Artificial intelligence in healthcare: contracts set the first guidelines

Artificial intelligence in healthcare: contracts set the first guidelines

Health is the area where the utmost caution is required: the 2025–2027 legislative term has therefore introduced specific regulations

For feedback, please contact english@ilsole24ore.com

For some time now, the debate and media coverage surrounding artificial intelligence have become a daily occurrence, and AI is perhaps the most significant yet least understood issue within society. Consequently, great caution is required in all areas of civil society, and the field of healthcare is naturally one of the most affected by the development of these automated tools: the economic stakes are enormous and, compared to other areas of interest, healthcare is undoubtedly the most important and deserves the utmost prudence. Having said this, necessarily in brief, collective bargaining could not ignore the issue, and the 2025–2027 round of collective agreements has introduced specific provisions. This is somewhat of a move to keep up with trends, perhaps less so out of genuine necessity or concrete requirements.

Estimates of the impact on healthcare

Obviously, the collective agreement is not concerned with AI in general, scientific, technological or philosophical terms, but only with the implications it may have for the employment relationship. In light of the above observations, it seems at the very least risky to claim that AI can ‘reduce clinical errors’. Medicine is not an exact science but a humanistic one, and this is all too often forgotten. Nevertheless, the Anaao Research Centre has estimated that annual savings of €21.7 billion could be achieved through the systematic implementation of artificial intelligence in healthcare, which could reduce total costs by 10–15 per cent. On this issue, it has been found that around 30 per cent of hospital admissions are inappropriate, that 40 per cent of clinical time is spent on bureaucracy and, finally, that around 15 per cent of tests are duplicated or inappropriate.

The sectoral collective agreement

The preliminary agreement on the National Collective Labour Agreement for the sector, signed on 29 July, includes three specific articles on this issue (11, 12 and 13). There are many areas of intervention, and on a couple of occasions the agreement refers to employee appraisal, which is undoubtedly the most sensitive issue. AI applications are certainly being used, and will continue to be used, for vertical and horizontal career progression and for the allocation of roles. It is to be hoped that these are merely exploratory and supportive measures; otherwise, all the risks outlined in the National Collective Labour Agreement could materialise. AI has also been used for some time in the management of attendance and absences, but as regards possible applications within the employment relationship, it is to be hoped that no one will consider managing disciplinary procedures through fully automated processes.

Provisions in the National Collective Labour Agreement for Local Authorities

From a negotiating perspective, the National Collective Labour Agreement for Central Government Departments of 6 August 2026 set the precedent, containing three specific provisions which have been incorporated into other agreements with very similar wording. In a nutshell, the negotiating parties sought to introduce specific contractual provisions dedicated to artificial intelligence, with the aim of supporting technological innovation whilst ensuring transparency, the protection of workers and the centrality of human decision-making. The agreement provides for the right to prior information on the use of AI systems, a ban on fully automated decisions affecting the employment relationship, and specific training programmes to promote informed and responsible use.

Article 11 serves as a sort of introduction to the subject matter and sets out general principles. Paragraph 2 goes into greater detail, clarifies the ‘healthcare sector’ and provides for the prior provision of information to employees. In this respect, the healthcare sector diverges significantly, as for civil servants the information must be provided to trade union representatives rather than to the individuals concerned: this is, in practice, the most significant difference between the two agreements. In concrete and operational terms, the employer must provide employees with all the information set out in the six points of paragraph 2 in a transparent manner, in a structured, commonly used format that is machine-readable. The possible impacts on the employment relationship are subject to supplementary bargaining pursuant to Article 9, paragraph 5, letter s), added by the new contract. This matter also concerns the Opi (Joint Body for Innovation) and is covered by letter f) of Article 8.

As regards staff training, artificial intelligence technologies and systems are covered by the provisions of Articles 30 and 31. Ultimately, the most significant statements of principle are that AI does not replace humans and that the final decision remains a human act; and, naturally, on a practical level, it is very clear that workers have the right to be informed of the general operating criteria of AI in an ‘understandable form’, an issue that will undoubtedly be one of the most contentious in the implementation of the agreement. Provisions dealing with artificial intelligence will undoubtedly also feature in the forthcoming National Collective Labour Agreement for healthcare managers.