About AI Governance | Best practices and organizational model from Telefonica’s Responsible AI

12 min readJul 10, 2023

DISCLAIMER: This article contains a non-official translation from OdiseIA “GuIA” report with PwC (originally in Spanish). Sources:

Telefónica, with an almost century-long history, is one of the emblematic multinationals that have accompanied the Spanish and international market over the years. Since its origins in 1920, telecommunications services and the infrastructures on which they are based have evolved to become the cornerstones of innovation, sustained by data. The company has been proactive in adapting to the sequence of technologies

disruptive initiatives, from the early adoption of big data, to well-known data and artificial intelligence (AI) initiatives such as 4th Platform (“cuarta plataforma”), and the company’s pioneering role in terms of data governance and artificial intelligence.

This case study aims to contextualize Telefónica’s role as an innovation company and pioneer in artificial intelligence, with a special focus on its own approach to responsible AI and its contribution to the state of the art in this area, which is a pending task for most companies, both at the national and international industry level, as well as in comparison with other relevant players in the telecommunications industry.

Context: Origins and responsible AI at Telefónica

Telefónica’s history as a technology company, which goes beyond pure telecommunications activities, cannot be understood without a data strategy, such as those generated by the networks, those related to service users, and the knowledge generated, in the form of insights. All of these have been the source and driver of the rest of the brand’s activities over the last few years.

The relationship between Telefónica and data (traditionally known as big data) is not new. The period between 2011 and 2015 marked the initial step between the company that until then was considered a relevant player in the telco industry and the total transformation to give rise to a company of innovation in data and artificial intelligence. The internal experience gained with big data initiatives led in 2016 to the creation of a business unit for customer projects (LUCA), now integrated into Telefónica Tech, and the creation of a new executive role: the Chief Data Officer with Chema Alonso at the head, which at the time was a pioneering decision by the company’s current Chairman, Jose María Álvarez-Pallete.

This new stage gave rise to today’s Telefónica, the same one that attracts attention at the Mobile World Congress (MWC), year after year, with cutting-edge presentations, and with innovation as its DNA. This innovation allows it to take advantage of all the opportunities for progress and business offered by the current digital ecosystem, characterized by its dynamism and speed, and with which it has forever changed the way of understanding data and the relationship between the telco and its customers, being the 4a Platform its maximum exponent and all the digital services that are created on it. using their capabilities.

The 4a Plataforma is the technological platform at the heart of Telefónica’s digitization that provides an end-to-end view of the customer. To achieve this, it relies on different capabilities of identity management, “apification” (APIs) and data abstraction that it implements under the maxim of privacy-by-design. With this ecosystem of capabilities it offers, applications, products and digital services can be developed faster, with less friction and also enables the creation of products that incorporate artificial intelligence to enrich the customer experience in Telefónica. Some of these products are: Aura, Telefónica’s virtual assistant with artificial intelligence; and the My Movistar application. Both offer a new form of relationship with customers to manage their digital experience with the company. Solutions to which we must add others such as Living Apps, applications integrated into Movistar Plus+, which offer a new consumption experience from the television or market solutions with third parties through Telefónica Tech, leader in digital transformation that integrates, among others, the Big Data and IoT business of the Telefónica Group.

Telefonica’s vision: artificial intelligence principles

Telefónica focuses all its actions on the Sustainable Development Goals (SDGs), as indicated in its Principles for Responsible Business and Human Rights Policy. This allows it to ensure the sustainability of the business as well as to contribute, in a significant way, to a decisive project for the future of people, promoting social and economic progress through digitization while generating trust to ensure a digital transition focused on people.

The main premise is that technology should contribute to creating a more inclusive society and offer better opportunities for all, leaving no one behind, and artificial intelligence can contribute to these goals. In order to guide the company and its employees in their application of artificial intelligence and big data in all Lines of Business, Telefónica published in 2018 its “AI Principles”:

The reason for the creation of these principles was precisely to point out the importance of ethics for artificial intelligence, which, according to Telefónica, cannot be complete if it is not accompanied by such a vision. Each of the five principles has a specific rationale:

Sources:

Octubre 2018 — Telefónica AI Principles https://www.telefonica.com/wp-content/uploads/sites/7/2021/11/principios-ai-eng-2018.pdf
Infografía sobre los Principios de IA Responsable en Telefónica https://www.telefonica.com/es/wp-content/uploads/sites/4/2021/03/infografia-ia-esp.pdf

The presentation of the strategic alliance between Telefónica and Microsoft to boost artificial intelligence and create new services, with the top managers of both companies present at MWC’19, illustrated a further step in Telefónica’s AI maturity stage, characterized as Cognitive Intelligence or cognitive intelligence by the company. Since then, the adoption of artificial intelligence has continued as a natural trend, and there have been very significant contributions in the field of ethics and responsible AI.

