Artificial intelligence has definitively entered a phase of maturity. After a few years marked by hype , grand models , and a narrative focused almost exclusively on technological power , 2025 is shaping up to be the year in which the focus shifts to structural issues: governance , accountability , strategy , sustainability , and public value .
The articles published on the CIDAI blog throughout this year clearly reflect this paradigm shift , which is also confirmed by the large number of proposals presented as success stories at the AI Congress Barcelona . Taken together, these contributions demonstrate a clear evolution in the debate: AI is no longer valued solely for what it is capable of doing, but for how it is responsibly integrated into real-world environments .
Based on the outreach activities promoted by CIDAI during 2025, we can conclude that the debate can no longer remain focused solely on the capabilities of these systems, but must address how these systems are incorporated into decision-making processes with social and economic impact . A brief summary and analysis of this evolution throughout the year follows.
The deployment of AI in real-world contexts highlights an uncomfortable but unavoidable truth: a model can be technically excellent and, at the same time, inadequate for decision-making . Furthermore, according to the MIT study, State of AI in Business 2025 , 95% of generative AI pilots are failing in companies. It should be noted that this study is based on US companies and that, in most cases, these AI-based pilot projects do not have a direct impact on the bottom line of the companies analyzed.
In complex environments, trust stems not only from the accuracy of models , but also from the ability to interconnect them with business processes and, in turn, explain, justify, and contextualize the results. This approach is particularly relevant in the European context, where the legitimacy of technology is closely linked to transparency and accountability.
Ultimately, AI can only scale sustainably if it is perceived as a reliable and governable tool, and not as an opaque artifact.
Another central element of the 2025 debate is the reassessment of the usefulness of large language models, essential engines of generative AI. Despite the undeniable advances these technologies have brought, the CIDAI articles highlight the value of smaller, more specialized, efficient models adapted to specific applications and contexts. This shift is not only technical but also profoundly strategic. Energy sustainability , cost control , data protection , and the ability to deploy proprietary solutions are becoming key factors for technological sovereignty. Initiatives such as AI Factories and European projects like DIH4CAT respond precisely to this need to equip the productive sector with infrastructure, knowledge, and the capacity to act.
From Catalonia, the Catalonia AI 2030 Strategy represents a clear commitment to AI rooted in the region, connected to research, geared towards SMEs, and aligned with public policies . It’s not just about adopting technology, but about building capabilities, talent, and innovation within our community.
The issue of delegating decisions is probably one of the most mature, as well as most delicate, topics addressed in CIDAI’s articles. In 2025, it makes clear that the question is no longer whether we will delegate to AI, but to what extent, under what conditions, and with what control mechanisms.
Explainability becomes a functional requirement, not merely an ethical consideration. Without interpretability, there is no effective oversight, and without oversight, there is no accountability. In systems involved in critical decisions, human judgment cannot be replaced; rather, it must be reinforced through the use of these technologies.
This implies explicitly acknowledging the limitations of models, designing hybrid decision-making processes, and accepting that error, both human and algorithmic, is inherent to the system. Technological maturity consists precisely in knowing how to manage these limitations consciously and responsibly.
2025 does not represent the end of the road, but rather the beginning of a more demanding stage. A stage in which AI ceases to be a promise and becomes a critical infrastructure for society and the economy. It is in this context that Catalonia and Europe have the opportunity and the responsibility to lead the way in an ethical, humane, and strategically sound model of artificial intelligence , capable of generating trust, public value, and technological sovereignty.
