. So some of the low-hanging fruit, I think, is where the masses are. So where there is masses of data, where there is masses of people, where there is masses of costs these, I would say, are the low-hanging fruit because the more data you have, the more intelligence you have, we believe we can have a bigger impact, definitely call centers, service delivery, bill experience, customer care, those are the areas that we think is really "low-hanging fruit for us. Sorry, Avishai.

Yes. I would add the network domain being one of the first place that we see a lot of potential place for GenAI.

Yes. And there's a question here from Edward as well, if I can maybe jump flip to that one. Yes. Edward, look, the good thing about our amAIz framework is it doesn't necessarily have to sit on an Amdocs product suite or an Amdocs space, right? A classic example around service delivery and truck rolls, we don't necessarily have a product that manages truck rolls or anything like that. There are other partners we work with. However, we're working on a use case where it sits on it and you can provide intelligence and direct it. So on the broader scale of things, absolutely, we think that should have an impact in terms of addressable space where we can go after because it's not necessarily just the alley of BSS, OSS, things like that. So definitely, we hope and we think that there's some potential there.

Just tying Ed's question back a little bit to the Microsoft customer engagement product. And this is one of the use cases that must offer sort of highlighted around the CPQ, is to what extent is the enterprise just a massive opportunity? I mean it's a very sort of -- it's an area where really there isn't a lot of automation in terms of what we do on like maybe on the customer side, where we've been very good over the years and automating much of that process. Maybe you can just talk a little bit about that, Anthony, and how we're applying GenAI to the CPQ process to really drive some potential -- huge benefits potentially for...

Yes. Look, one of the questions -- I kind of jerk around that. I've had more C-level meetings with our customers in the last month on this topic than any other topic I can imagine in my 25 years in Amdocs, but one of the discussions always goes to, okay, so is it about revenue uplift? Is it about cost efficiencies, like where can generative AI help us. And of course, we know that there's a huge component around driving cost efficiencies in our customers in their like billing operations, customer care. But the example of the stuff I gave is an interesting one for me because if you look at the enterprise space and CPQ, if you look at today's world, you're talking about average sometimes of 40, 50 days of provisioning an enterprise order and collecting the money for getting the job done. We believe generative AI can accelerate the recognition of this revenue from 2 months to maybe a week or even less. So we believe that generative AI can play a part in acceleration of revenue recognition for our customers, which is also a very important thing. Let alone the kind of the NPS uplift and the satisfaction of a customer because here you are, you're sitting in front of a customer and you can generate at least 60%, 70% of a framework proposal rather than say, "Hey, thank you for the information. Let me get back to you in 2 weeks". So that's kind of the tie in from both ends, if you like.

Yes. I think another really cool use case is around the catalog and Avishai, I think you referred to that as being the brain of the CSP's offers and promotions. Can you talk a little bit about how that can really help service providers in the way that they accelerate time to market with new promotions and so on?

Sure. So I'll start maybe with the obvious thing to say, which is -- with GenAI, the ability to utilize the catalog is now open also to business people because now it becomes a straight for mechanism, a chat with a catalog as opposed to being a more technical person that needs to master a lot of in-depth knowledge to understand what is good and what is bad, this is one. But with GenAI and the catalog, you can also now create a very interesting segmentations and bring a lot of data coming from different sources. The marketing pillar, be the business pillar, be obviously, the underlying different components of the BSS and OSS solutions and create an offering which the GenAI can assist you with. A, by harnessing all the data and consolidating and B, by now trying to offer you as a business person or as the person that wants to create a new marketing offering with recommendations about the best fit based on a very tailor-made segmentation and market capabilities and needs. These are game changers in this industry because it will allow you to understand much better the segment that you are offering to. It will allow you to better cater them and it will bring you the best offer possible for a given scenario.

