The disruption of Artificial Intelligence (AI) is today indisputable in almost all areas of our society. Recent advances in large Natural Language Models and Generative AI They have placed this technology in the spotlight and have demonstrated its immense potential. However, other AI models are already deeply integrated into our daily lives, often imperceptibly. This is the case, for example, of vehicle driving assistance, the music recommendations we consume, the content shown to us on social networks or the advertising we see daily.
Today, AI is contributing to drastically increasing the productivity and automation of many of the processes and tasks we carry out in both personal and professional spheres, and this trend is expected to become even more pronounced in the coming years. AI is even playing a key role in generating new knowledge through the generation ofartificial hypotheses that are accelerating new discoveries that, without this technology, would have taken many years to achieve. This includes the discovery of new proteins, support in medical diagnoses and advances in experimentation with new energy sources, among many others.
While the great potential and benefits of AI are clear, at the same time it also carries a series of risks that cannot be ignored. The European Union (EU) has been a pioneer in legislating the use of this technology and, recently, a agreement between the US and European governments to work together on its regulation, an announcement that has been joined by Canada and Argentina, among others. Therefore, it is essential that any sector, but especially the public sphere, ensures that the use of AI is carried out under ethical principles, minimizing risks and protecting human rights.
This ethical and security component is crucial, especially in the public sector. Great care must be taken with the use of AI when it comes to managing documents, personal data or sensitive information. From the administration, we are focused on theanonymization and the clustering of sensitive data for inferences made with AI. For more complex projects, licenses of the type are contracted Enterprise, which include clauses that guarantee that information and data remain within the environment Lieutenant or Cloud of the domain contracted for the service. This is volatile data used only in AI inference, thus preventing supplier companies from having access to or being able to use it for re-train models or sell them to third parties. Another security solution that large administrations have already begun to implement is thehosting Open Source models on our own servers, although this option presents scalability difficulties at the local level.
Now, what can local administrations do to introduce AI into their public service? As in any company, and taking into account that administrations must also be examples of efficiency and productivity, the first step is to provide training in the use of AI and equip public workers with AI tools to accompany them in their usual tasks. The entire public sector value chain, from programmers and planners to administrators, operators and managers, can benefit from the use of AI to improve productivity, automate processes and increase efficiency.
In addition, local administrations can begin to implement pilot projects linked to AI that can be scaled in the future. A good example is the case of theBarcelona Metropolitan Area (AMB), which already has a citizen service chatbot based on AI to answer and resolve questions about the new public transport access support, the T-Metropolitana. This chatbot is connected with APIs to GPT-4o and specializes in all the documentation – public and without sensitive data – of the T-Metropolitan, including manuals for citizen service operators, web content and documentation on frequently asked questions (FAQ) of this product. Railways of the Generalitat (FGC) also already has a very similar tool in production on its website focused on offering service status information and the routes of its railway network.
It is recommended that local administrations rely on collaborations with technology centers, foundations and universities specialized in these fields to develop more complex joint projects, which can benefit from funding. There are already many initiatives in the public sphere that are in the development and testing phase, ranging from citizen care to analytics and prediction projects to improve the planning and maintenance of public services, as well as initiatives to facilitate administrative management, especially in the field of public procurement and fraud detection. For example, AMB Information and Services is already working on projects supported by public subsidies to develop AI models of predictability of the Barcelona metropolitan transport and mobility network, with the aim of obtaining advanced tools that improve the planning and information of public transport and micro-mobility services.
However, none of these projects would be viable without a proper data governance. An AI project cannot succeed without excellent data management. It is not just about having databases, but about placing data at the center of organizations, applying processes, policies and technologies that guarantee the quality, efficiency and accessibility of information. For this reason, it is essential to deploy appropriate architectures for hosting, ingesting and preparing data, with the necessary technical and human resources. Although this is a considerable challenge, it is important to make it clear that this is the starting point that administrations must follow to achieve a more efficient provision of public services, more informed decision-making and a more personalized, satisfactory and transparent relationship with citizens.
In parallel, the Generalitat of Catalonia, alongside the State and the EU, is also encouraging the deployment of Data Spaces, a key infrastructure to promote data sharing between public and private entities in a federated, secure and standardized manner. In the case of the private sector, with the participation of public entities as well, these data spaces are also being coordinated around initiatives such as: Gaia-X. The purpose of Data Spaces is to create environments that facilitate data interoperability, offering a structure where participating actors can share information without losing control over it, guaranteeing their privacy and respecting the rights associated with the data. The importance of these shared data ecosystems lies not only in the technological and operational benefits, but also in the development of governance models and economic viability. The joint participation of companies and institutions is key, since without a common, agreed, regulated, economically viable and maximally standardized federation, neither the private nor the public sector will be able to fully take advantage of the value of data, and therefore the transformative potential of Artificial Intelligence.