In the last two years, we have seen a general rise of responsible AI concepts, which started from a theoretical reality that has been evolving into more grounded practices. Telefónica has been no exception, and has worked on defining new governance practices and new related roles.

Differentiating factor: Internal IA governance

Artificial intelligence governance and Telefónica’s own approach are reason enough for this success story. Over the past three years, the company has developed a responsible AI system with three tiers, where each person has the ability to analyze potential risks.

The main points of this Responsible artificial intelligence governance system are:

AI principles as a starting point for any implementation involving artificial intelligence and/or big data.
The introduction of the governance model as a key part of the formal product and service development processes.
The creation of a self-assessment questionnaire to be completed by product managers and development teams during the design phase of products and services that use artificial intelligence and subsequently in successive phases throughout the product life cycle.
The definition of an escalation process between levels to ensure detailed analysis.
The figure of RAI Champions, as part of the Ethical AI governance process, who are close to both business units and developer teams.

Role of the Responsible AI (RAI) Champion

The figure of the responsible AI Champion is part of the new granularity of roles related to artificial intelligence initiatives, beyond the typical cases of data scientists and engineers. The RAI Champion is more a role than a specific position within an organization, usually occupied by professionals with very diverse profiles, from data scientists, privacy officers, to corporate social responsibility, with more or less technical knowledge, but who bring a different point of view when analyzing the ethical impact of an artificial intelligence project.

This role is also used by some companies that are well advanced in the implementation of responsible AI, but Telefónica has adapted this profile to internal processes and needs. The main premise is that a RAI Champion is needed whenever the business unit uses or plans to use artificial intelligence or big data systems in its solutions. If the level of adoption is intensive, there may be more than one Champion.

The functions of an RAI Champion include:

To inform about the importance of applying the principles of artificial intelligence.
Train their areas of influence on how to apply these artificial intelligence principles and governance model.
Guide and escalate, and be the point of reference for doubts and ethical issues, as well as escalate to the expert group when necessary.
Coordinate with different areas such as Privacy, Security, Chief Data Office, Corporate Social Responsibility, Legal, etc.
Connecting and fostering the building of a community of experts in ethical artificial intelligence AI.
Manage change, so that ethical considerations become part of the teams’ business as usual.

Three-level scaling system

As mentioned above, scaling is part of RAI Champion’s scope of activities. This scaling does not correspond to the traditional company hierarchy, but rather Telefónica has created a structure dedicated to ethical and responsible AI. There are not many companies that have reached this level of internal maturity in the world, and in the case of Telefónica it includes not only the RAI Champions of each area (defined as level 1 of the model), but a core RAI team (level 2), which coordinates all the RAI Champions and other actors such as CDO (Chief Data Office), CSR (Corporate Social Responsibility) or DPO (Data Privacy Office), and the Responsible Business Office (level 3). Specifically, the relationship between levels and scaling factors are:

Level 1 to 2: If there are doubts about the questions or results of the questionnaire (details in section 4), and these doubts cannot be resolved by the RAI Champion or its community of experts, then the case is escalated to the level of the core RAI team for consultation.
Level 2 to 3: In the extraordinary case that the core RAI team and its extended network of multidisciplinary experts are unable to find a solution, the final escalation level is reached, where the Responsible Business Office can intervene and solve the case.

Such a system promotes individual responsibility, and an ability to scale on demand, rather than a traditional top-down approval system.

Implementation: From ethical principles to action

Key point 1. Employee training as a knowledge and awareness tool

With artificial intelligence and ethics in artificial intelligence being relatively new fields of knowledge for many people, Telefónica has given the necessary weight to internal training to democratize access to such knowledge, and for each employee to understand the absolute importance of ethics in artificial intelligence.

Today, this training program has trained more than 3,500 company employees, and the number continues to grow.

Aspects covered in this training include the basics of artificial intelligence and big data, Telefónica’s Responsible AI principles and their implementation depending on the type of project, the governance system and day-to-day roles, the tools available to implement ethical and responsible AI, and of course practical examples to ground the concepts learned.

