You cannot share, reuse or exploit what is not known and governed. At a time when all public bodies want to participate in the data economy, deploy artificial intelligence use cases or join European Data Spaces, the first step remains the most difficult to climb. And this first step has a name: data governance.
The temptation to connect everything, and why it's a mistake
When an administration decides to open its data or feed a corporate catalog, the temptation is great: to directly connect transactional systems or data lakes to a public portal and let the citizen, company or agent access it.eressat takes care of it. It looks fast, it looks efficient, it looks modern.
But exposing raw data, without prior inventory, sanitation, and semantic enrichment work, is the perfect recipe for three problems we know well in the public sector: inconsistencies that erode trust, legal risks stemming from improper exposure of personal data, and, finally, repositories that no one consumes because no one understands what's inside. Building an interoperable data ecosystem is not a technical export problem, but a strategic journey that transforms informational chaos into real public value.
Think in domains, not systems
The journey begins, paradoxically, by moving away from technology. Before scanning databases or deploying tools, an organization must understand its information landscape through so-called Information Domains: groupings of data around a coherent and delimited thematic area, such as mobility, health, education, sustainability or social services.
Defining these domains allows us to answer the fundamental question of all governance: who is accountable for this information? When responsibility is assigned to business areas and not to IT departments, we ensure that governance survives organizational changes and inevitable system migrations. It is the step that allows us to break down silos and move towards federated governance models, such as those proposed by the Data Mesh paradigm.
The technical inventory and the role of artificial intelligence
With the map of responsibilities drawn up, it's time to get down to business.eres: the technical inventory of data assets. Systems, tables, files, APIs, document repositories. Doing this work manually in a moderately complex organization —and administrations are— is unfeasible. Automation is not an option, it is a necessity. Self-discovery tools (data crawlers) that connect to databases to extract schemas and metadata are essential to keep the record alive.
However, here we must remember one of the most common pitfalls of data governance: technology is a necessary condition, but never sufficient. Buying a catalog platform does not equal having governance. Data Stewards and business leaders need to provide the context that no tool can infer on its own.
This is where generative artificial intelligence is changing the game. Where once a person had to decipher what an opaque technical field meant, today AI-assisted tools analyze structures, example data, and sparse documentation to propose automatic semantic mappings. AI not only helps build business glossaries, but can also cross-reference consumer demand with existing inventory and suggest which asset combinations have the most potential to become consumer-ready data products.
From raw asset to data product
Having an automated and governed inventory is an important milestone, but inventory is a tool for internal consumption. Not everything we inventory deserves to be cataloged or published. A selection and design process is needed that transforms raw assets into true data products.
A data product is always designed with the consumer in mind, and therefore will rarely be a còpia The exact form of a table. Creating it is an exercise in curation: translating technical language into business vocabulary, linking each field to an unambiguous corporate glossary, applying structural transformations, aggregating microdata to generate useful statistical indicators, and often merging multiple assets to provide complete context.
This transformation cannot be separated from privacy by design. Before any publication, it is imperative to apply Privacy Enhancement Techniques (PET), such as anonymization or pseudonymization, to mitigate ethical and legal risks. Only in this way do we obtain a standardized asset, securely packaged and accompanied by quality metrics and clear usage contracts.
DCAT-AP-ES: the common language to speak to the world
Once the data products have been forged, they need to be exposed. The data product catalog is the enabling service, but if we want our catalog to not be an island, it must speak a universal language. This is where the adoption of DCAT-AP-ES and its sectoral derivatives comes into play.
This application profile, as a structural standard for metadata, is what guarantees interoperability. It allows describing not only what the data product contains, but also who edits it, how often it is updated and under what licenses it is distributed (with vocabularies such as ODRL). Adopting this standard is the gateway to European Data Spaces. A well-built catalogue on this basis ceases to be a simple file finder to become the foundation of digital trust: it enables verifiable identity, observability and, ultimately, the automation of transactions between organizations.
Towards agentic AI and knowledge markets
The journey does not end with publication. The organizations that dominate this production chain are laying the foundations for an immediate future in which the catalog evolves into a true Knowledge Marketplace. In this scenario, the consumers of our catalog will no longer be just human analysts: they will be artificial intelligence agents and smart contracts capable of locating information in federated ecosystems, understanding its legal constraints, negotiating access to it, and composing complex responses in an autonomous and governed manner.
For public administration, this horizon is not science fiction: it is the logical consequence of having done the previous work well.
Conclusion: there is no righteres
The enthusiasm to participate in Data Spaces, to share information with other administrations or to deploy advanced AI use cases cannot make us forget the fundamentals. Skipping the internal discovery and governance phase is the surest way to build unsustainable silos and expose our organizations to legal and reputational risks.
Automation and artificial intelligence offer us an invaluable shortcut today to govern technical assets at a speed and scale unthinkable just a few years ago. But the leap from technical inventory to business catalog continues to demand a user-centric approach, rigor in product design, and commitment to open standards.
The data ceases to be a technical liability hidden on a server when it travels the full path: of → inventories → product → catalog → ecosystemOnly then does it become a reliable, governable asset ready to connect with the world.
At the AOC, we believe that this is the path that all Catalan administrations must take if they want to be actors —and not spectators— of the data economy. And it is a path that cannot be taken alone: it requires common standards, shared infrastructures and a country vision that places data governance at the center of the digital transformation of the public sector.
This is how, from the Smart Local Governments Network, we are already working to make this path affordable for all local entities, regardless of their size and technical capacity. The objective is to make available to the local world some minimum, common and reusable instruments for data governance: reference models, methodological criteria, shared vocabularies, guidelines for inventory and cataloguing and common services that prevent each administration from having to start from scratch. For city councils, county councils and dependent entities, this shared effort must mean less uncertainty, less dispersion, less dependence on closed solutions and more real capacity to transform their data into knowledge, better public services and smarter policies at the service of citizens.
Nota: This article is inspired by the article of Carlos Alonso Pena, Director of Division at the General Directorate of Data, published on Linkedin in May 2026 under the title "From hidden data to hidden assets: The path to Data Products".
