26

Immagine creata con IAAI-generated image

UNI 11621-8: the twelve professional profiles of Artificial Intelligence, between national standard and European CWA

On 30 April 2026 the standard UNI 11621-8 was published, the first national standard in Europe to systematically define the professional role profiles of Artificial Intelligence: twelve figures, from Chief AI Officer to AI Research Scientist, certifiable under accreditation. A few weeks later came CEN’s CWA 18398 on AI professional and educational profiles. Two documents alike in theme but deeply different in nature: let us clarify, including on the far-from-simple front of implementation.

The author took part personally, together with the other experts of the working group, in developing the standard discussed here: this is therefore the right place to present its framework and to address a knot emerging forcefully in practice — the relationship between national standardisation on professional profiles and the new European deliverables.

The standard: twelve third-generation profiles

UNI 11621-8:2026, “Unregulated professional activities — ICT professional role profiles — Part 8: Professional role profiles relating to Artificial Intelligence (AI)”, entered into force on 30 April 2026. As the official summary states, “the standard defines the third-generation professional role profiles relating to activities carried out in the field of Artificial Intelligence, using the principles defined in UNI 11621-1”. The significance was underlined by the Department for Digital Transformation of the Presidency of the Council, which coordinated the work: “this is the first national standard in Europe to define, in a systematic and structured way, the professional role profiles operating in the field of Artificial Intelligence”. The regulated profiles are twelve: Chief AI Officer, AI Consultant, AI Product Manager, AI Prompt Engineer, AI Algorithm Engineer, AI Deep Learning Engineer, AI Data Engineer, AI Data Scientist, AI Security Specialist, AI Machine Learning Engineer, AI Natural Language Processing Engineer and AI Research Scientist. For each, “mission, main tasks, expected results, competences, knowledge, skills and key performance indicators (KPIs)” are specified, within the UNI 11621 series which since 2021 declines ICT profiles on the methodology of the European e-Competence Framework (UNI EN 16234-1, e-CF).

The certification architecture: 11506, 17024, Law 4/2013

Here is the point distinguishing the UNI 11621 series from a mere repertoire of job descriptions: the profiles are built to be certifiable. The standard operates in conjunction with UNI 11506:2021, the “mother” standard describing the elements for assessing conformity of knowledge, skills, autonomy and responsibility (as defined by the e-CF). The certification of professional competences, following assessment of the candidate, takes place under accreditation, per UNI CEI EN ISO/IEC 17024; and it is from June that Accredia Information Circular DC No. 21/2026 defines “the rules for starting and maintaining the accreditation […] of bodies certifying persons intending to operate, in conformity with UNI CEI EN ISO/IEC 17024:2012, in the certification scheme referring to UNI 11621-8:2026 and UNI 11506:2021”. Note, for completeness, that the new 2026 edition of ISO/IEC 17024 has meanwhile been published, set to replace the 2012 one recalled by the circular according to the transition rules the accreditation bodies will define. The legislative frame is that of unorganised professions: Law No. 4 of 14 January 2013 and Legislative Decree No. 13 of 16 January 2013 on the national system for the certification of competences.

The CWA 18398: what it is, and what it is not

A few weeks after the Italian standard, on 24 June 2026, CWA 18398:2026, “AI Professional Role Profiles and Educational Profiles”, was published — the outcome of the CEN Workshop launched in April 2025 to complement the European ARISA project. Two documents on the same theme, then. But beware: their legal nature is profoundly different, and confusing it generates application errors already visible in practice. The CWA (CEN Workshop Agreement) is, by official definition, “a CEN and/or CENELEC deliverable, developed by a […] Workshop, which reflects an agreement between identified individuals and organizations responsible for its contents”: an agreement among workshop participants, developed by a fast procedure with participation limited to the interested parties, without the national public enquiry that characterises actual standards. The consequences are set by CEN-CENELEC Guide 29: “the CWA is not assigned the status of a European Standard […]. CEN and/or CENELEC National Members are not obliged to adopt it nor to withdraw national standards in conflict with this CWA”; moreover “a CWA is not designed to support European legislative requirements”, and has a limited validity in time (three years, renewable). In other words: CWA 18398 is not a European standard, imposes no withdrawal or adaptation of UNI 11621-8, grounds no presumptions of conformity and coexists with national standards. The two documents operate on different planes: the UNI standard, developed with the full consensus process, grounds in Italy accredited person-certification schemes; the CWA offers a shared European reference, precious above all on training paths.

The implementation difficulties

Finally, we cannot conceal that implementing this framework is meeting operational difficulties. First, the coexistence of references: certification bodies, training providers and clients face several maps of AI profiles — the national standard, the European CWA, market repertoires — and the temptation to treat them as equivalent is as strong as it is wrong. Second, the construction of examination schemes: defining examination requirements, minimum experience and independence rules descends from ISO/IEC 17024 and the scheme regulations, and requires examiners who master the e-CF competences of the profiles at an adequate level. Finally, the coordination with the regulatory framework: the profiles recall conformity with the AI Act, the GDPR and UNI CEI ISO/IEC 42001, and the qualification of persons will have to integrate, without overlapping, with the organisational obligations on providers and deployers of AI systems.

Conclusions

With UNI 11621-8 Italy has made a choice of field: to give the AI professions names, competences and a tool of third-party verification, within the frame of Law 4/2013. CWA 18398 confirms that the theme is now European; but the difference between a workshop agreement and a consensual standard remains, and it is well that the market knows it. In the light of the above, one wonders whether the ecosystem — certification bodies, trainers, businesses and professionals — will translate this framework into serious certification schemes, and not into the mere issuing of certificates, and whether the certified qualification of persons will keep pace with a technology that redraws its own professions faster than any standard can capture them: a challenge that concerns not only standardisation, but the very future of qualified work in the age of Artificial Intelligence.