The Six Five Connected with Diana Blass: How Can Telcos benefit from Gen. AI?

By Patrick Moorhead - September 26, 2023

Telco services are vital to our everyday activities yet many operators face an uncertain future, challenged by the high volumes of data traffic on their networks and slow subscriber growth. Under pressure to develop new services and cut costs, operators have become the latest customer base to invest in Gen AI. In this episode of The Six Five Connected with Diana Blass, we explore the Gen AI solutions spearheaded by AT&T and DISH Wireless. We also chat with IBM about the infrastructure investments needed in a telco’s network and operations for Gen AI to thrive.

Be sure to subscribe to our YouTube channel, so you never miss an episode of The Six Five Webcast.

Watch the video here:

Or listen to the full audio here:

Disclaimer: The Six Five Webcast is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.

Transcript:

Diana Blass: Telco services are vital to our everyday activities.

Marc Rouanne: Telco has to work. People don’t understand when it doesn’t work

Diana Blass: Yet many operators face an uncertain future.

Steve Canepa: In the consumer space, the truth of the matter is average revenue per user really isn’t growing in most markets at the moment. But meanwhile, the consumption of data is growing dramatically, which means cost. So you have to move more data within a network, but you’re not getting more revenue.

Diana Blass: Telcos must expand into new business areas to better compete, and that means looking at their customers, their network assets, and even themselves in non-traditional ways. All now possible with the emergence of generative AI.

Andy Markus: There’s not a component of AT&T that will not use this to make what they do a better experience.

Diana Blass: It’s a technology that has the power to streamline network operations.

Marc Rouanne: So we are using generative AI to analyze and create language-based interactions with developers when it comes to the way we implement the network.

Diana Blass: It can also enhance customer interactions, open up new services, and even power the next wave of connectivity.

Daniel Newman: Things like future cars connecting to networks and even XR or augmented and virtual reality applications, which will require incredibly high bandwidth and throughput, but at the same time, we know will be part of the future narrative.

Diana Blass: A narrative that will be defined by the investments telcos make today, investments in their architecture, data models, and governance. Because as big as these opportunities are, the risks associated with the tech could be even bigger. I’m Diana Blass, and it’s time to get connected to generative AI for telco.

Every day the telecom industry manages massive amounts of data on its network. Depending upon how you look at it’s either a key asset or a big risk. Just think about the regulatory concerns and the potential for cyber attacks. Generative AI has put those risks in the spotlight for telcos. On one hand, generative AI adds to the security features on a telcos network through its ability to spot security threats and detect anomalies. On the other, it can introduce a number of data privacy concerns with potential leak and hallucination of sensitive data. It’s why experts urge telcos to manage and build their generative AI applications using a consistent platform with governance features as seen in IBM’s debut of watsonx for enterprise.

Steve Canepa: So think about this in the context of an AI plus world. I’m rethinking the way I’m going to deploy services by taking advantage of the AI capabilities, the generative AI capabilities, from the very beginning. So what kind of capabilities do I need in my enterprise to do that exceptionally well? We think there’s really three core things.

First, you need a studio that can help you consistently develop, tune these models so that they can be introduced into the enterprise, into the telco environment, in a very efficient and trustworthy way.

The second is, a model is only as good as the data that’s being used to train it. So do I understand all the important aspects about the data I have? Do I have a governance platform for my data that allows me to get to data wherever it happens to be in the enterprise if that data is critical to getting a better outcome from the model? Do I understand the lineage of that data? Where did it come from? If it’s used in the training of the model, do I maintain an understanding of that? These are all really important aspects of deploying generative AI in a telco or in any enterprise, for that matter. And based on what I just described is exactly why IBM launched watsonx.

Diana Blass: Steve Canepa is a general manager of IBM Global Industries. IBM was an early innovator with AI through its Watson tech. For background, here’s analyst, Daniel Newman, CEO of The Futurum Group and host of The Six Five Podcast.

Daniel Newman: The industries that are going to see the most complexity as it pertains to leveraging the potential and the benefits of generative AI are going to be the highly regulated industries. So healthcare, financial services, and of course, telco, all highly regulated. So while these tools can do everything from optimizing networks to driving more sustainable business practices, the privacy and data compliance that is required in telco is going to require a robust and well-designed solution that considers the needs of a telco, which are different than any other industry.

Diana Blass: Thanks, Dan. Watsonx provides telcos with a solid foundation from where they can build their generative AI applications, applications that can then scale across all areas of operations. The question is, where to get started.

