ai-adoption
The One-in-Five Threshold
Italy has surpassed the symbolic threshold of 20% in the adoption of artificial intelligence in businesses, reaching 19.5% during the period 2024-2025. This data, emerged from Unioncamere-Dintec surveys based on over 20,000 observations through Selfi4.0, indicates a growth of 46% in just one year. This is not a gradual evolution or structural consolidation, but a quantitative leap that reveals profound qualitative tensions within the national production system.
The 19.5% metric is not an indicator of uniform technological maturity, but rather a snapshot of a fracture. On the one hand, the Italian AI market reached a value of €1.8 billion in 2025—with an increase of 50% compared to 2024, according to the Artificial Intelligence Observatory of the Polytechnic University of Milan—on the other hand, there is an infrastructural disparity that risks undermining the sustainability of this development. The technology is not spreading like a homogeneous fluid; it is creating areas of high operational pressure in some sectors and deserts of expertise in others.
The Istat ‘Companies and ICT’ 2025 data confirms this asymmetry: 16.4% of companies with at least 10 employees use AI technologies, doubling from 8.2% in 2024. However, the gap between large companies (over 50%) and small SMEs (15%) is a significant 37 percentage points. This difference represents not only an economic scale difference, but also a disparity in computational and organizational capacity that is redefining power dynamics within the Italian industrial value chain.
The Skills Bottleneck
The main bottleneck is not technological, but human. At least six out of ten companies, despite showing interest in AI, are forced to abandon it due to a lack of internal skills. This data, reported by Ansa based on Unioncamere surveys, transforms technology from a strategic asset into an operational constraint. The adoption of AI requires specialized human capital that the Italian labor market has not yet been able to produce at the speed required by the expansion of the market.
The direct consequence of this imbalance is the emergence of a structural dependence on external experts and consultants. SMEs, unable to hire specialized personnel in-house, rely on third-party providers for the implementation of AI systems. This dynamic creates a development model that is external to the internal production capacity of companies, making technological progress fragile and vulnerable to fluctuations in the professional services market. Expertise does not become an asset of the company, but a variable cost that is outsourced.
The mechanism is clear: the demand for AI exceeds the supply of qualified skills, creating a bottleneck that slows down the deep integration into the core processes of companies. Companies are not simply ‘using’ AI; they are negotiating with the labor market to gain access to capabilities they do not possess. This reduces operational resilience and increases transaction costs, transforming a potential competitive advantage into a strategic dependence.
The Governance-Business Friction
While businesses accelerate adoption, the governance infrastructure struggles to keep pace. A global study by the IBM Institute for Business Value involving 2,000 tech executives reveals that 70% of business teams implement AI technologies faster than central IT can monitor them. Two-thirds of CIOs and CTOs interviewed admit to being held accountable for AI systems they do not fully control.
This discrepancy between operational speed and supervisory capacity creates a systemic risk of ‘computational shadow IT’. Business units, driven by competitive necessity, bypass formal technology procurement channels, implementing AI solutions in a fragmented and non-standardized manner. Central IT, deprived of the necessary visibility, loses the ability to ensure compliance, data security, and strategic alignment.
The Italian regulatory framework is rapidly evolving to address this challenge. Law 132/2025, which came into effect on October 10, 2025, represents the first national attempt to translate the European AI Act into a comprehensive legal architecture. However, the introduction of sector-specific rules and the creation of a €1 billion fund for AI investments require implementation timelines that exceed the adoption speed of businesses. Legal governance is lagging behind operational reality.
The Cost of Strategic Slowdown
The asymmetry between large and small businesses, combined with the skills gap and slow governance, creates a hidden cost for the Italian economy. SMEs, which represent the backbone of the national production system, risk being excluded from the benefits of AI not for lack of interest, but due to a structural inability to absorb it.
The result is a polarization of technological progress: large companies consolidate competitive advantages based on proprietary data and algorithms, while SMEs remain confined to traditional operating models or rely on external solutions that do not generate internal know-how. This dynamic threatens the cohesion of the Italian industrial supply chain, creating weak points in the value chain.
The time window for correcting this course is narrow. With the AI Act becoming fully applicable in August 2026 and national Italian regulations in the implementation phase, companies have little time to build robust governance frameworks without first investing in human capital training. The risk is that quantitative growth in adoption (+46%) will translate into a qualitative fragmentation of the production system.
The most important signal to monitor in the coming months will not be the penetration rate of AI, but the ability of SMEs to develop internal skills and for companies to integrate central IT with business processes. The friction between operational speed and structural governance will determine whether Italy will be able to transform its technological potential into a sustainable competitive advantage or remain trapped in a fragmented and dependent adoption model.
Photo by Igor Omilaev on Unsplash
⎈ Content generated by multi-agent AI under Human-in-Command protocol in Epistemic Safety mode. Read the Operational Disclaimer.
> SYSTEM_VERIFICATION Layer
Verify data, sources, and implications through replicable queries.