In the midst of the European drive to reinforce technological sovereignty in Artificial Intelligence (AI), the bottleneck is not only in the infrastructure or computing capacity, but within the organizations themselves. That is the reading that MIOTI Tech & Business School makes in a scenario in which the adoption of AI advances, but continues to collide with a structural barrier: the lack of internal capabilities to understand the technology, govern the data, evaluate risks and make AI adoption decisions autonomously.
European data helps contextualize this scenario. In 2025, approximately one in five companies in the European Union with 10 or more employees used at least one artificial intelligence technology, according to Eurostat, reflecting the progressive growth of AI adoption. However, adoption continues to be highly concentrated by size: while 55.03% of large companies already use AI, the percentage falls to 30.36% in medium-sized companies and 17% in small companies, which shows inequalities in the adoption of AI. Even more relevant to the debate on sovereignty is that, among the companies that considered adopting AI and did not do so, the main obstacle identified was the lack of relevant expertise, a direct brake on the adoption of AI.
For MIOTI, this gap demonstrates that sovereignty in AI cannot be understood only as a question of data centers, chips or cloud providers. Real autonomy begins when an organization has discretion to decide what models it uses, what data it works with, what dependencies it assumes, how it monitors results and under what risk, compliance and business framework it deploys the technology, thus facilitating robust AI adoption. Without that internal knowledge, AI may be available, but it is not truly under control, limiting effective AI adoption. This idea also gains weight in a context in which the European Parliament has warned of the high concentration of the European cloud market around large non-European providers, which increases exposure to technological lock-in and reinforces the need for strategy, governance and own capabilities within companies for a correct adoption of AI.
“We are talking a lot about sovereignty in Artificial Intelligence in terms of infrastructure, and it is logical that this is the case, but the most immediate blockage for many organizations is somewhere else: in their ability to understand AI, govern it and apply it judiciously,” says Fabiola Pérez, CEO of MIOTI Tech & Business School. “Without prepared talent, without data governance and without profiles capable of connecting technology and business, autonomy is very limited, even if the technology is available, which makes the adoption of AI difficult.”
From strategic debate to real adoption capacity
The MIOTI diagnosis fits with the European regulatory moment. The European Union’s AI Regulation, the AI Act, came into force on August 1, 2024 and, from February 2025, requires vendors and organizations that deploy these systems to strengthen the AI literacy of the teams that use and monitor them. The underlying message is that AI adoption must be approached through structure, judgment, and organizational readiness.
In this context, MIOTI defends that this bottleneck for sovereignty in AI plays out on three key levels to promote the adoption of AI:
- Talent: have professionals capable of understanding how models are trained, adapted and evaluated, and of translating that logic to the business, facilitating the adoption of AI.
- Governance: have clear criteria for data use, traceability, risk assessment, human supervision and regulatory compliance, essential elements in the adoption of AI.
- Adoption criteria: knowing in which processes it makes sense to deploy AI, under what architecture, with what suppliers and with what degree of technological dependence, ensuring efficient AI adoption.
Eurostat’s own data show that AI is already being used in areas closely linked to business operations, such as marketing and sales, administration, language analysis, content generation or cybersecurity, which makes this internal decision-making capacity in the adoption of AI even more critical.
“The question we must ask is which person or people in the organization have the real capacity to decide how to integrate these technologies without blindly depending on third parties,” adds Fabiola Pérez. “Sovereignty, for a company, is not only about where the technology is hosted, but whether it has the necessary knowledge and structure to evaluate it, govern it and align it with its business objectives and with the European regulatory framework, thus guaranteeing correct adoption of AI.”
For many leaders, the first step in AI adoption has been purely assimilative
The European Commission has warned that the EU is not yet at the pace necessary to meet its Digital Decade objectives, and has pointed out persistent gaps in both digital skills and the adoption of advanced technologies. Only 55.6% of the European population has at least basic digital skills, and the EU is still far from the goal of reaching 20 million ICT specialists by 2030. At the same time, the community objective is for 75% of companies to use cloud, big data or AI before the end of the decade, which means accelerating the adoption of AI. This distance between ambition and implementation points again to the same problem: without internal capabilities, technological sovereignty runs the risk of remaining a discourse and slowing down the adoption of AI.
For many leaders, the first step in AI adoption has been purely assimilative. In seminars and conferences, AI is presented as a totem of innovation and efficiency. This conceptual familiarity has inflated statistics and in markets such as Spain, around 95% of managers say they are studying how to maximize the value of AI, according to the Kaspersky Gen AI Business Infiltration report.
However, we are facing a mirage of knowledge. Most executive training programs have focused on the “what” but ignored the “how.” This leaves leaders with a peripheral understanding that they know the benefits, but lack a map to translate that data into robust decisions.
