Italy’s AI Act under Legislative Decree 179/2026: 3 things to do if you build software or consult for SMEs

If you sell digital services to SMEs, the first job right now is understanding how your clients already use AI. Since 23 October 2026, Italy’s Legislative Decree 179/2026 has been in force. It puts the AI Act into effect in Italy and sets specific rules on using AI in decisions about workers. Meanwhile, in almost half of SMEs, employees are already using AI on their own initiative.

What changes with Legislative Decree 179/2026?

Legislative Decree 179/2026 is dated 7 October 2026, was published in the Official Gazette on 8 October and came into force on 23 October. It aligns Italian law with the EU regulation on artificial intelligence, the AI Act.

For anyone building software, the most concrete part of Decree 179/2026 concerns employment. According to EC News’ analysis, the decree establishes that:

• final decisions on hiring, performance reviews, disciplinary measures and dismissals cannot rest solely on automated processing;
• the final decision belongs to a person with real, independent authority;
• consent from the worker or candidate does not make a fully automated decision lawful;
• a dismissal that breaks the ban is void;
• workers can request an understandable explanation, stating the weight of the AI system and the main parameters it considered.

How much are Italian SMEs already using AI?

Italian SMEs already use AI, often without really knowing how to manage it. That is what the Nomisma-UniCredit AI.DEA study shows. It was presented on 16 October 2026 and covers over 200 SMEs with 10 to 250 employees in the machinery, agrifood and Motor Valley supply chains. The key figures, reported by ANSA:

• in 45% of cases AI is used informally, on the initiative of individual employees;
• 50% of companies name the lack of in-house skills as the main obstacle;
• two SMEs out of three do not use AI in a structured way.

The picture is clear: AI is already inside companies, and it often runs without shared rules. For anyone building software or offering digital consulting, there are at least 3 things to work on.

1. Why should human oversight be designed in from the start?

Human oversight has to be designed early because adding it later can require major changes to the system. If you build tools that handle CVs, performance reviews or shift schedules, you need to decide from day one who checks the AI’s decisions and what gets logged.

In practice, that means designing the flow with an explicit approval step: the AI proposes, a person decides. Decree 179/2026 requires exactly this for final decisions about workers. An approval step only works if the approver has the time and information to do it properly: I covered this when writing about approval fatigue with AI agents.

When I build custom AI-powered software for companies, this is one of the first points I lock into the specs, before writing any code.

2. What should you explain to the client about the controls?

When you propose an AI solution to an SME, you need to make three things clear:

1. who can approve or correct the system’s decisions;
2. which activities are logged;
3. what happens when the system gets it wrong.

For a company that wants to use AI without losing control, these answers matter a lot. They are also the questions a worker can ask: the decree gives workers the right to request an understandable explanation of a decision. If the software records who decided and on which parameters, answering becomes simple.

3. Why start from the AI tools the company already uses?

Before proposing new automations, ask employees which AI tools they use and what for. With AI entering 45% of SMEs through individual initiative, the real map of usage is often different from the one the owner has in mind.

Understanding what data employees enter and which tasks they hand off to AI brings concrete problems to the surface. Sometimes it is a risk to company data, as I described when writing about company data sold to train AI. Sometimes it is a repetitive task that deserves a well-built tool with the right controls.

What does this mean for software developers in the coming months?

In the coming months, building useful AI solutions and helping companies manage them well can become a major advantage. Half of SMEs say they lack the in-house skills: whoever brings those skills, together with a system that has clear controls, meets a need that already exists.

Often the first opportunity is understanding what is happening inside the company, before writing a single line of code. The right time to ask your clients that question is now.