Artificial Intelligence and Big Data in the Telecoms Industry
The global business value of Artificial Intelligence (AI) is estimated to be in the trillions of dollars in the coming years. There are great applications of AI such as improved diagnosis of cancer, selfdriving cars, automatic translation, personalisation, and even art creation or music composition. What has enabled AI to make such leapfrog in the last decade that it has become so powerful? There are basically three reasons: i) the abundance of data; ii) democratised access to economic processing power; and iii) new Machine Learning techniques such as Deep Learning (many-layer neural networks). It is the combination of those factors that make AI applicable to almost any tasks there is, and hence the estimated global business value.
However, it is not easy for large enterprises to become data-driven and AI-powered; it is a long and difficult journey where progress is made through different phases. The journey usually starts with an exploration phase where the potential value of data and AI is tested through some initial pilots. If successful, the next phase is a transformation phase where the organization becomes serious about data: a global roadmap is established, key use cases are implemented in the main business areas, and data silos are broken. Once in the third phase—the data-driven phase—most of the organisation’s important decisions are informed by data using machine learning techniques. Data is democratised to the rest of the company, the company culture is becoming more data-friendly, and new, data-driven business models see the light. In the last phase, full advantage can be taken of AI by applying it at scale across the company. When data and AI have taken such an important role, also other aspects like trust, privacy, ethics, and social good are becoming key for sustainable growth.
Current Natural Language Processing and Machine Learning techniques are good enough to automate a significant amount of customer interactions
Many companies have learned the hard way that one of the main problems for scaling big data and AI is related to data access and fragmentation, leading to the annoying phenomenon that repeating the same project in a different business does not result in time gain—most of the time is dedicated to data integration. Some enterprises therefore have taken the decision to define a common data format across their organisation and subsidiaries allowing a rapid “lift & shift” approach of successful use cases. Telefonica has defined its so-called “Fourth Platform” that consolidates in a common data format with all data coming from its operational systems. All big data and AI applications are then executed on top of this common format. Other lessons that we have learned during our data and AI journey relate to decisions such as: to whom should the Chief Data Officer report to? What is the relation between IT and Data/AI? How to select use cases? How to measure economic impact? How to assess data maturity? How to ensure a responsible use of AI? At our blog “data-speaks”, we explain several of those questions in detail.
Let us dive a little deeper in the specific opportunities for AI in the telecoms industry. Firstly, there are many applications to optimise the core business. Think about network planning optimisation, point of sale optimisation, churn reduction, intelligent pricing, device recommendation, B2B sales optimisation, and so on. There are literally dozens of possible use cases, which together can generate hundreds of millions of value if applied at scale across the organisation.
Secondly, AI and data can be used to improve the interaction with the customers. Think about chatbots that not only answer generic questions, but have access to personal customer data to attend to each customer in a personalised way, and allow for multiple types of interaction such as text, speech, and point & click. Current Natural Language Processing and Machine Learning techniques are good enough to automate a significant amount of customer interactions. If you take into account that those interactions happen in many different places including websites, mobile apps, and social networks, call centres, and shops, then you can imagine the huge opportunity implied by implementing this technology at scale. This reverts in better, more consistent customer service, where AI takes care of the repetitive and boring interactions, while human agents can focus on the more complex and interesting interactions. But it also allows for significant cost savings through less outsourcing of customer relation services.
The third opportunity is based on the differential value that mobile network data embodies. Every interaction with a telecoms network, be it from a person or an IoT sensor, generates a digital trail. By anonymising and aggregating all those data points in a privacy-respecting manner, this data can be turned into highlevel insights about mobility patterns in cities or countries (how crowds move around) and about footfall (where crowds are at certain times). And those insights provide very relevant information for sectors such as transportation, tourism, retail, finance, and public administration. As an example, Telefónica’s Data & AI Unit—LUCA—has performed over 400 of such projects with clients in the past three years; really a new business. However, apart from business value, the same insights can generate value for societies: Data & AI for Social Good. In several projects with international organizations such as UNICEF, FAO, IABD, UN Global Pulse, and World Bank, we have used those insights for predicting spread of pandemics, assessing the impact of natural disasters, improving air quality in large cities, and supporting governments in understanding forced migrations.