There are enormous potential benefits to using AI technology, including making operations more sustainable. While we work toward taking advantage of these opportunities, it’s crucial to understand both the challenges and the potential pitfalls of reliance on AI. On the way ahead, the importance of domain knowledge, customer focus and collaboration with partners will only increase.
Benefitting from AI in telecommunications will rely on several key areas: data expertise, including specialised machine learning (ML) skills, leadership abilities that embrace data-driven company culture, finding new ways of working that drive experimentation with data and AI tools, and not least having the required critical connectivity infrastructure, essential for growing new AI businesses.
Optimal infrastructure for AI
AI requires the transfer and processing of huge amounts of data, which means the infrastructure must be able to secure safe, sustainable and efficient data handling. Telenor has established a new business area consisting of infrastructure for digital communication, including communication towers, fibre, data centres and edge data centres. Through optimal design of the digital infrastructure, we are securing low latency and high digital data handling capacity.
Intelligent masts increase efficiency
One of the interesting uses of AI in Telenor is likely to be saving energy in the radio access network. Telenor's ambition is to become carbon neutral by 2030 in the Nordic region, so it is high on Telenor’s agenda to reduce network energy usage.
This work started in 2018 with a project we’ve called Green Radio. Telenor Research and Innovation worked together with a team in Telenor Denmark, and later with other business units, to make mobile masts 'intelligent' in an effort to optimise their energy use. Using data and advanced analytics techniques, the project reduced power consumption and CO2 emissions, and most importantly gave us valuable learnings on what we can do, together with our partners, to reduce emissions without sacrificing the quality of network services.
How was this achieved? The Green Radio team analysed data reflecting how often the mast was used and at what time of day. They were subsequently able to calculate at what time of day data usage was low enough to reduce the capacity of the masts, while keeping the coverage intact. Our initial tests in Telenor Denmark showed savings of 2-3% of total annual power consumption in the network.
With the Green Radio approach, we believe we can more than double the savings from using sleep-mode functions on Telenor´s 4G network. Similar tests in other Telenor business units show a savings potential of up to 8% of total annual power consumption in the network.
Our AI journey
Currently, our commercial operations use AI and machine learning in three areas, either through development of technology in-house or with partners:
Network AI: Automating network operations, reducing CO2 emissions and increasing resilience against cyber-attacks
Customer-facing AI: Increasing the value of our customer service, marketing and sales operations through data-driven decisions
Beyond connectivity: Developing new services on top of connectivity offerings and enabling growth in other industries
In Telenor, we began working on AI in 2012, when we created our first dedicated AI research team, and later invested in a Norwegian Open AI lab at the Norwegian University of Science and Technology, NTNU together with academic and industry partners.
Data analysis and forecasting
Going forward, sophisticated data analysis and forecasting methods can support network rollouts and investment planning. Anomaly detection and root-cause analysis help troubleshoot and resolve issues in the network. By gathering insights and building our knowledge about AI/ML use cases in simulated mobile networks, we are preparing for the eventual deployment of similar technologies in our live networks. Telenor is also exploring how more advanced ML methods can be applied to personalise customer interactions, automate sales and marketing campaigns and increase the security of our critical IT and network infrastructure.
Language models and speech recognition
Another area of longtime research is within language models. In Telenor we have worked on this within the related field of speech recognition. To enable good transcriptions of recordings, we need a model that puts the phonetics into context. One example: the difference between “during the reign” and “during the rain”. To distinguish between the two, context is key. Language models must be able to predict which is the correct word in many contexts. This technology is what gives technologies such as ChatGPT and Bard the ability to write fluently on almost any cue.
In 2020, Telenor Research delivered a proof of concept with a small sample of Telenor Norway’s customer service speech data, showing that it is possible to produce relatively good and meaningful transcriptions for challenging types of data. Our team continues to advise Telenor on automated speech recognition and text analytics solutions for automating and improving customer service.
Our AI learnings
Our learnings after so many years of research are many. Overall, we understand that AI is "just" a piece of software, but one which has become more powerful. Reaping the true benefits requires high domain knowledge, data readiness and ability to integrate and scale it in our operations. We are looking forward to working on responsible AI solutions in the coming years, in collaboration with our partners.