Organisations buy models and get demonstrators; few turn Artificial Intelligence into products that people actually use. The difference, argue Stanford faculty in the Harvard Business Review, lies in the disciplines of product management. A portrait of the AI Product Manager: the craft, the figures of a growing global demand and the Italian benchmark.
The thesis is sharp and comes from one of the most authoritative venues of international management: to adopt AI in earnest, the decisive work consists of “defining valuable problems within workflows, evaluating possible solutions, rapidly experimenting, and integrating new practices sustainably into day-to-day work – disciplines that are core to the work of product managers” (Harvard Business Review). Not the algorithm, then, but the product discipline: identifying the right problem, experimenting quickly, integrating the solution into daily work. This is the territory of the AI Product Manager, the figure who governs the life cycle of products built on AI systems, from priorities to metrics, from requirements to compliance.
Three differences from traditional product management
Product management is a mature discipline; its application to AI, however, introduces three specificities that change its practice.
The first is the product’s constitutive uncertainty: a system built on machine-learning models has probabilistic performance, which varies with the data and over time. The product manager must be able to define acceptable quality, the thresholds for human intervention and the pathways for the cases in which the system errs, because err it will.
The second is the centrality of data feasibility: before the roadmap comes the question of whether the data exist, at what quality and at what cost. More than a few initiatives fail for having chosen AI as the answer before framing the question.
The third is compliance as a product requirement: transparency towards users, human oversight, the handling of personal data under Regulation (EU) 2016/679 (the GDPR) and, for systems falling within the high-risk categories of Regulation (EU) 2024/1689 (the AI Act), documentation and controls prepared in good time. Regulatory requirements enter the backlog like any other requirement, with priorities and owners; treating them as a final rubber stamp is the most expensive way to discover them.
As for the operational repertoire, the industry’s professional pathways converge: the AI product manager is asked to “create high-level product strategies and detailed roadmaps that outline objectives, milestones, timelines, and priorities to guide product development and ensure alignment with business goals” (Microsoft, via Coursera), together with user research, the writing of requirements and the measurement of performance after launch, with Key Performance Indicators (KPIs) defined beforehand, not afterwards.
A demand growing on both sides of the Atlantic
That the labour market rewards those who can build products with AI is suggested by the data on job postings. Stanford University’s AI Index, in its 2026 edition, records – on the basis of data from the analytics provider Lightcast – that “demand for AI talent continued to rise around the world, with AI skills claiming a larger share of overall job postings in market after market”, with peaks, among European markets, in Luxembourg, at 3.4 per cent of postings, and Spain, at 3.3 per cent, above the United States itself, at 2.6 per cent (Stanford AI Index). In Europe the push is also systemic: Decision (EU) 2022/2481 on the Digital Decade sets the target that “the number of ICT specialists employed in the Union is at least 20 million” by 2030, and the figures able to turn technology into product are, by definition, the multiplier of that target.
The Italian benchmark
The UNI 11621-8:2026 standard has included the AI Product Manager among the professional role profiles of AI, with competences set out according to the methodology of the European e-Competence Framework, the UNI EN 16234-1 standard. The assessment and certification of professional competences take place at bodies accredited under the UNI CEI EN ISO/IEC 17024 standard, within the framework of Accredia Information Circular DC No. 21/2026: for a craft that until now has been told mostly in CVs, it is the chance to prove it.
Conclusions
The AI Product Manager is the point of balance between what technology can do, what users ask for and what the law allows: where this figure is missing, AI projects swing between engineering enthusiasm and compliance paralysis, and demonstrators never become products.
In the light of the above, one wonders how many of the AI initiatives now under way in Italian organisations have a product owner worthy of the name; where the answer is uncertain, that is where it is worth investing before buying more technology.
