Is GenAI a panacea for the operational complexities of telco networks?

A look at why the technology is expected to be such a gamechanger for the telecoms industry in the next three years

Hakan, a Network Operations Centre (NOC) engineer for a large mobile network, reports for work at 9am. He begins to chat with his network.

Hakan: What will be the busiest cells in my mobile network at 4pm today?

Network: Here are the 3 busiest cell sites, by subscriber numbers: London RS, Manchester CM, and Leeds UB.

Hakan: Do you see problems impacting any cell sites in Leeds?

Network: Yes, there are developing performance impacts on 14 cells in Leeds. Do you want to know what these problems are?

Hakan: Yes, what are the worst 5 alarms for network vendor X in the 14 cells?

Network: Here are the top 5 alarm types and the number of times they have occurred in the last 24 hours…

This chat between NOC engineer Hakan and his mobile network happens within seconds. Within a few minutes, Hakan identifies impending problems, and identifies weak spots in the network. So, rather than scanning auto-generated (push) reports in a NOC, Hakan can pull responses to his niche questions, and prioritize his actions.

He does not use query language or write complex commands. He chats in plain English and receives responses from the network in plain English. This is the simplicity of generative AI in a telco network environment; it recognizes and communicates using natural language.

Let’s take another scenario.

The Head of Products at a leading Communications Service Provider (CSP) wants to know how a newly launched service is doing on the second day of the launch.

Deborah: We launched a new mobile data service on 5G yesterday. It is called Holiday Entertainment package. How many subscribers have registered for it?

Network: There were 12,000 registrations on the first day

Deborah: Have any customers used video streaming? How many accessed Spotify and Instagram Live through this service?

Network: I have all this information, and I can tell you how the new service performed for Spotify and Instagram. Would you like to know this as text or graphs?

This is what Generative AI (GenAI) is capable of – supporting network engineers and offering business insights.

In a recent Gartner report, Artificial Intelligence, GenAI was named as top game-changing technology in the next three years. Gartner reports that 79% of survey respondents (all of whom were from CSPs) voted for GenAI as the technology most likely to be implemented by 2026.

An AWS study (carried out by Altman Solon) shows that mobile network operators (i.e., LTE/5G) find a particular use case a ‘clear win’: guided employee assistance for installation, troubleshooting, and maintenance. This use case allows to retrieve and respond to network service engineers with information to aid in installation, troubleshooting, and maintenance of network devices and infrastructure.

The appeal of GenAI

Now, what makes GenAI so interesting for the CSP technical teams?

First, it uses different forms of data to make its outcomes intelligent, analytical and effective for network operators. If the CSP allows GenAI to crunch multi-source data, (Network performance, Quality of Service, billing, sentiment analysis from social media, point of sale and CRM data) it can generate responses to complex data queries which would otherwise be highly difficult to correlate.

Second, it offers deep, generative models within the artificial intelligence (AI) umbrella, in comparison to Machine Learning, which focuses on predictions and enabling of actions. GenAI creates new content after crunching the multi-source data, thus establishing itself as a simple tool for human-to-network communication.

In the mobile industry, the rapidly developing high interest in GenAI adoption is triggered by the network complexities posed by 5G and its applications. 48% respondents in the AWS survey said that they would adopt GenAI use cases within a year and a half from now. In MYCOM OSI’s conversations with CSPs, we hear that network optimization is a very strong use case for GenAI.

NOC applications

An ideal GenAI NOC application would supply Hakan with all the information that he needs, and, in the near future, as detailed multimodal reports. It will intelligently report on the network health, network utilization, identify the root causes of incidents and provide high-quality diagnostic information in text, charts, and graphs, which enable Hakan to resolve problems faster.

GenAI gives him real-time advice on the next best action, and predictive insights on what is round the corner. It could even generate code for him, which could be transferred via APIs to systems that can use it for new applications.

Other than the network optimization aspect, there is a strong commercial angle to GenAI. Instead of producing daily reports for the C-level (and other business users in the company) GenAI can free Hakan of overheads and let him focus on critical operational tasks, where his skills will be best used.

GenAI can instead have daily conversations with the product strategists and business planners to generate new business insights for them, inform them of revenue impacting issues and their extent, generate customer usage/footfall/behaviour, billing insights and recommend new offerings. All using simple text-based natural language chats.

Large language models (LLMs) are at the heart of GenAI. They are the underlying technologies on which to build and scale GenAI applications, suitable for use by both technical and business users. While GenAI use cases are extremely promising and there is high interest in them, the LLMs need to be trained for the desired outputs for specific domains.

CSPs are more interested in defining niche LLMs for their use cases or their domains, rather than use generic LLMs, and rightly so. These fine-tuned models (or Foundation Models) and niche expert models will deliver more reliable, accurately summarised and sophisticated responses to natural language queries.

MYCOM OSI is working on developing use cases using GenAI for NOC users and for business users of CSPs. Collaborating with their teams on multi-source data helps us understand the vast possibilities that GenAI offers. And how it can transform the analysis, distribution and consumption of network and non-network data to simplify operations and create new business insights.

This blog was first posted on Mobile Europe on 22nd Feb 2024: https://www.mobileeurope.co.uk/is-genai-a-panacea-for-the-operational-complexities-of-telco-networks/