How Artificial Intelligence Is Helping to Revolutionize the Telecom Industry

The development of artificial intelligence and the presence of technological solutions based on such technologies will accelerate digitalization in the telecommunications industry. New solutions will assist in the development of new digital communication channels such as chatbot or voicebot solutions, will also impact network optimization, and provide solutions that can proactively eliminate issues within networks or technical systems, contributing to user experience and changing and optimizing numerous sales processes. Domestic telecoms are aware of this, having not only recognized the potential that artificial intelligence brings but are already developing their own AI-based solutions.

Thus, A1 Hrvatska has been developing an internal team for the past six years that deals with the implementation of AI solutions in various domains of the company. They are developing AI solutions that are involved in various sales processes, IT system optimization, communication with users, and improving various processes, and that is not all.

– We also recognized the potential in NLP (natural language processing) solutions very early on and developed several AI systems that assist in daily communication with users. One such solution is an AI tool that participates in processing incoming emails in customer service and speeds up the communication process with users. To accelerate the development of various AI projects, we are intensively investing in the development of the internal team; however, we are always open to collaboration with other companies in this area and currently have several ongoing projects where we are jointly developing AI solutions.

The biggest challenges in implementing NLP solutions have been the lack of a corpus in the Croatian language from which AI solutions could learn outside the A1 domain. Our internal AI team, together with colleagues from customer service, has been continuously gathering inputs for training the model for months, and that was the most challenging part of the project, explained Andrea Pirša Ilić, head of cognitive computing at A1 Hrvatska.

Training Staff Internally

She also noted that there are currently not many experienced AI engineers in Croatia, as the domain of artificial intelligence has only been rapidly developing in recent years. Despite this, Pirša Ilić says, the interest in working at A1 in the field of artificial intelligence is indeed high, and there are currently no challenges with staffing.

– We have experienced experts who invest in young colleagues in this area, and we believe that this is the recipe for successful development of AI teams, she emphasized.

Research and development of artificial intelligence in other industries, including telecommunications, largely depends on funding from European Union resources and private capital from companies that see potential in developing this area. A1 will continue to invest in the development of AI solutions used in various domains of the company, especially since the company has large amounts of data, and the development of AI solutions creates additional value from the data.

Ivan Runje, director of the Product Management Sector at Hrvatski Telekom, told Lider that technologies such as artificial intelligence are an important part of the digital transformation of Hrvatski Telekom, which strives to get as close as possible to the concept of a ‘data-driven’ company in its business practice.

– This implies the use of concepts of data science potential, and thus AI models in everyday business for the realization of various projects that have a wide range of applications, from preventing technical errors, recognizing market trends, increasing customer satisfaction, optimizing processes, and so on, Runje emphasized, adding that the results of these solutions contribute to a better understanding of their users, as well as the market, cost optimization, and also serve as input parameters for making business decisions.

He explains that artificial intelligence will enable telecom companies to provide a better user experience and simultaneously, based on network automation, proactive management systems, virtual assistants, and robotic process automation, create added value for their users, which Runje explained with an example.

Namely, through mathematical models, user needs are simulated and recognized to respond to them in an adequate and relevant manner. For this purpose, a three-layer architecture is used, where the first layer contains predictive models that assess user satisfaction based on recognized patterns from historical periods, the second layer uses expert systems based on ‘fuzzy’ logic that recognize disruptive events that may jeopardize user experience, while the third layer is based on processing the Croatian language, and its task is to recognize hidden patterns that affect the level of user satisfaction.

Adapting to Croatian Cases and Speech

– Technology helps us recognize rare events with users, which was not the case before, as resolving user requests was done through the prism of frequent events. All of this leads to better quality of our actions and responses to users. Many of the solutions we use have been developed internally, such as the text mining solution that was developed to adapt to the Croatian language, cases, and everyday speech, as well as the specificities related to communication patterns in the text corpus. Expert systems based on ‘fuzzy’ logic have also been developed internally in the company, as well as solutions based on predictive models that contain HT’s ‘know-how’ related to the specificities of problem spaces that have been addressed. Additionally, at Hrvatski Telekom, we are also testing ChatGPT, Runje said.

When implementing any new processes and technologies, challenges can arise, and they have not been absent in this telecom, mainly related to the development of their own software library for processing natural Croatian language and adapting to the area and the specificities of telecommunications.

They have also encountered challenges in optimizing learning processes in a ‘big data’ environment for the purpose of developing recommendation systems, where traditional AI algorithms need to be adapted to the ‘big data’ environment, especially in conditions where combinatorics plays a major role, and it needs to be reduced/optimized to successfully carry out the machine learning process.

Runje particularly emphasizes that user satisfaction is at the center of everything they do, and with the help of technologies, HT aims to respond as well and as qualitatively as possible to their needs and desires and to assist in resolving their complaints. Accordingly, they are also looking at other areas of business and how to optimize operations and energy consumption. He also states that every problem that can be solved with intelligent data processing is a potential focus of interest in a ‘data-driven’ organization like HT.

There Are No ‘Cookbooks’ for Data Challenges

– In ‘data science,’ you will rarely find ‘cookbooks’ that will help you solve complex business problems, which is an additional motivation for further development of such solutions, especially if they ultimately solve a problem, save costs, or increase the level of user satisfaction. It is particularly challenging to mathematically describe and predict and manage something as intangible as user experience using data science methodology. Hrvatski Telekom has already found solutions in this domain and is looking for new ones that integrate a whole range of algorithms and its own solutions from the AI domain, striving to better understand user needs and accordingly adapt its business.