Steve Canepa: I think there’s really four core domains that I see most of the energy and focus being put on today, and I think that’s where a lot of return or value lies.

So the first is in customer engagement. With these advanced systems where we’re marrying now generative AI into these processes, as much as 70% of the inbound traffic that’s coming in from a customer with requests or inquiries or meets could be handled through the AI processes. And what this meets to is probably somewhere in the range of a 30% reduction in the pre and post activities that have to happen to make that a really compelling customer experience. And then when you extend that to things like revenue generation, we’re seeing projects we’re doing now with clients drive 6-8% revenue uplifts because you can just focus in so much more effectively on how to actually enhance the relationship and value with that customer. So customer care, I think, are really cool.

Second, the network itself. We all know the networks now a software-based platform, and we’ve been infusing automation into the network for a long time. But as it’s grown, think about the billions of IoT devices that are now attached to a network, that threat surface where the network has grown dramatically too, to the extent that no human can keep track of all those potential entry points, those threats that exist in the market. So our security tools that are embedded for zero trust security into the network environment are now being infused with these generative AI capabilities, and that’s really critical.

Another thing is that just in the operations of that network. The number of alerts and the number of tickets that are being produced from that interoperability that happens between different parts of the network, whether it’s the RAN or the core, that’s scaling exponentially. So having AI figure out new patterns that are emerging about inefficiencies or optimization opportunities, those are all really powerful.

Another great one is thinking about the enterprise itself, the business of the telco, all the processes that happen. I’ll just give you a couple examples. Human resources, we’re seeing the opportunity for as much as 40% of those common tasks to get done in human resources to be automated through generative AI. And then think about the other aspect of the business, just what it takes to run a telco network. Truck rolls, out in the field, optimizing those trucks to be at the right place at the right time at all times. Thinking about vegetation management, oftentimes making sure that you’re getting those trees and bushes trimmed around those lines to make sure they’re efficient. We’re using video analytics now to do real-time analysis with drones on where they are. That can feed into that whole operation system and dramatically save money and make it more efficient to know exactly when you have to go do those kinds of tasks.

And the final one I’ll talk about is just IT, the technology piece of it. One of the models that we are talking about now that we’ve announced is the ability to take a lot of the older code that sits inside the telcos and use generative AI to modernize those code so the processes run faster, they run more efficiently, and that you’re able to leverage the skills of today in keeping the systems running. That has a tremendous productivity improvement, maybe 20 to 50%. And then the underlying IT, with AI operations driving the technology performance around ticketing or around usage metrics for technology, that’s another 30% savings opportunity in making the platform run more efficiently.

Diana Blass: Thanks, Steve. Telcos haven’t wasted any time to test and deploy generative AI. AT&T has developed an internal employee tool called Ask AT&T. It assists its workers in tasks ranging from coding to customer support to language translation. Here’s Andy Markus with more details.

Andy Markus: So it’s a really robust solution we have and we’ve taken that solution and we’re applying it into use cases like how we code for the company. Obviously generative AI can code out of the box, but with the right process in place, it can do anything from taking a user story, creating the unit test, creating the test data, writing the code, testing the code, and even on the backend, checking for vulnerabilities and adjusting for vulnerabilities. So it’s a full functioning process for managing the software development lifecycle.

We’re working right now on what we call the Ask Data functionality, where we’re connecting our internal databases, and in a human-like way, the subject matter experts can use generative AI to connect to the data and ask really sophisticated questions that you would always need a data scientist or a data analyst at your hip to do that. Now you don’t. So now, that subject matter expert in finance or legal or marketing can ask really detailed questions against data and get responses back near real time.

Diana Blass: Now, how challenging was it to train these models to a place where you’re comfortable with their answers?

Andy Markus: When we first started, and this feels like we’ve been on this journey for years, it’s been really since last November where it’s been supercharged, but applying generative AI to AT&T specific information, even with the right prompting and stuff, it was wrong probably 40% of the time out of the box. And so, getting the right information to it and getting it prompted in the right way to answer and in the context of AT&T took a lot of work, a lot of effort. And now that hallucination rate is very small, very manageable. But out of the box, companies have to go through that. We, as consumers, use it for our own personal lives, but when we use it for company specific applications, you have to make it smart on your company’s data.

Diana Blass: You’re also using generative AI in your customer help center. Now, this is an area where we’ve seen AI chatbots used for years. So how is this application with generative AI any different?

