Inside Amdocs AOS: How Unified AI Reinvents Telecom

What if a telecom could think with one brain? We sit down at MWC with Gil Rosen CMO from Amdocs to unpack how the company is shifting from building point solutions to delivering AOS, an agentic operating system that orchestrates many AI agents as a single, contextual layer across the business. Instead of bolt-on automations, AOS connects decisions end to end—governed, explainable, and secure—so outcomes are consistent and measurable.

We walk through why domain expertise matters: telecom-specific ontologies, prebuilt process libraries, and open integration with non-Amdocs BSS, financial, and logistics systems. Gil explains how trust becomes non-negotiable when AI touches money and identity, outlining controls for repeatability, policy adherence, auditability, and scale. Then we go practical. With Market One data across tens of millions of subscriptions, operators can see which OTT bundles trend, how device cycles forecast upgrades, and where local behavior shapes winning offers. The result is smarter pricing, higher ARPU, and faster testing without rip and replace.

The conversation takes a human turn with “waiting is gone.” Voice-first agents shift service from taps to talk, powered by two new layers: personality engineering to encode brand tone, and a customer digital twin that aggregates thousands of signals to tailor language, depth, and empathy. A techie hears throughput and latency; a casual user gets plain guidance and reassurance. Early results are striking, with personalized agents showing up to 3x NPS gains versus generic interactions. Beyond speed, customers feel understood—and that feeling moves retention, referrals, and revenue.

Change is urgent. Three years ago the story was 5G; now it’s outcomes customers can feel. AOS offers a collaborative path forward, coordinating multi-vendor agents and existing stacks while upgrading the experience layer. If you’re ready to reimagine service, unlock new bundles, and replace hold music with helpful conversation, tune in—and then share your take. Subscribe, leave a review, and tell us what part of the customer journey AI should transform next.

Transcript

From Tech Solutions To Intelligence

SPEAKER_01 0:01

Hey everybody, I'm at MWC with Gil from Amdocs. Gil, how are you? I'm very well. How are you? Congratulations on an amazing presence at MWC again. You guys certainly kill it every year. But this year has been extraordinary. So much new. Before that, how do you describe Amdocs these days?

SPEAKER_00 0:21

I think what we're transforming into is from a company that created technology solutions into a company that provides intelligence to run the business and operations. And intelligence is a different layer than a function or a capability. And it's much more profound and it's much more embedded throughout the organization. So I think we kind of rose a level and actually created a level that didn't exist before in the IT stack, because the IT stack was kind of a connection of different subsystems. And now between them, we need to embed intelligence to create one thinking mind, and that's what we bring to the table.

SPEAKER_01 0:55

Fantastic. And tell us about your agentic strategy. It's rather unique, rather differentiated. Give us a bird's eye view.

Telecom Context And Open Ecosystems

SPEAKER_00 1:01

Yeah, so as you see here, we're all about AOS. What we realized and what AOS is about is the evolution of agentic capabilities. When we started as an industry and as Amdocs, we looked at use cases and we proved that agentically or with Gen AI, we can actually create efficiencies in vertical processes. But to scale, you need to actually connect everything. You cannot just do bolt-on intelligence, you need the intelligence to be all connected. If you think about it from a paradigm perspective, it's like you cannot have a brain in my hand, a brain in my feet. You need one brain that actually sees everything and controls it in context. So we basically moved up and said, okay, how do we actually create a layer that for once is able to orchestrate all the different agentic processes in an intelligent and contextual manner? The second level is, or the second pillar, is that this layer needs to be very telecom-specific. You know, we come with decades of experience of running telecoms, and the ontology, the context is built in. We bring in pre-built libraries of processes, of agents, of sub-agents that really understand how to run the business. And I have to say, it's also not about AMDOs. We have to be able to collaborate with different agents from third parties, other vendors, our customers' agents. The third part, which is probably one of the most important, is trust. How do you trust that the agent does what you mean it to do? There is uh, I think in the last uh year or two, what's been really profound about the technology is that you're looking at a technology that sometimes you cannot explain. You know, when you when you converse with ChatGPT, it's like it's like magic. And we are not an industry that can trust magic. We need to see the behind the scenes explainability, repeatability, the ability to configure, the ability to governance. So all these and scale, and of course, security. So all the kind of the foundations that enable us to go on scale. The last thing is again, uh our customer base is global. Our customers are on different journeys or different stages of their IT journey, which means that we have to be able to coexist on any BSS system, amdocs or otherwise, within the Amdocs ecosystem on any version. If it's a if it's non-AMDOX, on non-AMDocs. And of course, telecoms run on multiple systems, not just uh BSS, they have financial systems, logistics systems, and we need to be able to interact with everything. On top of that, we also bring third-party data, which is which comes from our different platforms. I'll give you an example. We we have uh Market One, it's our subscription platform. We already have something like 50 million plus subscribers running on it. We know which third parties are trending, we know which OTT apps are doing well, we know it, we know it also based on consumer behavior, so we can actually recommend to our customers listen, this bundle can work better than that bundle. You should take HBO Plus versus Netflix or Netflix versus NetHBO. We've seen in a different country that a bundle between Disney and app of X and this works well. So we also bring a lot of rich data to the you know to the ecosystem. And I think it's the first time when we're actually really showing the open modularity architecture, but not in an IT architecture perspective, but how we plug in and create a I would say Uber intelligence that makes the system run. All of this translates not into again an architecture but a different model. This model is in fact, if you think about it conceptually, is an operating system or an agentic operating system, therefore AOS. That's that's my short answer.

