Machine Learning in Telecommunications: Prepare for AI Revolution

Their business is doing so well that, unlike in many other sectors, Artificial Intelligence (AI) and Machine Learning (ML) is more of a solution to their problems than a challenge. Telecoms are not far behind such big tech companies as Google, Amazon, or Microsoft and could use the 5G transition as an opportunity to cut the distance.

Telcoms have traditionally navigated quite well through tech change. Globally, they managed to transform from landline to mobile carriers, then move from voice calls to messaging and data-centric networks. In most of the developed markets, telecoms are creating ecosystems for the data-driven economy.

Telecoms are bound to grow even bigger

Large telecoms have predominantly succeeded in retaining their leading market position and now face challenges related to the growth and expansion to new business areas. Telecoms are not only holding firm their role in the tech business but also embarking on new ventures.

Telecoms are in the middle of an important transformation from the fourth to the fifth generation of cellular mobile communications (5G). The 5G technology will provide higher data transmission speeds, ultra-low latency, enlarged system capacity, and device connectivity suitable for a world controlled by the Internet of Things (IoT) and autonomous cars, making smart cities possible.

During the ongoing 5G transformation, the global number of mobile phone users is expected to surpass five billion by the end of 2019. Seven billion people are already creating more than 8 billion mobile connections. The number is expected to explode with the introduction of 5G and IoT to 25 billion networked devices expected by 2020, according to GSMA Intelligence.

Telecoms are responsible for delivering the infrastructure that becomes more indispensable and more valuable for the society every day.

Machine Learning is a must

Practically every telecom is investing heavily in AI. There's a consensus about this, as 93% telecom representatives see ML as a game-changing technology and 76% are planning to incorporate it into the business within three years, according to a survey by Digitalist Magazine at Mobile World Congress. The reason is obvious: telecoms need it, and they have the resources – money, people, and the data – to do it.

The need is urgent. Telecoms need Machine Learning to be able to process and analyze the data in many areas: customer experience, network automation, business process automation, new digital services, and infrastructure maintenance.

Here are some examples of what telecoms do.

Customer service chatbots

For telecoms, which serve millions of people every day, customer service automation creates substantial savings. Each day telecoms receive more tickets from their customers, and the structure of each case is getting more complicated.

Chatbots are a machine learning trend that offers a valuable solution to this problem. As human consultants cannot process all the data, telecoms need chatbots to make customer service faster and more scalable and improve client satisfaction.

Additionally, customer support chatbots in telecoms can be trained quickly and effectively. The companies have a long history of data retention due to strict regulatory obligations. Machine learning engineers can cross the transcripts or recordings of customer service with the actual logs that show what was wrong with the network.

ML algorithms can automate inquiries, route customers to the optimal agent, as well as direct prospects directly to salespeople. Chatbots are already able to ask the customer service-type questions and lead them through a troubleshooting process.

The use of chatbots can help increase the competence of the agents as they can spend more time on training, investigating issues, and start solving problems they are experts at.

Voice services and churn rate reduction

Voice services are telecoms’ natural area of expertise. Some companies are partnering with the leaders in speech and voice services, joining, for example, Alexa ecosystem. Others develop their own solutions or acquire smaller startups. South Korean companies are leading the pack. Recently, SK Telecom has introduced its artificial intelligence-based voice assistant service for the home, which was an answer to the move by its local competitor – KT applied its artificial intelligence assistant to a hotel in South Korea with English language support.

Machine Learning is also useful in reducing churn rates, which can average annually from 10 to as much as 67 percent. Telecoms can train algorithms to predict when a client is likely to turn to other company, and what offer could prevent them from doing it.

Predictive maintenance

Mobile towers are the perfect object for ML predictive maintenance solutions. They are difficult to access and require time-consuming on-site inspections of complicated modules such as power generators or air conditioners. Moreover, towers are vulnerable to intrusions, as they contain a lot of valuable equipment.

There are various possible applications of ML in the maintenance of mobile towers, such as empowered surveillance, where video and image analysis can help detect anomalies. The telecommunications infrastructure is already equipped with various sensors. The data those sensors collect can be used for training ML models, which will predict possible failures – this would reduce downtime and repair costs, and also improve the coverage.

Nokia uses ML algorithms are to adjust the best configuration for 5G antennas and indoor positioning of objects or configuring uplink and downlink channels. An active network configuration information can be cross-checked with asset management systems to maximize network utilization and improve coverage.

Infrastructure for IoT and ecosystems

European telecoms are engaging in 5G projects with Chinese telecommunications equipment giant Huawei, while the US, Australia and, recently, India are ready to sacrifice the pace of transformation for the independence from Chinese technology.

Huawei is becoming the leader of 5G solutions, which also include machine learning: smart buildings that use analytics data collected from video recordings, autonomous cars and smart cities, where engineers use big data to train ML models to control traffic better.

As providers of the gigantic infrastructure for the new world of interconnected things, telecoms will get their hands on vast datasets. What's more, they will tap into the very source of information – they will be in control of the world's nervous system.

Development of their own AI platforms

These are the reasons why many telecommunication companies are considering following the tech giants – like Google, Microsoft, Amazon or Facebook – and launching their own AI platforms. Vodafone, NTT, Telefonica, Orange, SK Telecom, Telenor, and AT&T are among the ones that have already declared it.

The Norwegian company Telenor has become a local leader of AI solutions. The company gave the funding and anonymized data to establish an AI research center in Trondheim and is a part of a Norwegian "powerhouse for AI" that uses ML to increase the security of gas and oil industry. In Spain, Telefónica launched its Aura platform, which uses AI assistants to interact with customers.

With the development of AI and ML, data is power. While big tech companies are leading the AI race, it seems that the telecommunications industry is feeling quite comfortable as a runner-up, and is ready to make the leap using the 5G transformation as a trampoline.