Andy Markus: We see the conversation… and of course the customer’s very aware of that they’re interacting with a machine, but we see the conversation being much more human-like than traditional NLP, that the conversation is engaging, that the person on the other end, the customer on the other end, is satisfied. And often, we end the call with a thank you, the customer is saying, “Thank you.” You never see that with traditional NLP.

Diana Blass: Telcos are also using the technology to bring new services to an enterprise customer, something crucial to the industry success going forward as consumer subscriber growth remains relatively stagnant. For that reason, Dish sees innovations around machines like smart cameras as a key investment area for generative AI. Here’s Marc Rouanne describing a new imaging solution under development.

Marc Rouanne: The 5G has been designed for machines. It has not been designed for humans. Of course, it’s better, it’s faster, but it’s really much faster than humans can see because it’s designed for machine, machine to machine. And machines need a lot of data. AI and analysis and automation needs a lot of data. And a lot of that data is about sensing the physical world. Sensing the physical world can be a video camera, it can be a sensor, it can be heat monitoring, it can be traffic management, and so forth. When you sense the data, especially when you do it through video, a lot of the video you have is useless because nothing is happening and all of a sudden something is happening. So if you want to extract that metadata and make sense of it and expose it back through a digital twin to a user so that they understand what’s going on without going through hours of video, then you need generative AI on video, and that can completely change the way you interact.

Diana Blass: So when do you expect to have this solution on the market?

Marc Rouanne: So we’ve built the architecture in order to deliver. What is difficult in our world in the telco is that it’s an infrastructure, an investment driven business, where you need to have the right infrastructure, a bit like the data centers. You need to build it for the purpose. And our purpose has been data-centric and data analytics from day one. It has not been consumers. Consumers for us is a given. In 5G, of course, you can do smartphones, but right from the beginning, we knew we could have a cloud native GPU-centric, data-centric network where we can put smart connectivity. So instead of just being a dumb pipe that is carrying a video, we are connecting the video to engines, AI engines and analytics engines and others, that are giving you the smartness and extraction of the data. So we are testing that.

And the speeds and gains we’re seeing are just crazy. It’s a complete change of paradigm in terms of what you extract and how you extract the data from those videos. I’ll just give you an example, video surveillance. So you’re sitting at a security desk and you have 20 cameras. Very hard to follow, very boring because nothing is happening until something happens. And when something happens, which camera do you look at? How do you mix the cameras? Well, we can shrink all of that into one view, which is a multi-camera view and like a 3D view, and then you can start moving around as if you were having your own cameras. And then we can extract what’s happening and the signals that are happening, alarms, threats, or other things that are happening in this 3D modeling based on cameras.

Diana Blass: Well, lastly, we need to touch on the investments that Dish has made in building a cloud native 5G network for the enterprise. Why is all this critical for modern day telco to remain competitive?

Marc Rouanne: Well, that’s very simple. If you use traditional telco to carry the amount of data that those AI models need, it’s just not competitive, the cost of carrying that data. Imagine you have 20 4K cameras for one emergency dispatcher. Who’s going to pay for all the traffic of those 4K cameras? So you need to have a new architecture that we call cloud native 5G standalone with deep edge processing in order to compress and shrink those costs.

So it’s really a survival for being competitive. Remember those networks were designed for voice or for human interacting with the web, which is so slow. When we go to a webpage, we read it, it’s extremely slow. AI is so much faster, so the amount of data is much bigger. So yeah, only the next generation networks, purely cloud native built on the distributed edge compute will be competitive. It’s just a fact.

Diana Blass: So it’s clear that generative AI is a vital tool for telcos. One that can drastically streamline operations, cut costs, enhance services, and grow an enterprise customer base. Seems too good to be true, right? Well, it might be if the right architecture, data models, and compliance standards aren’t in place. Generative AI could put telcos on a profitable path. The question is, will they be ready?

Patrick Moorhead
+ posts

Patrick founded the firm based on his real-world world technology experiences with the understanding of what he wasn’t getting from analysts and consultants. Ten years later, Patrick is ranked #1 among technology industry analysts in terms of “power” (ARInsights)  in “press citations” (Apollo Research). Moorhead is a contributor at Forbes and frequently appears on CNBC. He is a broad-based analyst covering a wide variety of topics including the cloud, enterprise SaaS, collaboration, client computing, and semiconductors. He has 30 years of experience including 15 years of executive experience at high tech companies (NCR, AT&T, Compaq, now HP, and AMD) leading strategy, product management, product marketing, and corporate marketing, including three industry board appointments.