If Matt, maybe I think Edwards asked an interesting question here, if I can just jump to it for a second. Look, I think Edward just -- in this multiple things you raised here. But I would say, if you look at a macro perspective in terms of sales motion, 2023 was if I would say, from our customers' perspective, was the year of digestion, exploration experimentation. 2024 would be the year of productionization. So we are moving now from potentially POCs and seeing early results of going -- like I just did a review with one of our groups on a customer a few days ago. And we were blown away with the results because it's a result that none of us expected, those are very interesting, by the way. And we're like, wow, this is something that is very cool because finally, we have a view of customer sentiment, connected to customer history, connected to information from call center and customer care, connected to information with offers and looking at it very holistically. So I think if we look at it from a sales perspective, we'll be moving from POCs to productionization to broader expansion to multiple use cases in the next year. So that's kind of the trajectory, I would say. The comment about -- is an interesting one, but also a good one. When we started this process, believe it or not, we've been in the industry, we knew regulation was coming, right? You could not let this just run ramp it without having some type of guardrails. And we did it. We built amAIz with this as a foundation. So things like explainability, observability, knowing which data sources you use, knowing the priority and categorization of the data sources, being able to explain how this information was gathered together and accumulated, we knew that you had to do it because of the types of customers we work for, not even because some government was going to come out of the EU or anyone else than legislated, right? So I feel like we're a little bit of ahead because that's one of the design principles on our platform. But of course, as regulations come to bear, we are all over it. We're looking at what's coming out in the East Coast starting 1st of January. Of course, the use and discussions we're on several panels being part of the discussions. And to us, it's BAU, right? Like we are in a regulated industry in many different aspects. And we're okay with this, and we knew what was coming, by the way.

Great. And as we're talking with these customers and going through the POC process, I mean talking -- we've said that we're dealing with several of our flagship customers Anthony, what sort of a cross-section of customers are we actually speaking to where is the interest coming from globally [indiscernible] and so on?

Yes, it's a good question. So I can say from our biggest customers to first movers, that's where the interest is. So if I draw a spectrum and I would put like a smaller first movers on the right-hand side, and I would put our biggest customers on the left-hand side. Interestingly, those 2 is kind of where we're getting the first interest from, right, like not the distribution curve in the middle in terms of the masses. So I think that's okay. Our big customers realize there is a bigger bang for the buck, right? I mean they have huge call centers with hundreds of thousands of employees, huge amounts of customer care costs. Our small customers are like, "look, we can do something very cool. We can be a digital native in the space." So I think it makes sense, and that's really where I would say that the interest is coming from. Just one of the last questions here, Matt, if you mind, I'll just pick up here. Yes, sure, sure -- so it's for along, it's like what are the implications on GenAI on Amdocs organization and structure and routines. I would say that we are trying to reimagine everything, right? Look at how we do things, look at the impact on things like I'll let Avishai add here in a second as well. Avishai is rethinking the entire -- how the R&D organization operates, where you focus on, where you book people. And I think -- it's not necessarily like I think about digital, right, and back -- go back 10, 15 years ago, everyone needed to have a digital officer. And then you go, look, I mean, digital is pervasive, digital should be in everything. You should be a digital native in everything we do. And I think GenAI is also a little bit like that. It's not about having a GenAI organization. You may need to do that to kick it off to accelerate stuff. But at the end of the day, to give you a simple example, every ops review we have with myself, with [indiscernible] every group has asked to come and present what their plans in to incorporate GenAI into what they do, right, whether it be a corporate support function, whether it be a technology function, whether it be an operations function or even a marketing function, for example. So it's very, very pervasive throughout the organization. I don't know, Avishai, if you want to add anything from an R&D perspective.

Obviously, it changed everything we're doing. So in the sense of the way we develop, the way we test, the way we look into quality all of those are being injected with GenAI. So in every piece of "the SDLC process" within the software development life cycle process, we're injecting GenAI. So this is definitely something that is reshaping almost every touch point within the R&D life cycle. It has a lot to do also with new ideas. We're holding many hackathons people are coming with ready ideas. And the beauty about it is that some of them are materializing as we speak within a very short time frame. And suddenly, we find ourselves with new initiatives. Some related to stuff that are technology driven, but many are related to the way the organization is being the way we work and we form.

