Il 2025 was a turning point for Artificial Intelligence in Italy. The national market has reached quota 1,8 billion euros, scoring a 50% jump compared to the previous yearIt is an unprecedented growth which certifies AI as one of the central directions of the country's digital transformation, according to the data of theArtificial Intelligence Observatory of the Milan Polytechnic.
Nearly half of the total value, 46%, is attributable to Generative AI solutions or hybrid projects, while the remaining 54% continues to rely primarily on machine learning applications. This ecosystem is also expanding on the supply side, with over a thousand registered companies currently operating in the sector and 135 startups funded in the last five years occupying vertical niches, especially in areas such as healthcare and financial services.
From experimentation to business impact
La spread of AI In large Italian companies, it is now a structural fact. In 2025 71% have started at least one Artificial Intelligence project, a significant increase compared to the previous year. Only one in five, however, manages to use it pervasively across multiple business functions. Despite this limitation, six out of ten companies report a significant impact on their business model, whether it concerns the value proposition, customer relationships, or operational structure.
The most evident phenomenon is theexplosion of ready-to-use solutions. 84% of large enterprises have purchased Generative AI licenses, an increase of 31 percentage points in just one year. ease of adoption and low cost of access They are pushing AI out of the lab and into everyday processes, although full organizational integration remains a challenge.
Work is changing: productivity, new capabilities, and Shadow AI
Artificial Intelligence is already redesigning the work. On average, 47% of workers use tools of AI in the company and a significant portion perceives concrete benefits. Four out of ten workers estimate a saving of more than half an hour in the last tasks performed with the support of AI. But the most significant data concerns the skills. 41% say they can, thanks to Artificial Intelligence, perform tasks that they would otherwise be unable to do or could only partially do.
Alongside the opportunities However, evident critical issues emergeOnly 19% of users say they use company tools exclusively. Eight out of ten workers use external solutions, reporting a widespread phenomenon of Shadow AI which risks escaping the control of organizations.
The Race for AI Skills in Job Postings
La transformation is also strongly reflected in the labor market. In 2025 the number of ads which require Artificial Intelligence skills grow by 93% on an annual basis. Of over 3,2 million positions advertised, approximately 44 include AI-related skills. Demand is primarily focused on highly qualified white-collar profiles; these skills are now considered a requirement in 76% of job postings.
AI also enters non-technological top roles, appearing in over a quarter of Chief Human Resources Officer ads and a significant share of Chief Marketing Officer ads. Demand is shifting toward advanced skills such as Generative AI and AI Agent, confirming that Artificial Intelligence it is no longer a technical support, But a strategic leverage to rethink organizational and production models.
Governance and rules: the great Italian delay
Faced with a rapid spread, the AI governance remains the real weak link. Only 9% of large Italian companies have a government structure of Artificial Intelligence with clear responsibilities and alignment with ethical principles and corporate objectives. More than half are working to close the gap, but adaptation to the new European rules is proceeding slowly and only 15% have started structured projects in line with the AI Act.
The risk is that of a disorderly growth, driven more by enthusiasm than by a long-term vision.
The challenges of 2026: investments, sustainability, and people
“2025 confirmed the huge market growth and the technological development of AI, which is now absolutely central to the agendas of top decision makers", observes Alexander Piva, director of the Artificial Intelligence Observatory, "this enthusiasm, however, requires us to stop and think. First, on the still limited ability to recognize the right ways to rethink entire processes with AI in every sector and area: we need people with extremely high domain and technological expertise to deconstruct, reimagine, and re-establish the operating model. And then on the need to move from the simple individual adoption of AI, which is now widespread, to the structural transformation of organizations, which is still limited, which requires well-organized and usable data, widespread technical skills, and a corporate culture open to experimentation."
For Giovanni Miragliotta, director of the Observatory, the new year opens with decisive challengesThe first is finding a balance between expectations and the actual benefits of adoption, which often materialize only after progressive and personalized implementation processes. The second challenge is to continue with research and training programs despite the depletion of PNRR resources: the lack of a strategic plan to finance AI development in Italy risks undermining the development of recent years. The third challenge, of global scope, concerns the financial sustainability of the enormous investments underway, which add to the risks of predatory profit-making, the expulsion of people from the labor market, disinformation, and systematic surveillance.
In the meantime, 2025 has already delivered its key word"Agentic AI"As Nicola Gatti, director of the Observatory, explains, it symbolizes the "power of combinatorial innovation," destined to fully mature only with a convergence between probabilistic models and logical reasoning capabilities. Until then, the "human-in-the-loop" approach remains not only recommended, but necessary.
"Agentic AI will reach full technological maturity when there is full convergence between probabilistic cognitive engines and native logical reasoning and self-correction capabilities, which will ensure robustness in complex processes," explains Gatti. "Until then, the 'human-in-the-loop' approach is not only recommended, but necessary. Furthermore, it will be very important to monitor the impact of the end of the extraordinary investment plan linked to the PNRR in 2026 and how Europe intends to intervene to strengthen its current position."