This program, on a par with any ethical and responsible AI quality training initiative, enables the development of knowledge and a common understanding of the principles, jargon, and realistic potential of artificial intelligence and big data.

Key point 2. The evaluation questionnaire as a starting point

After training, employees have the information and knowledge necessary to apply the principles in their projects. In order to facilitate the analysis, Telefónica has created a dynamic questionnaire to be integrated into the design of products and services using artificial intelligence.

This questionnaire is available through online access and contains a series of relevant questions on each of Telefónica’s IA principles. These questions provide a reflection based on the answers entered and recommendations to ensure ethics and accountability between the design and implementation phases. This questionnaire is completed by those responsible for the product or service to be analyzed.

Among others, the issues focus on:

Data sets and their variables, with special focus on those that may be sensitive (e.g., a person’s gender or nationality, or the presence of minority groups).
Correlation between variables, as part of the data preparation exercise to reduce unnecessary data that may introduce biases in the models.
The explainability of the systems and the possible impact on people’s lives.
Data privacy issues and regulatory compliance.
The relationship with third-party systems, and their alignment with Telefónica’s IA principles.

Finally, and as we have seen in section 3, the corresponding RAI Champion resolves doubts, proposes measures, and even escalates the case if necessary.

Key point 3. Inclusion of governance as part of the formal product development process.

The Responsible AI governance system aims to favor a Responsible AI first approach in which all types of stakeholders (from the most technical to the business units and executive levels) are part of the analysis and reflection to ensure an ethical and responsible implementation of AI in products and services. This system is a day-to-day reality at Telefónica, and aligns the entire company around previously defined processes and principles. Telefónica uses existing product and service development processes to integrate this evaluation of Ethical AI principles as another phase of development whenever the product or service incorporates artificial intelligence or big data.

This system, shared and adopted internally, together with the rest of the tools and training actions previously explained, allows to have a standardized and validated approach that guarantees the effective implementation of the system on a daily basis and that takes into consideration ethics in all phases of product development, starting from design. This maturity of thought and organization has allowed Telefónica and its employees to develop a very valuable experience, with interesting learnings that will be presented in section 5.

Lessons learned and recommendations

If we consider Telefónica’s journey to become a data and artificial intelligence company since 2011 (section 1), as well as the level of innovation and solutions presented over the years, it is fair to think that Telefónica has developed a knowledge and opinion formed around best practices to facilitate an ethical and responsible adoption of artificial intelligence. Few national examples can better illustrate this type of journey and its lessons learned.

Telefonica’s key learnings

The Chief Data Office and Chief Responsabìlity Office share their final thoughts:

AI principles are an excellent starting point to unify the approach to responsible and ethical artificial intelligence innovation in any company. Any company starting out in the world of data and artificial intelligence should adopt a set of principles. It is recommended to create your own set of principles, based on the company’s objectives, business context, company values and vision, and the specific use or application of artificial intelligence.
Defining and adopting a responsible AI governance model allows to ground the concepts, and facilitates the implementation in the day to day of the company. Introducing such governance in the official product development processes is key, and a successful methodology proven over the last few years by Telefónica.
Employee training is key to ensure understanding of the concepts and the model. Hybrid trainings that cover technical and business aspects should be favored, and that are accessible to any employee of the company, beyond their role, as potential adopters of artificial intelligence and big data in their products and services.
Creating the figure of RAI Champions and their knowledge communities allows to have a reference point for the correct adoption of a responsible and ethical artificial intelligence. These Champions are ambassadors. They often adopt this role as part of their responsibilities, without having to dedicate themselves exclusively to it. The desire to continue learning and helping development and business teams is the key motivation for this type of professional.

Benefits of ethical and responsible AI

Telefónica is aware of the importance of its stakeholders (customers, suppliers, employees and society as a whole), as well as its shareholders and investors. These are increasingly aware of the importance of developing ethical and responsible technology, as well as the growing requirements and demands to be taken into account.

For all the above reasons, Telefónica was one of the first companies in the world to define a framework of ethical principles for artificial intelligence, which, in addition to doing things right, would allow it to anticipate possible compliance risks and threats that could arise, thus reinforcing its risk control and prevention systems.

Companies such as Telefónica that have a solid system for controlling their IA risks will be particularly attractive and interesting for shareholders and investors, who for years have been stressing the importance of ESG (Environmental, Social and Governance) aspects in company strategy.