AI Doesn’t Decide Alone: What Italy’s First AI-Regulating Labour Agreement Requires

There’s a date worth remembering: June 9, 2026. That’s the day artificial intelligence stopped being a side note in an Italian national collective labour agreement and became the subject of an entire dedicated section. The draft renewal of the CCNL Funzioni Centrali (Italy’s collective agreement for central government functions) 2025-2027, signed at ARAN, introduces explicit rules on how AI systems can be used in processes affecting employees for the first time.

This isn’t a technical footnote for specialists. It’s a signal that AI governance is moving out of corporate strategy documents and directly into employment relationships, with specific obligations, limits, and accountability.

What Actually Changes

The new agreement sets three firm points.

First: advance transparency.
Before introducing an AI system or automated algorithm that could affect employment, the administration must inform trade unions about its purpose, the data processed, and its possible effects on work organization and performance evaluation.

Second: no fully automated decisions.
Professional evaluations, task assignments, scheduling — whenever a decision significantly affects an employee, human oversight capable of reviewing or overturning it must remain in place.

Third: the right to understand.
Every employee can request, in clear terms, the criteria by which an AI system influenced a meaningful aspect of their work. No more algorithmic “black boxes.”

Why This Matters to Leaders, Not Just Lawyers

It would be easy to read this as a purely legal or contractual matter. That would be a mistake.
The three rules above shift the center of accountability toward the people in leadership positions: whoever decides to introduce an AI system must be able to explain it; whoever uses it must be able to govern it; whoever is affected by it has the right to a human counterpart, not an algorithm.

This is precisely where the quality of leadership is tested today: the ability to adopt advanced technologies without offloading responsibility, keeping the decision — and with it the possibility of error, correction, and judgment — firmly with people.

For HR functions and for those involved in executive search, the message is just as direct: organizations need leaders capable of operating within hybrid, human-machine decision-making processes without losing control or accountability for what they decide. That’s a competency to be actively assessed and developed, not assumed.

An Agreement That May Set the Standard

The CCNL Funzioni Centrali comes shortly after two government decrees establishing Italy’s first rules on the use of AI at work, confirming the same principle: technology can support decision-makers, not replace them.

It’s unlikely to remain an isolated case. It’s more likely to become a reference point for private-sector bargaining too, pushing organizations across industries to formalize what has, until now, been left to informal practice: who decides what, when an algorithm enters the process, and what safeguards are in place for the people affected.