AI and Open Networks – What CSPs Need to Know

The convergence of AI and open networks is reshaping telecom infrastructure as we know it. CSPs see tremendous potential for the automation, optimization, and predictive analytics that AI can deliver, which will create new opportunities for growth.

MeriTalk recently sat down with Manish Singh, chief technology officer for Dell Technologies’ Telecom Systems Business, to discuss the key considerations for CSPs as they develop business models and AI strategies to position themselves effectively to realize the full benefits and value AI can deliver to their business.

MeriTalk: With the rapid adoption of AI, how should CSPs start thinking about AI relative to their network transformation plans?

Singh: When we think about AI and the transformation opportunities for CSPs, I would say the sky is the limit.

CSPs should be thinking about AI in three key areas. The first is revenue generation. In many markets, especially in Europe and Asia Pacific, telcos have already started to set up sovereign AI capabilities, which are tuned to local language and culture, as well as local data privacy, security, and sovereignty issues. These sovereign AI capabilities are creating tailwinds for CSPs to assess how and where they can provide new revenue-generating services such as AI-as-a-Service and GPU-as-a-Service.

The second area is the network itself. We’re seeing operators build digital twins, including closed loop automation, for use cases such as prediction. Eventually, as they get to higher levels of network autonomy, they can drive a lot of operational costs out of the network, improve network up times, and reduce downtimes. So, there are tangible operational benefits on the network side with improved productivity and performance, as well as energy efficiency.

The third area of opportunity is customer engagement. We’ve had chatbots for a while, and now with large language models, the next generation of chat agents are emerging.

As an example, we have been working with SK Telecom across the full customer experience lifecycle with customer care solutions that enable CSPs to offer new services, new devices, and new plans. It can all be done in a very automated fashion, because now you can capture more context and better engage the customers to deliver those services.

MeriTalk: As a CSP engages in network transformation and adopts their AI strategy, what are some of the changes they need to make to their organizational mindset and business processes?

Singh: Any type of transformation eventually comes down to people, process, and technology. To transform existing processes, you need clear definitions and processes that are simplified and standardized. That paves the way for automation with AI. Technology is the easier bit. The hard work of defining, simplifying, and standardizing is a people and process challenge.

Then there’s the mindset changes around data. Data is how you fuel AI, and today a lot of data lives in silos and of course, it is nowhere near clean. To unlock the potential of AI, you have to break down those silos to really get at the end-to-end value that AI is going to generate.

Leveraging AI requires a new approach to processes and skill sets. That’s where multi-vendor ecosystems play a big role in supporting CSPs through this transformation to help define processes and upskill people.

MeriTalk: If a CSP is at the very early stages of transformation, how should they be thinking about their data management and storage when it comes to AI?

Singh: First things first, you have to start where the data resides. Data has gravity, as well as privacy, security, and sovereignty issues.

This is why you want to bring AI to your data and not the other way around. You should be thinking about where you want to do your AI work, which helps drive whether you’re going to build your AI infrastructure, whether it’s in the public cloud, multi-cloud, or private cloud. The goal is to make it easier for your data scientists and your models to effectively tap into that data. This is where the work we’ve been doing on the Dell Data Lakehouse comes into play. We’ve all had data warehouses and data lakes where the data is just sitting in there, unused.

The Data Lakehouse is an AI-ready data platform that provides a unified data layer for ingesting and accessing data. It’s a highly secure, single point of access to all data, so your data scientists can put it to work.

Another question I get a lot is which use cases should you start with. At a high level, we advise CSPs to focus on the use cases that will have the biggest business impact and assess how good the data is. The use case that generates a high business impact and has good data is generally the one to start with.

MeriTalk: How does the Dell AI Factory differentiate you in the eyes of CSPs?

Singh: The Dell AI Factory is one of our major areas of strength in helping our CSP customers in their network transformation. Think of a traditional factory where you need raw materials coming in, you have infrastructure and processes within the factory, and finished goods come out of the factory. For CSPs working on AI, the raw material going into the AI factory is the data, the factory floor is the infrastructure, the open ecosystem, services, and the data engineering and data strategy. The output is the use cases and the business outcomes.

At the infrastructure layer, we bring a breadth of solutions, from training to inferencing. Through the open ecosystem, we help CSPs identify partners that can simplify AI deployment and operation. We’re also developing quick reference designs that help CSPs get started quickly.

Together with our ecosystem partners, we start with the data, select the right models, help to identify the right use cases, bring in the right infrastructure, and focus on the outcome so that the CSP can realize the operational and financial benefits of AI.

MeriTalk: How are some CSPs using AI in their networks today?

Singh: Beyond customer experience applications, another growing use case that’s operational today is around network engineering and troubleshooting. Think of the AI as a copilot for a network engineer that can look at large call flows, files, and logs and quickly identify where the error flows are coming in and troubleshoot the network.

Another network engineering use case is predicting equipment failure with high accuracy. If you’re a chief network officer and you get predictions that are 90 percent accurate eight to 12 hours in advance of a line card in a BNG going down, you’ve got unprecedented ability to get ahead of the curve with network workflow, workflow planning, and getting trucks out into the field. Ultimately, you are reducing network downtime which in turn improves the customer experience.

AI also allows for network engineering to create a digital twin of their network, where they can run scenarios, test, and predict to proactively mitigate issues.

With all these use cases and practical applications on the network side, you can draw a straight line to positive impacts on the bottom line, which validates the original business case for network cloud transformation and AI.

MeriTalk: What’s next for CSPs and AI?

Singh: Over the next few years, we’ll see AI proliferate across CSP networks. There are really two things to keep in perspective. First, realize that today’s AI is the worst it will ever be. The technology is moving and improving at incredible speed and AI’s opportunities and use cases are only going to grow. Second, while a number of CSPs have already gotten started, there are many who haven’t and it’s absolutely critical they ramp up.

Those who have gotten started with AI in their networks already have a competitive differentiator, and the gap between those who are realizing the benefits and those who haven’t gotten started will only widen.