Trust, Governance, And Scale

SPEAKER_01 4:48

Wow, that's that's a mic drop moment, but I'm not gonna drop the mic. I do have a couple more questions. Go for it. Um, lots of talk about AI with productivity and saving money, reducing capex, etc. etc. But tell us about the upside in terms of new revenue opportunities, new customer and consumer experiences, and improvement to some of the old models we've seen with telcos in the past.

Market Data And Bundle Insights

New Revenue And “Waiting Is Gone”

SPEAKER_00 5:10

That's a that's a great question. You know, we have one of the taglines here is uh waiting is gone. And what do we mean by waiting is gone? The paradigm, my my you know, my 17-year-old daughter, she's not gonna wait on the line for anything. She wants an answer from a service provider, she just needs to open the phone and start talking to it. So we bring a different, you know, this is a different uh approach to customer service. Also, we introduced a concept called customer digital twin. We actually understand, okay, a step a step back. You know, what you and I are doing now is talking. We're not uh pressing buttons on an app, we're not uh browsing through a website, which were the traditional customer experiences. If you want to find an offer from your service provider, you have to go into their app or to their website. Spend at least at minimum two, three minutes finding your information. What we do now is allow the customers to talk. Now, talking has never been personalized because we've had people running the talking, and people know how to personalize their conversation because this is what humans do. When we're launching agents, we're actually, and I know it sounds kind of science fiction, we're launching humans. As far as the end customer is concerned, he's talking to a person to uh an avatar that's for him relatable in the same way that we ask very personal questions to ChatGPT. So we created an entire new domain that has two sides to it. The first side we call personality engineering. How does your brand, how does your agent stay on brand? How does he manifest the brand archetype, the values into the conversation? The second part is what information he uses to relate to every customer differently. So using all the data that we have from various data sources, Andocs and otherwise, all the way from the network, through to the devices that are used, through to the behaviors of, for instance, do you change your device every time a new Apple, uh a new iPhone comes out, or do you wait three years and just get uh, I don't know, like a generic uh uh uh Android device, not to uh offend anyone. So that means that you're a different type of person. We take that into consideration. So we aggregate thousands, if not hundreds of thousands, data points into creating a single customer digital twin that our customers can use, and then the agent, before interacting with somebody, says, Okay, Evan, you're you're a techie. I'm gonna go and drill down and talk about gigabytes and throughput and maybe speed and use terminology I would never use with if my dad called a call center. For him, I'll do plain and simple, plain vanilla, just the bottom line. Do I need to be empathetic? Think about it. We need to create software that's empathetic, that's never been in our vocabulary, not as AMDOs, as an industry. We've built systems. Now we're building systems that have a psychology. So that needs to be baked into the interaction, and that's a new domain. It's also super interesting because it's a new layer to the service provider architecture. Service providers currently do not have cognitive information about their customers, they have statistical information, they have actionable information, they have transaction information, but so they know what to say. They don't know how to say it. The CSR, as a human, knows how to say it. But even then, do they speak the right language? Do the right tone of voice? Did they have a tough day before? Did they wake up late? Are they just, you know, first time on the job and they're not an expert yet? Replacing that will be an agent that's first in the entire data set of the organization, knows exactly all the policies to the last point, understands when to pull what and also how to converse with a customer. And I think again, that's gonna create a totally different experience. As a coming at it from the CMO side, NPS is not a statistic I think the industry has been proud of. And what we try to show, and we already uh proved it in uh research that we did, that an the NPS for an agent that's personalized is three times threefold, then an agent that's generic. Wow, that's a huge benefit for the customer just from the experience. You know, I think many people perceive telecoms, you know, one of the things that we do in our job, and I'm gonna kind of take a step back, we make we connect the world. It's a very tough thing to do. But customers, it's like, oh, it's there, right? It's like Wi-Fi just exists.

SPEAKER_01 9:52

Phone calls 100 years old, you know.

Customer Digital Twins And Personality

SPEAKER_00 9:54

They don't understand the amount of assets and compute and software that needs to make this seamless. Now, on top of that, what do they think of the so they don't give us credit for that? What do they think about us? How was our latest interaction in the call center? I had to wait five minutes, ten minutes, thirty minutes, the CSR couldn't resolve my issue, it took him five minutes, he had to call the supervisor. That could be gone. So we're really looking at uh like a turning point in the industry where we can create a different level of customer service, and I think it'll make us a much better service provider, and we're there to help our customers do that.

SPEAKER_01 10:32

Fantastic. Final question. Um, there seems to be a sense of urgency in this industry now to reinvent, to transform in a way that I haven't seen in the past. Um, how are your customers keeping up with all this incredible technology that you're developing and offering them?

SPEAKER_00 10:49

Yeah, it's a really good question. I think it's um unlike anything I've seen before, we talked about it a little bit before. If you would have kind of zoomed around here three years ago, everything was about the network. 5G, 5G, 5G. Like I mentioned before, our customers are oblivious to it. And I think what our customers realize is that they need to turn their focus into the outcome, the experience with a customer. So we're with them on this journey. By the way, you know, humbly, it's also a new journey for us, but you know, we've been at it for two years, and uh now in the AIA, just like dog years, like two years, it's like 14. So, you know, the the Chat GPT launch in November 2022 seems like uh a decade ago. And ever since then, we've been on a on a race, we've invested a lot of dollars, effort, personnel, and thinking strategically for our customers. How can they actually get it done without, and this is really important, it's not a rip and replace, it's not a unless you take Amdocs, you can't do this. It's a different world, it's a world of collaboration, it's a world of ecosystems, and we're there to help them make it happen sooner rather than later.

SPEAKER_01 11:57

Well, congratulations, what an amazing journey. Can't to see what uh wait to see what's next.

SPEAKER_00 12:02

Yeah, uh thank you. Thank you.