Data on AI adoption in Italy paint a varied picture depending on what is being measured. According to the EY Work Reimagined Survey 2024,<\/a> 80% of Italian businesses have adopted generative AI in the workplace, a figure that places the country above the global average of 75%. Conversely, the Istat ICT Report 2025 <\/a>notes that only 16% of companies with at least ten employees use artificial intelligence in a structured manner: a figure below the EU27 average of 20%, light years away from the Nordic benchmarks (Denmark at 42%, Finland at 38% \u2013 Eurostat data). <\/p>\n\n The contradiction is only apparent. The EY survey measures individual use in work contexts, including personal and unauthorised use of consumer devices. The Istat figure, on the other hand, reflects formal, company-managed adoption. The gap between the two figures is, precisely, the scale of the problem: a large proportion of usage takes place outside the scope of any governance. <\/p>\n\n This gap is substantiated by the research report *The State of Marketing Automation and Artificial Intelligence in Italian Businesses*, conducted by Alessio Sorrentino with the support of Floox and in collaboration with Professors Simone Guercini and Matilde Milanesi of the University of Florence, and presented in Milan on 27 May 2026. The study is based on semi-structured interviews with C-level executives and managers from companies of varying sizes and sectors, representing a combined turnover of approximately \u20ac227 billion, and focuses on \u2018how\u2019 rather than \u2018how many\u2019: not how many companies adopt AI, but how they integrate it, what they gain from it, and what impact it has on people. <\/p>\n\n One of the most revealing findings of the research concerns precisely the undeclared use of AI tools. Among the barriers to adoption documented by the interviewees, one condition stands out which the researchers describe as \u2018clandestine use\u2019: the technology is perceived as useful, but has not yet been officially endorsed by the organisation. One manager describes it with a phrase that speaks volumes: \u201cThere\u2019s someone who\u2019s started using it in a bit of a clandestine way. Whether you use it or not, it\u2019s best not to say so, because who knows what they might think.\u201d <\/p>\n\n In other words, the tool is operating behind the scenes because the organisation hasn\u2019t yet decided what to do with it, and in the meantime, the data is being processed.<\/p>\n\n The Floox-Unifi survey paints a picture of adoption that is still far from true integration. 40% of the companies surveyed have not yet integrated AI into their processes, whilst 35% use it as a tool for individual productivity \u2013 to generate copy, assist with translations or carry out analyses \u2013 without incorporating it into decision-making or automated workflows. Only 25% use it in a predictive or advanced manner. It is precisely in that middle ground\u2014that of individual and unstructured use\u2014that the phenomenon of shadow AI lurks: tools used by individuals, outside any corporate framework, with data flowing out without anyone recording it. <\/p>\n One of the most counterintuitive findings of the research concerns the nature of the barriers. Among those surveyed, 60% identify human factors \u2013 resistance to change, lack of skills, internal scepticism \u2013 as the main obstacle. Technological barriers account for 30%, economic ones for 10%. It is not the budget, nor the platforms; it is the organisational culture: SMEs do not have the time to implement what would save them time, a self-perpetuating vicious circle. Then there is a cognitive reason: the lexical gap precedes and causes the technological gap; those who cannot name a concept cannot buy or use it. The third reason is a question of power: those who resist adoption are not acting irrationally; they are defending an established information balance that technology would reshape. <\/p>\n\n This internal dynamic is also borne out by the quantitative data from the Deloitte-CSA report on security: only 9% of security managers say they have full visibility over the AI initiatives currently underway within their organisation, and 86% of companies have not yet allocated a dedicated budget for the security of AI systems. There is no governance because there is still no awareness of what needs to be governed. <\/p>\n\n There is a direct link between unauthorised use and tangible risk. When an employee uses unauthorised AI tools, the data entered into prompts \u2013 such as contract texts, customer information, source code and business strategies \u2013 is transmitted to third-party infrastructure, often based outside the European Union, under contractual terms that the company has never negotiated. Only 12% of Italian organisations have integrated risks related to generative AI into their formal risk management frameworks. Less than 10% carry out ongoing assessments of AI technology providers. 97% have not implemented DevSecOps processes specifically for generative AI. <\/p>\n\n In its *Global Cybersecurity Outlook 2026*, the World Economic Forum reports that 87% of security experts identify AI-related vulnerabilities as the fastest-growing cyber risk in 2025. Data leaks associated with generative AI are cited as the top concern for 2026 by 34% of the global sample. <\/p>\n\n Law No. 132 has been in force in Italy since 10 October 2025; it transposes the principles of the European Artificial Intelligence Regulation and introduces specific obligations for high-risk systems: inventories, roles, data assessments, technical documentation, logging and human supervision. For organisations that do not know which AI tools their employees are using, building a compliant inventory is a challenge that begins long before formal compliance. <\/p>\n\n Italian businesses that have integrated advanced digital tools demonstrate 12% higher productivity and a 67% greater propensity to export compared to those that have not (Istat ICT 2025). The gap in structured adoption between large companies and SMEs (37 percentage points, widening from 20 in 2023) translates into a measurable competitive disadvantage, which is also replicated geographically: 31% of SMEs in the North-West adopt AI, compared with 20% in the South. <\/p>\n\n But the unregulated use of AI is, of course, not the solution to the lag; on the contrary, it widens the gap. An organisation where employees use unauthorised AI tools is not adopting artificial intelligence; it is exposing its data without reaping the systemic benefits. Shadow AI is not digital transformation: it is a digital headache (opening image created with ChatGPT) <\/p>\n","protected":false},"excerpt":{"rendered":" The adoption of generative AI in Italian businesses is gaining momentum, largely outside the scope of planned initiatives, according to new research<\/p>\n","protected":false},"author":234,"featured_media":174918,"comment_status":"open","ping_status":"open","sticky":true,"template":"","format":"standard","meta":{"_acf_changed":false,"_substack_draft_id":"","_substack_draft_url":"","_substack_last_pushed":"","_substack_premium":"","footnotes":""},"categories":[1335],"tags":[1324],"companies":[],"journalist":[1512],"class_list":["post-174919","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-report","tag-artificial-intelligence-en","journalist-donatella-cambosu-en"],"featured_sizes_urls":{"thumbnail":{"src":"https:\/\/www.startupbusiness.it\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-9-giu-2026-10_45_34.jpg","width":150,"height":84,"crop":false,"srcset":false,"alt":"Italian companies AI"},"large":{"src":"https:\/\/www.startupbusiness.it\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-9-giu-2026-10_45_34.jpg","width":1024,"height":576,"crop":false,"srcset":false,"alt":"Italian companies AI"},"2048x2048":{"src":"https:\/\/www.startupbusiness.it\/wp-content\/uploads\/2026\/06\/ChatGPT-Image-9-giu-2026-10_45_34.jpg","width":1672,"height":941,"crop":false,"srcset":false,"alt":"Italian companies AI"}},"acf":[],"yoast_head":"\n“In a somewhat secretive manner”<\/strong><\/h2>\n\n
An adoption that has not yet led to integration<\/strong><\/h2>\n\n
![\"\"](\"https:\/\/www.startupbusiness.it\/wp-content\/uploads\/2026\/06\/Screenshot-2026-06-09-alle-10.51.06.png\")
<\/figure>\n<\/div>\nTechnology is not the problem<\/strong><\/h2>\n\n
No one sees, no one governs<\/strong><\/h2>\n\n
The cost of absence<\/strong><\/h2>\n\n