The Italian business fabric records the highest gap in Europe between the trust placed in artificial intelligence systems and the maturity of the control and governance infrastructures necessary to manage them safely, as highlighted by the results of the study carried out by SAS with research insights from IDC. During 2026, the index measuring perceived trust in technology stands at 72.9 points in the national context, compared to a reliability index stuck at 58.8 points.
Although the gap has narrowed compared to the previous year, going from 18.4 to 14.1 points, the Italian gap remains the most marked at a European level, highlighting a persistent distance between algorithmic expectations and the actual ability of organizations to demonstrate the transparency, explainability and regulatory compliance of the systems used. Such asymmetry directly affects the profitability of digital projects, according to the Data and AI Impact Report, organizations that adopt consolidated practices of reliable artificial intelligence have a significantly higher probability of obtaining a significant economic return compared to the market average.
Despite the improvements recorded in terms of accountability, model governance and data quality, all the dimensions analyzed remain below the global benchmarks, highlighting operational criticalities linked above all to the lack of clarity of the responses generated by the systems. The problem is not just using AI, but understanding why it responds in a certain way. According to the research, 37.7% of organizations indicate lack of clarity in AI-generated responses as the main reason that leads to modifying or correcting the outputs produced by the systems.
Next comes the lack of context, indicated by 26.4% of organizations. Concerns related to the safety of outputs accounted for 14.2%, factual inaccuracies 13.2% and risk of bias 8.5%. With the increase in automation and the entry of artificial intelligence into increasingly relevant business processes, explainability thus becomes an essential component of trust: it is not enough that a model works, organizations must be able to verify how and why it produced a certain output.
Further slowing factors derive from the fragmentation of databases, with a share of companies that manage information in organizational silos almost tripling in just one year, while only a minority adopt automated quality controls on each algorithmic project.
Overcoming this imbalance represents the central issue for transforming technological adoption into a measurable competitive advantage, linking future productivity growth not to the mere experimentation of models, but to the structural consolidation of control, audit and transparency processes.