Skip navigation
Like what you’re reading?

How to make better use of network insights with Generative AI

Explore the transformative impact of Generative AI within telecommunications and how to simplify network data communication and harness data-driven insights for superior business outcomes. Ready to revolutionize your data strategy with AI?

BCSS BOS SDU Solution Architect at Ericsson

System Manager (Data Science) at Ericsson

BCSS BOS SDU Lead Data Scientist at Ericsson

BCSD BOS Data Scientist at Ericsson.

BCSS BOS SDU Sr. Data Scientist at Ericsson

Strategic Product Manager

BCSS BOS Technology Evolution & Strategy, Technical Manager

Generative AI for better network insights

BCSS BOS SDU Solution Architect at Ericsson

System Manager (Data Science) at Ericsson

BCSS BOS SDU Lead Data Scientist at Ericsson

BCSD BOS Data Scientist at Ericsson.

BCSS BOS SDU Sr. Data Scientist at Ericsson

Strategic Product Manager

BCSS BOS Technology Evolution & Strategy, Technical Manager

BCSS BOS SDU Solution Architect at Ericsson

Contributor (+6)

System Manager (Data Science) at Ericsson

BCSS BOS SDU Lead Data Scientist at Ericsson

BCSD BOS Data Scientist at Ericsson.

BCSS BOS SDU Sr. Data Scientist at Ericsson

Strategic Product Manager

BCSS BOS Technology Evolution & Strategy, Technical Manager

Application of Generative AI in telecom data analytics

As 5G networks grow, solving complex queries about the status of the network in real time could push the network management system to the next level, rather than relying only on the usual periodic BI (business intelligence) reports. Communication service providers (CSPs) have a rich set of data, and network analytics helps them utilize this data to address operational challenges and discover new use cases. For example, they can use the data to understand how the performance of various device models differs with the varying quality of traffic or track the trend of incidents across the cells or regions). However, it can be difficult for CSPs to take advantage of all these data because significant barriers leave most of the data and its potential unused. Generative AI (Gen-AI) combined with analytics could help lower this barrier and democratize data, enabling a broader set of people to explore using natural language-based queries in a useful way.  

In the telecom domain, the integration of Gen-AI's capability to convert natural language to SQL and execute complex SQL queries to fetch data from databases presents numerous opportunities for innovative use cases. This advanced functionality allows for seamless interaction between users and data, enabling natural language queries to be translated into SQL, executed to retrieve the relevant data from databases, and then further transformed into natural language to present the query results. By leveraging this capability, telecom companies can empower users to effortlessly access and analyze data, enabling them to make data-driven decisions with ease. 

Figure 1: Real time network insight delivered by using LLM powered bot

Figure 1: Real time network insight delivered by using LLM powered bot

Furthermore, using Gen-AI's chain of thought, RAG (Retrieval-Augmented Generation), and pre-engineering the output can provide insightful interpretations of the query results, enhancing the overall understanding and value derived from the data. With the continuous evolution of language models like LLMs, the potential for creating a user interface that facilitates natural conversation for querying and analyzing telecom data has become increasingly feasible. This advancement opens the possibility of transforming the traditional search process into a more intuitive and conversational experience, offering a user-friendly approach to accessing and comprehending telecom data. As a result, Gen-AI has the potential to revolutionize data accessibility and analysis in the telecom industry, empowering users to extract actionable insights through natural language interactions with the data. 

CSPs around the globe are thinking about the potential commercial feasibility of Gen-AI technologies. One emerging area of application is making analytics insights more accessible 24/7 using natural language queries and responses. Transformer architecture-based LLMs, which are the backbone of Gen-AI, show promising performance in areas such as Text to SQL problems using in-context learning. As a prototype, we leverage one of Ericsson’s Business and Operations Support Systems solutions from the Data and Analytics portfolio. Ericsson Expert Analytics (EEA) has been used as a data source for the prototype. It stores real time and historical information related to network health, incidents, subscriber experience, and many active device related information. With the proposed solution, the user can query all relevant asks and can download them in a format of small contextual reports with graphs, and geographical maps. The usefulness of the insights will always depend on the quality of the data available in the system, therefore, the underlying analytics solution is crucial. 

Network operations can be made easier with LLM-powered assistant 

With regard to LLM-powered assistants, two key areas are considered a high priority in the CSP community: 

  1. Humanizing AI – Helping to reduce or remove concerns regarding AI as this black box that is about to control the world.
  2. Democratization of data – Making large and complex datasets more easily accessible to larger receiver groups inside the CSPs by leveraging the cross-domain learning 

Along with these, the availability of pre-trained LLMs realizes the power of AI in designing intelligent products and bringing operational efficiency. The power of prompt engineering has been used to interact with various open-sourced pre-trained LLMs and evaluate what works best for the purpose. The key learnings are the requirement of large computing resources, to be able to provide the most accurate answer to Natural Language Questions (NLQs), and having consistently good results in multi-round Q&A. 

