Tony Poulos, TelecomTV (00:04):
Hi, Tony Poulos here at FutureNet World 2025 in London. Today I have with me Usman Javaid, who is the Chief products and Marketing officer at Orange Business. Usman, welcome.

Usman Javaid, Orange (00:17):
Hi, Tony, great to be here

Tony Poulos, TelecomTV (00:19):
Today. You were on a fabulous panel earlier today at this event, and we talked about, although we're talking about decoding the AI value chain, but how do you find that AI value chain?

Usman Javaid, Orange (00:31):
I think that's a great question. I think you need to define it from a customer lens. So let's put ourselves into the customer shoes. So any enterprise who want to get on the AI journey, what are the three things that they need? I think the first thing that they need is the right infrastructure. So you need to make sure that your data is at the right clouds or cloud or clouds. You have the right strategy for managing your data. You need to know that how your network is set up in terms of getting access to different type of data, which is sitting in different places. And then you need to have a cybersecurity layer across that to make sure that you are doing it in a way that it's secure by design. So that's the infrastructure piece. That's the first thing. The second thing that you need is now you need to get into this complexity of LLMs models, fine tuning, training, inference and so on.

(01:20):
And not every enterprise will really get it, and then every day you get something new. Recently I learned about MCP, it's a model context protocol. It's the way that the agents would interact with each other between the models and the IT systems. So that's sort of the layer. I call it an AI platform layer. You need to sort that out. And then you go to the most important layer, which is the solutions you want to solve specific problems and for which you need to have a gen AI powered solutions to solve those problems. Now if we step back, there are three layers. There's an infrastructure layer and AI platform layer. And the solutions layer as telcos, I believe we have a lot to say and a lot to offer at the infrastructure layer. Historically, we've been connecting enterprises and we are today connecting enterprises around the world.

(02:07):
We can offer them the right connectivity to the clouds, to their on-prem environment and offer this network as a service for different type of AI traffic. We can also offer them cybersecurity. I think we've been doing that for many years and AI takes the security challenge to even the next level, right? Because with AI you can manage a lot more incidents and alarms and threats and so on if you use AI in the right way. And then you have got, historically, some telcos have been offering cloud services and increasingly enterprises are worried about compliance, regulation, sovereignty, and hence we have an position to offer sovereign AI for enterprises. So a lot to offer at the infrastructure layer. Then you go to the AI platform layer. Now as telcos, we are all trying to empower our own employees with AI capabilities. We are investing in terms of a data platform, buildings, right? AI tools so that employees can use it. If you look at most enterprises, 70% of the employees today in enterprises are using public ai. So there is a risk of shadow ai, which we can fix by offering some sort of an AI platform capability. And then going to the solutions layer, you can be very specific in terms of offering some solutions in customer care, which is something we are doing with contact centers in the past or collaboration and productivity with AI that can be further enhanced.

Tony Poulos, TelecomTV (03:31):
You mentioned in the presentation today that's 73,000 of your employees are already using AI internally. How are they using it?

Usman Javaid, Orange (03:40):
Well, I think it's a very interesting story because chat GPT came out in November and less than one year after that, we were able to launch an internal platform. It's called live Intelligence for our employees to start using ai. We partner with the LLM providers, Mistral Lighton, which is the European ecosystem players. We partner with Meta, with Anthropic, with Google, Microsoft, AWS. So we sort of build this capability for our employees to be able to go in and just start with their specific use cases and start building those AI assistance to start with. And that's how it all started. And we decided to launch this platform to enterprises because there's so much valuable for us, it's equally valuable for our enterprise customers. So we launched that platform last year. Now your question about how they're using it, well, it's very difficult to define it because we are given this tool to employees to be able to make the best out of it.

(04:36):
But in terms of use cases, we have software developers using it to write code. We have got marketing teams to use it. In terms of content generation, we have generally employees using it for productivity in terms of creating content, analyzing content. We have our sales team using it to creating sales proposals, going from the RFQ to creating proposals. We had our pricing teams to be able to predict the right pricing. So there are tons of use. We have HR teams who are using it to write job descriptions, to do the analysis of the market in terms of the talent, and then identify where we can spot talent and go about them. So there are many, many use cases.

Tony Poulos, TelecomTV (05:16):
Well, of course the big issue around AI is always data, but we don't always have structured data. How are you going about building use cases, presumably with unstructured data?