Maybe we can just drill down on that a little bit because we have talked about even though the growth potential of GenAI, at least for us is maybe next year, this year, at least, we're seeing some pretty immediate benefits that are contributing towards some accelerated profitability guidance that we've provided this year. And Anthony, maybe you can just drill down on that a little bit, certainly in the software development side of the house and maybe also services give some tangible examples perhaps as to how we're actually embracing GenAI.

Yes. Look, I think, first of all, when you map out your internal employee customer -- employee journeys in terms of all the different things they do. Going back to the example that Avishai is using, we mapped out our SDLC, our software development life cycle, and we said, look, here is an area where we can use generative AI at a very, very high level. And of course, there are other areas that you can't that you need a little bit more of like solution architecture and things like that, where you do need the human involved and you need their brand power and the creativeness, especially when it comes to creative areas, I think this is more of a sensitive area. But we believe we're constantly looking at automation previously to generative AI, right, in the same way as generative AI comes in. We're trying to look at how we can always contribute obviously to the profitability company, and we'll keep pushing the bar and increasing it with tools like GitHub, CoPilot and things like that, that come out. Of course, again, we're very sensitive in looking at IP issues and privacy and things like that you can't ignore these things, right? We -- remember, one of the things is we build enterprise-grade carrier software. So we're very conscious of that, and we need to keep doing that. But on the flip side, wherever we can use generative AI, this is something we will always do. There's a question here from Edward around kind of quantifying the cost structure in terms of using GenAI, whether it's APIs or data center or GPUs or labor. I would say a couple of interesting comments, Edward here. The cost models are evolving across the board. Some people are charging per tokens, other people have tiers. So we are -- we know that cost is becoming an issue very, very fast, right? It's one thing to say I'm going to just use X tokens. It's another thing to say, I'm using X tokens for Y transactions for Z, customers and then multiply it by 30,000 employees using it thousand times a day, right? So we are super cost-conscious. So everything we're building, actually, Avishai was telling me yesterday about a task. He's given all these general managers around focusing on the total cost of ownership including how all these components of generative AI are used. So we are, for example, using -- this is a key benefit of amAIz that we don't always use proprietary large language models. We use open source language models as well, like Llama 2, for example. And we use things like vector databases. So for example, I'll give you a simple, tangible example. If you're asking a particular question or a particular element, and you're happy with the answer, it's highly accurate and you know it. You don't have to constantly keep on asking it, right, and being charged for it, right? So you can retrieve it in a different manner. So this is where the architecture is constructed. And we're super, super focused on costs from that perspective. But I think the other answer is this is being played out. And I know I was at the AWS reinvent event a couple of weeks ago, and Adam was talking about lowering the cost structure on their CPUs and things like that. So I think both the infrastructure guys and people like us that sit on top of some of these stuff are very, very focused on driving down this cost and not this becoming another big cost center, right, because that's just counterintuitive to what we're trying to do.

This is also part of the design parameters of amAIz whenever we can, we're choosing the "cheaper" option. So we can use a different LLM if needed, based on fine-tuning some of the work we're doing. And by doing so, dramatically reduce the numbers.

Avishai, maybe you can talk a little bit. There's a question around what makes our telco taxonomy unique. Maybe you can spend a little bit of time sharing a little bit of your perspective on it being the father of this.

