In this episode of Carrier 2.0, host Steve Saunders explores what it really takes to build an AI driven network. In a rapid expert montage, Cisco’s Guru Shenoy, Rana El Desouky Kazamel and Masum Mir explain how AI agents will reshape traffic patterns, why the edge becomes critical, and the architectural pillars carriers need to put in place. Verizon’s Yago Tenorio offers a clear-eyed reality check on Open RAN. The episode then shifts to a “Truth to Power” segment from China, asking what AI and automation mean for human workers, and whether a hybrid human plus machine model represents the most ethical and resilient future for Industry 4.0.
Steve Saunders:
Hi, and welcome to a chilly New York City. This is Carrier 2.0. In this episode, the AI is coming. And what that means for carrier businesses and their networks.
Guru Shenoy:
All of these are new characteristics on the network driven by AI that we haven't seen to this scale before, and it's coming really fast at us. So the network has to be prepared, and this is what we are building for.
Steve Saunders:
Also, Open RAN myth and reality, separating fact from fiction.
Santiago "Yago" Tenorio:
If your vision of Open RAN was like a playout of small players that you can kind of mix and match as you wish and it's plug and play, that was never going to be the case.
Steve Saunders:
And what does AI and automation mean for your job? A lot depends on which country you work in. But let's begin with the topic of the decade, maybe of the century. I talked to leading experts at Cisco about the reality of running artificial intelligence over carrier networks. It feels to me like we're overweighted on AI in the core of the network, but isn't the future of AI, it's at the edge of the network, isn't it? Because it has to connect to something, doesn't it? It takes place where people and robots live out at the edge of the network. I mean, is that how you see it?
Guru Shenoy:
100%. You're right about that, because for all the talk that's happened around AI and all of the excitement, a lot of what's happened in the recent past has been about what I refer to as the production of AI, large language models, massive training clusters, and it was really the big hyperscalers and large AI data center builders that were doing that. But now we are started moving into the age of consumption of AI. And this is where we are starting to see agents that support humans, support machines, support enterprise applications that are leveraging AI to do things differently, and we will start seeing the impact of that on the network and on just the way we do things going forward.
All of these agents are going to be talking back to maybe an inference cloud or some other cloud location. Now when you have these millions of agents potentially on a network talking at various points in time, and not just talking in a chatty manner, a lot of them will be doing tasks that often run into minutes, hours, or even days. So that means a lot more sustained traffic coming from these endpoints across the network, and this traffic needs to be secure. This traffic is going to go into different directions, because there isn't a single location where it's going to be serviced. These agents will reach out to multiple inference clouds. So your network needs to be able to handle that kind of multidirectionality of traffic. Some of it could be extremely latency sensitive. So all of these are new characteristics on the network driven by AI that we haven't seen to this scale before, and it's coming really fast at us.
Steve Saunders:
New network traffic patterns require new architecture. Here's how Cisco is tackling that part of the puzzle.
Rana El Desouky Kazamel:
And first of all, we need to look at what's really changing from a networking perspective when it comes to AI. We'll see that there's going to be a lot more upstream traffic than we have today traditionally because you have a lot of data being sent upwards to those AI LLM models, as opposed to before when we relied a lot more just on content caching and traffic going downstream. So that's one change. Second change, obviously there's going to be a lot of work happening at the edge. So AI inferencing, where service providers will be really in the midst of this, is going to be quite key, which means that we need to distribute more the edge towards the end customers and make sure that we're building the right platforms to fit in those locations to enable basically networking, security, and automation to be able to deliver on those new AI services.
And if we take a look at it, we've got three main pillars that we're working on. The first pillar is all about best-in-class networking when it comes to silicon that's very high performance and very energy efficient, but also having all the right services that are required to go all the way down to the edge. Second pillar is on AI agentic operations, which is absolutely key for us to get towards autonomous networking. And then the third pillar is security, and how do we infuse security more into the network fabric? And that's going to become really critical. The era of AI is actually going to position service providers pretty well because the network asset, the power and the data center assets that service providers have that are very distributed across every single country are going to be absolutely critical.