Figure 2: Generating dynamic prompt and interacting with the LLMs

Figure 2: Generating dynamic prompt and interacting with the LLMs

Gen-AI’s potential applications are rich and varied  

In the telecom domain, the potential of Gen-AI to act as an enabler for new and innovative use cases is vast and it is indeed expanding. For instance, in network management, Gen-AI can be utilized to optimize network performance, predict and prevent outages, and automate troubleshooting processes. In customer service, Gen-AI can enhance personalized interactions through chatbots and virtual assistants, providing tailored recommendations and solutions to individual customers. Moreover, in the realm of cybersecurity, Gen-AI can be leveraged to detect and respond to threats in real time, bolstering the overall security posture of telecom networks. Overall, the integration of Gen-AI in the telecom industry holds the promise of driving efficiency, enhancing customer experiences, and unlocking new revenue streams through the creation of innovative use cases.

What’s next for Gen AI and network operation? 

By leveraging Gen-AI, the development of an intelligent assistant, which is capable of processing voice or textual commands during high-profile techno-business meetings of CXOs is not far from reality. This assistant can seamlessly retrieve relevant information from the stored data and present it in a manner tailored to the specific type of required output. For instance, if a CXO inquires about the cells in a particular region with the highest number of incidents, the Gen AI-based assistant can promptly fetch the data and, in addition to providing a concise summary of the results, generate a plot diagram to visually represent the output. With this innovative solution, no external manual assistance is required, providing the service provider with a streamlined, efficient, and insightful way to engage with their data. 

Figure 3: a) CE Insight

 

 Figure 3: LLM powered Ericsson Expert Analytics chat bot creating an intelligent assistant prototype

Figure 3: LLM powered Ericsson Expert Analytics chat bot creating an intelligent assistant prototype

The integration of Gen-AI in the process of augmenting the AI capability in the existing and new product lines significantly streamlines the process of accessing real-time data. This eliminates the need for time-consuming research and enables gaining valuable insights, with no need for further interface, SQL GUI, or report GUI. This empowers CXOs to make informed decisions swiftly, enhancing the efficiency and effectiveness of their decision-making processes. 

Security aspects of LLM-based framework

In developing a Generative AI-based chatbot with a locally stored, open-sourced Large Language Model, a multi-faceted approach to security is essential for maintaining the integrity of the AI system. The Open Sourced LLM model can utilize the prompt engineering functionality to get a reasonably good result in this Text to SQL conversion. Successful conversion of SQL query would fetch the analytics insights from the in-house databases that are protected with all the standard data security norms. Strong authentication and authorization mechanisms must be in place to control user access. This includes implementing input sanitization to remove all malicious content, employing tokenization and parsing for validating user inputs, and maintaining a whitelist of allowed prompts to restrict input formats. Overall, a proactive and comprehensive security strategy is vital to safeguard the AI chatbot from various cyber threats.

Help technical and non-technical users get a better view of the network's health 

The Gen-AI prototype would offer network service providers a 360-degree of the network. With the Gen-AI approach, it’s possible to query specific questions, download relevant data to do further analysis, and generate reports. The automated reports consisting of useful graphs, and data are generated which saves time, energy, and operational cost and a better network experience for the subscribers. 

Gen-AI has immense value in providing a paradigm to mine massive amounts of untapped data in the form of texts and images to extract meaningful insights. Gen-AI helps to reduce operational costs with marginal capex expenditure. It will support CSPs in making more informed decisions in B2B and B2C space, which will contribute towards more sustainability.

Access and use network data in a better way 

The Gen-AI prototype offers a new approach to enhance the accessibility of network analytics insights, fostering a higher adoption of analytics data among users. The integration of a natural language to SQL interface expands the applicability across a broader portfolio, marking the initial stride in introducing Generative AI as a new technology to adopt in telecom operations. The prototype, featuring an LLM-based chatbot interacting with Telecom Analytics data, highlights the innovative application of in-context learning and RAG for natural language to SQL translation. 

Addressing concerns related to data security and intellectual property theft, the self-hosted pre-trained LLM ensures a compact deployment. The integration of LLM through a text generation API adds a layer of flexibility to the solution, catering to diverse needs. 

In essence, the strategic advancements signify a pivotal moment in Ericsson’s transformation journey and highlight the potential of new generative AI technologies in an increasingly competitive market for AI-integrated offerings.  This new application of AI will help to channel information by using cutting-edge technology and generative AI models will become more fluid, more useful, and more commonplace. 

Read more:

OSS/BSS evolution for successful 5G monetization - Ericsson

Telecom analytics: transform data into action - Ericsson

Data and Analytics - Ericsson

Expert Analytics - Ericsson

Intelligent IT Operations Insights - Ericsson

The Ericsson Blog

Like what you’re reading? Please sign up for email updates on your favorite topics.

Subscribe now

At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.