I think the -- there is a follow-up question. So why is it unique? And why can't others try to do it? So I'll start from the end. First of all, I'm quite sure that many will try to do it and many will try to make the same claim that they have at telco taxonomy. By the way, it's the same thing that happened before Gen AI in any other field, other can say that they have a better solution for whatever need. And this is the way the market behaves. So in terms of competition, I'm quite sure there will be competitors. The telco taxonomy, let's start to understand. What does it mean? I'll give you a very straightforward and a simple example. For instance, if I'm a financial person in an organization, PO means for me, a purchase order. And if you are using a telco taxonomy PO is a product offer. So this might be a very simplistic example. But when it comes to you trying to get some information and understanding that if someone is asking about PO, it might be related to an offering coming from the catalog or from your billing solution or from your charging engines or from your network has a lot to do with lost cycles, if you will not understand it, and you will try to find a solution somewhere else. Obviously, this goes much deeper into understanding the taxonomy of the telco, reducing the time it takes -- in the filters it takes you to understand what the question is and what are the possible alternatives to address a specific need. It also relates a lot to the underlying technologies and products that are providing a solution to a customer. So if someone is asking you about roaming, we know what exactly are the solutions and the systems that you need to tap into in order to have the right information A, save money, and B very accurate when addressing the question. As opposed to many others that might understand the overall question, but not necessarily understand how to tap into those data sources and how to understand the material and the data coming from the various places. And it goes on and on. It goes about understanding the taxonomy. It goes into understanding the different components that you need in order to assemble a question going back to a previous example that Anthony gave in the CPQ domain or in the B2B domain. You need to understand serviceability equals is my network ready? In 5G, would the [ slide ] be ready? It means I need also to understand what is the right quote that's coming from the right billing solution and not from a catalog, which is external, and then you're having different quotes and different promotations to the same answer. So it goes into a lot of things that usually people will do manually, many times, a lot of lab work and integration and all this is being solved by our tax economy and the amAIz framework.

There's a question here from [ Tal ] -- around the division of work between Amdocs and players like Microsoft in the market and asking if it's does it pan a new frontier of competition. Actually, I kind of see it quite the opposite. So for example -- if you look at the Microsoft Ignite event that happened a couple of weeks ago, [indiscernible], the NVIDIA CEO, and they are the slide in the background. And they were outside of Microsoft and NVIDIA, there were only 2 other logos on that slide. One was Amdocs and one was SAP. The reason Amdocs was on that slide is both of them see the benefit of having someone like Amdocs come to the table because #1, we are not building large language models, right? We are not building the foundational models that these guys are focused on. We are not building the cloud infrastructure. So when it comes to compute and GPU usage and large language models, that's not where we compete with. Where we compete with is taking that and making it pragmatic, making it relatable. An example -- another tangible example for what Avishai was saying, we ran a benchmark that showed on a -- I think it was -- correct me, Avishai I think it was GPT-4, accuracy was 60-odd percent. You add amAIz and you add everything that we bring to the table and you're getting to 92%. So it's a -- it's 30 plus percent jump in terms of the accuracy. Then you add to that all of the governance capabilities like observability, explainability, then you add to that all the data sources. Most of our customers, the data sources also come from us, right? So you normalize it, you ingest it, you feed it in, you train it using the data sources on top of foundational models, like if you're getting someone else to do it, they also need to understand the data. So of course, Amdocs understands the data on day 1, so we can run very, very fast on this one. And the other thing is also orchestrating it, right? So it's one thing to set in ChatGPT and say, "Hey, can you help me with my bill." It's another thing to automate these sorts automatically automated through the catalog through billing, through customer care, through CRM and actually have the actions go into execution. So we don't really see competition with. We're not in competition with Microsoft -- quite the opposite, same with NVIDIA. We're working with AWS. One of our customers, we're working with Google. So to us, it's like -- it's -- we look at it in a similar way that we went down the cloud journey.

Anthony, you touched on NVIDIA there and been a few weeks since we announced that collaboration and our involvement in the foundation there. But what's been -- what's the impact on the customer conversation as a result of that engagement? What sort of level of excitement or added confidence is it created in the way that we're advancing use cases and the way our customers are looking at the opportunities.