And then service providers having all those facilities and capabilities is going to become quite a critical asset. And then leveraging that to really come up with services to cater for the enterprise and the public sector set of customers that care a lot about security, care about data sovereignty to be within country, and also for us to start delivering very low latency use cases that are going to be needed, because in a multi-agentic framework, you're going to find that you're going to have a lot of agents communicating with each other to reach to an outcome.
Steve Saunders:
Well, once they brace for AI impact and put a new architecture in place, what's left? The big one, making money.
Adam Sandler:
I have no experience, but I'm a big fan of money.
Masum Mir:
But right now, we're at a point that we also have to go and leap and we have to transform ourselves. Start to think more outside in, take what we have learned connecting the entire population of a country, and build for things that is beyond just people connectivity. But you have to actually think about those verticals as the IT OT in the enterprise comes together, how you show up differently, how we start to think about solving the end customer's problem more holistically. Maybe start with few verticals where your brand and your presence is already strong. Like everyone, like every other industry, service provider has an opportunity adopted fairly fast for their own operational efficiency and customer relationship. Continue to invest and build a foundation, not only for your own efficiency, but be on the valuation of the AI traffic and agentics. My bold prediction is the big winners in telecom, there will be ... Not everyone is going to be big winners, but big winners will see majority of their revenue shifted from consumer revenue to business to business revenue.
Steve Saunders:
There's a lot to take in, obviously, so before we tackle the equally important topic of Open RAN, why not relax with this animated graphic showing the growth in the number of satellites around the world since the launch of Sputnik? It's not terrifying at all. Open RAN has been a controversial topic since it debuted seven years ago. To find out where is it now, I talked to Yago Tenorio from Verizon.
Santiago "Yago" Tenorio:
If your vision of Open RAN was like a playout of small players that you can kind of mix and match as you wish and it's plug and play, that was never going to be the case. And I think as carriers, we were all very clear that was not what was most likely going to happen. The paradigm of plug and play without any interoperability testing is just not possible. Of course you can mix and match vendors, but it doesn't mean that you can do that without proper testing. If you forget about that and you think of open interfaces that allow you to actually integrate different radios from different suppliers, and as a carrier, it allows you to procure radio units from different suppliers and then integrate them onto the software that you're buying from a third party, that has already happened.
Look at Verizon. We have over 30,000 base stations with Samsung software running on completely totally off the shelf tenders. All the interfaces towards the radio units are open. The answer to your question is that there is a nomenclature problem. It's not what you think, it's not what you wanted it to be, it's what it needs to be and it's happening before your eyes. So most of the industry is thinking, "Well, that's not happening anymore." Well, just wait.
Steve Saunders:
So investors can't get enough of AI, but what impact will it have on the human workforce? Time to speak some truth to power. Let's throw it to our man in China. Digital industrialization is shaping the future of the global economy, but what does all this mean for the human beings working in these industries or served by them? That's both an existential question and a more nuanced one. The answer hinges less on technology and more on the business philosophy of the executives overseeing these industries and the governments that regulate them. The key factor that will determine whether industry 4.0 is merely beneficial for customers or also a force for good is whether industry players harness AI and automation to eliminate the human workforce altogether, replacing them with robot armies.
Speaker 9:
Activate the droids.
Steve Saunders:
Now, that's clearly the default goal of big tech companies, including Uber and Amazon. But fortunately, this trend is being countered by an encouragingly large number of industries, including telecommunications, fiercely committed to keeping humans in the loop.
George Glass:
I'm not advocating a big buying approach to take your entire network and try and move from where it is today to level five. It's more likely, one, that you'll build it out incrementally. Two, you have to bear in mind that most of our operators have legacy technology and everything that we do within the forum, we try to make sure that the legacy technology can coexist with the newer technology that you're rolling out.
Steve Saunders:
They've worked out, correctly, that the safest, most efficient, and ethical use of automation is as part of a sort of hybrid mode, a kind of distributed cyborg system, if you will, where technology is used to support and enhance the capabilities of the world's human population.
Lisa Simpson:
Shouldn't you be working?
Homer Simpson:
I got someone to cover for me.
Steve Saunders:
So that's it for this episode of Carrier 2.0. I leave you with actual footage of one of the very first carrier implementations of 5G.
Speaker 12:
You're not going to get hit. I promise you. You've got it, son. You've got it. You've got it. You've got it. Keep going.
Steve Saunders:
I'll see you next time.