Usman Javaid, Orange (05:27):
Yeah, well, I think I give you one use case which is about customer care. So you have all those incidents and then you need to manage all those incidents and you need to sort of resolve those incidents, infrastructure related incidents. So we have a use case around that. Now you can take one approach, which is about taking all your ticketing systems and put the data of all the ticketing systems into one data lake. That could be a year project and then optimize the data governance around that and show data quality and so on. So by the time you have a use case, you're already in year two or year three. This is one approach. Gene AI helps with that because you can take multiple data pipelines and feed that into the large language model, which is by nature designed to manage with unstructured data. And then we were able to get to this use case in a couple of months, which may not give you the right accuracy as it may have given you with the two, three years of effort of data cleaning. But you get to outcome faster and then you can fine tune your model, you can feed your model with some good quality data in terms of improving the accuracy. So I believe gen AI bring an advantage of faster experimentation because you need good quality data, but you don't have to spend ages to clean the data before even you get started.

Tony Poulos, TelecomTV (06:49):
How do you open up AI to the market? You building these platforms for the use of your customers, but how do you open up to them? And you mentioned that 70% of enterprises are using public ai, but you obviously want to roll them into the way you are operating. How does that work?

Usman Javaid, Orange (07:05):
Absolutely. So I think you may have some enterprises who would like to control their destiny and they want to build their own AI platform. There are many enterprises, they want to just focus on what they're good at and they don't want to do undifferentiated heavy lifting for the things that others have already tried in the past. So we have done that. We have sort of rolled out this platform internally, it go to the right scale, and now we are bringing that to enterprises to say, okay, you don't need to start from scratch. Take this platform and we can help you to build use cases on top. And then I think the way we have built it is that you can decide where you want that data that you have to sit. If you are very concerned about the trust and sovereignty, we offer you an on-prem environment with GPUs in that environment.

(07:51):
So you can decide to run your models, you can choose your models, you can run it in that environment or you can use it in public cloud. So we give you that choice that helps you not to get logged in into any ecosystem, right? Historically we hear about this hyperscaler lock-ins. Now we have a new lock-in, which is the LLM provider lock-in. If you go to open AI and you start using their tooling, you are lock in into that. We free you up from that lock-in with that platform capability. So you can choose what type of model you want, you can choose what type of cloud you want to deploy. And by the way, given we use and our employees use that platform internally, we have already pre-built prepackaged use cases, which we offer what we call assistant library and agent libraries. So if you have a use case first, check the library, you may already have an answer or 80% of the answer.

(08:40):
So we have a lot of enterprises, mid-size enterprises, some very large enterprises and public sector for example, who are using this platform to do many things. We have one of the public sector customer who have built an HR assistant because they did not enough HR capacity to be able to respond to the employee queries. And they have built an HR assistant on this platform. We have another public sector enterprise who wanted their software developers to be able to use genai to generate code, and they're using it for that purpose. So there are many different enterprises each come from their own angle.

Tony Poulos, TelecomTV (09:14):
It's interesting because a lot of the telcos that are here are not even at the monetization stage. And it sounds like you are reaching that already with these models that you're creating for customers. But you also mentioned that AI leaders have a foundational platform. Secondly, that they don't compare humans with agents, human capability being augmented by ai, which is very refreshing to hear. Thirdly, they lead with responsible AI governance involving people from all departments. And of course then we've got the cultural change that's involved. How is orange or orange managing this internally and helping their customers manage those things?

Usman Javaid, Orange (09:50):
And Tony, I think it is difficult, but let me give you my take on this. I think it need to start with the very top down mandate and ambition. A strong ambition that what AI is for us as a company, right? And it has to be balanced between what it is to drive efficiency and what it is to drive growth. And I focus on that growth point because many telcos I feel have already given up and just focused on efficiencies. And in Orange we took this very strong stance that it is about efficiency and it's about growth.

(10:21):
So that's the top down mandate. Then the second thing that you do is to make sure that you enable your employees, you educate them, you give them training, you sort of show them the path, and that is very, very important so that you upskill them for that journey. And the third thing, when they are enabled, you need to empower them. And this comes to that platform approach. When you give them the right tools, when you give them the right intuitive environment to be able to build stuff, they don't need to be AI practitioners or AI experts to be able to build their AI use case. If I'm a managing incident, I'm an incident manager, that's all I know. I'm not an AI expert. So if I've been given the right tooling and platform capability and empowered in the right way, I'm enabled in the right way and I've got my management who give me the mandate to be experimenting, then things get easy. So I think it's very easy to be said. It's very difficult to be done, but I believe we got to a point of this journey in two years and we got now the momentum that we are reaping benefits of that.

Tony Poulos, TelecomTV (11:22):
Well, it sounds like you're in the right place with Orange business. Usman, So glad to be with you today. Thank you so much for the insights.

Usman Javaid, Orange (11:29):
Thanks Tony. Thanks for having me.