Yes. Look, the partnership, I have to say, has been fantastic. I mean for such a huge leading company, they have been absolutely amazing partners to us. We go to market. And we -- the discussions with the customers is we can come with the full stack, right? When I said I want to execute something and deliver results in 90 days. This is soup to nuts, right? This is getting your data sources, using a foundational model, using Amdocs amAIz framework, orchestrating the results and executing it. So being able to bring all of these together to the table, so you don't have to go and figure it out with 7 different partners to stitch this stuff together. That's really the value, I think that Amdocs brings to the table. And again, we come at it from a business results perspective. So we -- I know we speak a lot about technology. And as technologists, we're super excited about it. What I'm telling you that the start of my discussion, I had a discussion yesterday with the CEO of one of our large customers and the discussion, I would say, 75% of the time, the discussion was around business aspects of it, right, where you focus on it. So we take the business aspects of where they want to focus and where they believe they can deliver great results, and we translate that to the technology. Another thing, by the way, not related to technology that we bring to the table is we built a kind of a patented calculator to figure out what the cost benefit analysis is in terms of using a Gen AI use case. And underneath it, we have the amount of tokens being used, and we have this thing called a value multiplier. What's a value multiplier? Value multiplying in our model is something that says, look, a call coming into your center cost you X, we can execute that same transaction and resolve the issue with Y and the value multiplies essentially a division between the 2 and gives you a ratio. So when you prioritize your use cases, it's not just about, hey, I have a feeling that this can be nice. Like we come to the table with like 70% of your business case already built because we understand the data, we understand the metrics and we can help kind of build the plan and take it to execution from a technology perspective.

Maybe right to the last local strategies -- can you hear me now? So I'm saying I'll try to tie in Tal's question and the NVIDIA one together because I believe that NVIDIA brings also for us the ability to work with all players. NVIDIA is very strong and tightly coupled, obviously, with Microsoft and announced many things together, but they are working in their framework, allow us to work on all clouds. And by doing so, I think this also opens many opportunity for us working with other LLMs through the NVIDIA partnership.

To what extent is GenAI triggering a need for modernization more broadly within the CSB to move to the cloud and so on. Obviously, our customers needs to get ready -- they can't just adopt GenAI straight out of the box, right? So I presume there needs to be some broader investments as well? And how can we help?

Yes. When we talk about GenAI, of course, we're talking about 2 separate parallel streams, right? One parallel stream is if you take any of our modernized cloud native products, it is rolling out starting this month with GenAI capabilities built-in. So we have a customer that's taking a version of our catalog in March next year, that's going to benefit from all the GenAI capabilities that they're going to receive in production, right? The second stream is the use cases we've been spoken about that we can build on top. Now to benefit from all of these ideally doesn't have to be. But ideally, you have Amdocs new product set and you're running on the cloud, it's easy to do it. Now you don't have to do it. You can be on-premise, you can have legacy stuff. By the way, a customer, we're in the middle of a POC with has an older legacy version of Amdocs software, and we're still working with them. So it doesn't stop us, but absolutely, it generates another compelling event to drive transformations to you have that discussion to think about where your data sources reside, moving data from on-prem to the cloud. And of course, when you start talking about large language models, you don't want to build these in your data center, right? You don't want to start securing GPUs and things like that, you want to be able to leverage cloud infrastructure, infrastructure to be able to do it.

Very good. I'm not seeing any other questions in the chat.

Maybe I will add the beauty about amAIz is that it can work in all our permutations, which is also very strong. So it can support the new stuff of CES. It can support the on-prem, public cloud, and different permutations. So this also gives us a lot of credits with our customers.

Absolutely. We're coming up on time. Anthony, anything that any final comments or thoughts or a few points you would like to...

No, I think it's an exciting thing. I think what Avishai and the team have been building here is really, I think we build a lot of R&D and products and things like that, but this is really one of those -- we call -- there are some products in our portfolio that we call kind of cornerstones or controlling hills that are really -- the foundation for many transformations for many reasons our customers choose us. And amAIz, I think, is really one of the things that we're very excited about. Also because we think you can see a lot of tangible results. Sometimes you do an implementation of a product and you need to wait for a couple of years before you really see the full benefit. Here, we're talking about 3 months and you're seeing real benefits, tangible benefits to our customers. So that's really exciting for me.

Well, I think we are out of time. So Anthony, I'd like to offer a big thank you to yourself, Avishai, Rima and Mustafa for putting together today's event. It really was great and I hope everybody who joined find it useful. Apologies if didn't get to your questions but you can reach out to us here in the amdocs.ir department if you need anything further and with that have a great day. Thank you.