Javier Blanco Romero

At Telefónica, we have known for some time that talking about Data Governance means talking about much more than policies, committees, catalogues or quality. What we are really talking about is an organisation's ability to keep the wheel turning when it comes to deciding who can do what with data, under which rules, with what responsibilities and for what purpose. It is undoubtedly more psychological and human than technical. A Data Governance project has become an intellectual ambition. This idea becomes even more important at a time when data is no longer merely an asset for analytics, but has become the essential counterpart to Artificial Intelligence. AI is changing the way organisations produce, consume, classify, connect and use information. In doing so, it is turning Data Governance into an increasingly strategic capability. It is no coincidence that DAMA places Data Governance at the centre of its well-known “DAMA Wheel”, connecting it with all the other knowledge areas of data management. We discussed this a long time ago, including here. Governance acts as the hub that connects responsibilities, policies, decisions and control mechanisms across architecture, quality, metadata, security, integration, modelling and the other disciplines. The metaphor is particularly powerful: Data Governance is the hub connecting every discipline. We could even say that the other disciplines are becoming AI-tomated, while the one at the centre is becoming HU-manised. AI has intelligence, but it needs knowledge. DAMA-DMBOK 3.0: a framework preparing for a new era With all this in mind, DAMA International (Data Management Association) is working on the evolution towards DAMA-DMBOK 3.0, a project that began in 2025. DAMA-DMBOK 2.0 dates back to 2017, and the landscape is now completely different. The new version aims to modernise the framework for an ecosystem in which managing databases and traditional processes is no longer enough. DAMA explicitly states that the new edition will incorporate emerging areas such as AI, Cloud and modern data platforms, while retaining the model's fundamental principles. And there is one particularly significant aspect: DAMA is approaching this evolution as a collaborative, community-driven process, involving specialists, reviews and mechanisms for public consultation. In other words, even the profession's own reference framework recognises that the world of data is changing too quickly to be defined solely from a static perspective. DAMA Spain currently points to 2027 as the target for official publication, once the various drafting, consultation and review phases have been completed. That gives DAMA-DMBOK 2.0 a ten-year lifespan. Time will tell whether version 3.0 proves equally long-lived or needs to be revised sooner because of this exponential acceleration. The question, therefore, is not whether Data Governance will change with DMBOK 3.0. The question is: To what extent will Artificial Intelligence transform the very way we govern data? AI is accelerating Data Governance AI adoption is advancing at a pace that is difficult to ignore, or even absorb. According to McKinsey, in 2025, 88% of surveyed organisations reported using AI in at least one business function, up from 78% the previous year. Yet only 7% said they had fully scaled AI across the enterprise. That already feels like a long time ago: as we approach the end of 2026, agents are already moving and producing data, and they need to leave a governance trail. We might describe them as replicants or synthetic humans producing information, and the pace of acceleration is relentless. The trend is clear: more organisations are using AI, but many are still building the capabilities needed to govern and scale it. This is also changing the economics behind decisions about licensed and open-source software: the financial balance needs to be recalibrated. Open-source software is becoming increasingly attractive because it can be adapted more readily to specific needs through development, shifting more of the value creation towards people and AI. At the same time, Data Governance is also gaining ground. A study by Precisely showed that the proportion of organisations with a Data Governance programme rose from 60% in 2023 to 71% in 2024. The research also found that Data Governance was the leading data-related challenge for AI initiatives, cited by 62% of respondents. The message is clear: AI is making the need for better Data Governance impossible to ignore. It is no longer simply about governing data; it is about governing data better. And governing data means creating a semantic representation of the organisation's world. We are not talking only about governing the data that feeds models. We also need to govern the data they generate, the associated metadata, access rules, semantic models, audit trails, permitted uses, quality, explainability and the decisions derived from AI systems. From governing data to governing data for AI This is where a fundamental distinction emerges. A company may have a catalogue, a glossary and quality policies and still not be ready for AI. Gartner reported in 2025 that 63% of organisations either did not have, or did not know whether they had, the right data management practices for AI. The conclusion is particularly significant: having Data Governance does not automatically mean having AI-ready data. And that is before we even consider governance of the AI agents themselves as they move, create or alter data. In fact, this evolution is leading towards a new generation of governance in which conceptual boundaries increasingly overlap: Governing data → governing data for AI → governing AI The boundaries between these three realities are becoming increasingly blurred. The 2025 IAPP report illustrates this convergence: 77% of surveyed organisations were already working on AI Governance, rising to almost 90% among organisations already using AI. In addition, 10% identified the Data Governance function as primarily responsible for AI governance. The direction of travel seems clear: Data Governance is no longer simply a discipline focused on controlling the “data” asset; it is becoming critical trust infrastructure for AI. Data Governance as a human anchor in the age of AI There is no doubt that AI can accelerate many activities that have traditionally required significant manual effort, but there is still a long way to go for humans too. We need to confront and recalibrate the role of people in this new paradigm, what is known as Human-in-the-loop. We must not succumb, in our use of AI, to “shiny object syndrome” or to the siren song from The Odyssey. AI can already do a great deal, and more every day, but we need to be intelligent about what to do and what to stop doing among the things we still do today. Perhaps we should reconsider the need to document certain things when the result is ever larger, heavier deliverables, packed with more AI, that nobody will read. Perhaps our effort should instead go into being clear about where the human contribution really matters. AI can automate many Data Governance tasks, but it cannot, by itself, assume responsibility for governing data. The next frontier Governance is not simply about detecting patterns or executing rules. Governance means making decisions about who can do what, and under which conditions: Who produces the data? What rights do they have? What obligations does it create? What happens when data is produced by AI? Who is responsible for a piece of data? Which definition should take precedence? Who can access certain information? Who can use it? What level of quality is acceptable? Can a particular dataset be used to train a model? What level of risk are we prepared to accept? What happens when one business rule conflicts with another? Who decides when there is a dispute over data? ■ These are not merely technical questions. They are organisational, legal, strategic and, in many cases, ethical questions. Our view of the role of people, Data Governance and AI can be summed up almost poetically by borrowing Seneca's words: Docendo discimus. By teaching, we learn. Conclusion AI is still missing the knowledge held within organisations. Governing data means creating that semantic network, that ontological layer that gives AI the knowledge architecture it needs, and in doing so, all we are really doing is teaching. The organisations that are best able to capture that knowledge and provide AI with that “brain” will create the greatest value. And at that point, we will learn even more. We will have created an AI that knows everything, or almost everything, and that will give AI the knowledge it needs to understand our world much better. And there is no doubt that this is work done by humans, for humans. Telefónica Tech Data governance: key to becoming a data-driven organization April 21, 2026

Original content from Lourdes Tejedor, Community Manager and Editor at Telefonica. One city: New York; one problem: crime. How can we contain and control it? When and where is it the worst? Are policemen well located to dissuade robbers? Thanks to Big Data, the local government now knows that Brooklyn is the district with the highest crime, that the crimes usually take place between 15:00 and 19:00 with the lowest rate in January (maybe due to the Christmas holidays); that the majority are thefts on a small scale, with the greatest frequency on Fridays when there is the largest concentration of people near banks, where four times more crimes take place. Figure 1 : New York is the 10th most dangerous city in the world In order to access all of this information, the local government of New York (which was named as the tenth most dangerous city in the world in 2010 according to the World Security Institute) and the Spanish firm Synergic Partners, the consultancy department under LUCA from the Telefonica group, used data from the local police (complaints), census data (purchasing power, levels of unemployment etc.) articles about incidents taken from the New York Times, stock prices, events scheduled in the city, as well as meteorological data. In total, they studied 78 variables and 4,2 Tb of information. "We have achieved a precision average of 72%, in reference to the time when a crime will take place, as well as the identification of false threats" explained Santiago González, Director of Innovation at Synergic Partners. This level of precision can reach up to 83% for murders (the type of crime with the greatest amount of available information) and as low as 67% when it comes to predicting a kidnapping. "There are fewer cases of kidnapping and they are usually handled by the FBI, which means that we have access to less data," said González. These divs show an important potential, with real information. Furthermore, the idea for the coming months is to delve into the data, focussing on areas to cover in neighborhoods and residential blocks, with the analysis of more than 200 variables including times of real murders in order to better detect serial murders. Here is another example of how Big Data can make the world a better place by helping to predict incidents and by making the decision-making process easier when faced with, for example, natural disasters, expansion of epidemics, contamination, agricultural productivity, medical diagnoses...