Privacy slows down the use of AI among HR leaders

From salary transparency, to pay gaps, to performance and workload evaluation, to recruitment and selection of people, skills, and motivational surveys. While agentic AI would have high potential to manage all the data related to these topics, their sensitivity and privacy slow down its use by HR directors. Added to this are the indications of the EU AI Act, which classifies artificial intelligence systems applied in recruitment and resume screening as high-risk activities and imposes certain strictness on transparency, traceability, and especially human supervision. This explains why HR directors on one hand talk about the significant impact of agentic AI in the next two years and the high potential of the tools, as 83% say, but then use it very cautiously and only one in five, 21%, already has operational solutions. The gap between these two figures is very wide and emerged from a survey during a workshop held by the Italian Association of HR Directors of Lombardy and Dgs, which deals with implementing AI and cybersecurity, also applied in personnel management. Fifty managers were involved who addressed topics such as skills management, onboarding, worker experience, and salary transparency. While only 21% have AI solutions already operational in the various areas, 33% are exploring and 25% have active pilot projects. Skills management appears to be the most critical area. Indeed, 65% of HR managers say they do not have an updated and structured situation of available skills or only have partial mappings, poorly integrated into decision-making processes. Only 17% oversee the issue with development and reskilling plans. Artificial intelligence applied to human resources is a large, still open project. Elena Panzera, president of Aidp Lombardia, highlights that “AI is redefining models and processes, the dialogue between managers and experts offers valuable insights, but above all examples and solutions already applicable to make processes more effective, inclusive, and people-oriented and help the HR function transform technological innovation into a real advantage for organizations and those who work there.”

Read more Who is Natalie Harp, the Trump assistant who sparked the White House’s ire against the …

Translating momentum into practical applications is less easy than one might think in other areas. Vincenzo De Giovanni, Head of SAP Excellence at DGS and responsible for HR projects, observes that “applying agentic artificial intelligence to human resources does not just mean implementing algorithms and programs, as could be done in other business areas, such as supply chain for example. There is still very fragmented and in some cases incomplete data, such as those related to skills, especially in a field where data is significantly sensitive for privacy and compliance reasons. For analyses and outputs generated by AI to be reliable, the data must be actual, clean, secure, systematized, and guaranteed, fully aligned with regulations on personnel information processing. The problem is that most companies, even large ones, have not yet fully met this requirement, which hinders the smooth implementation of AI-based procedures.”

When talking about AI, most companies provide workers with tools to support knowledge, allowing them to ask questions and get guidance answers. In reality, agentic AI has much higher potential, De Giovanni points out. Take, for example, a large company where hundreds and hundreds of onboarding processes occur every year, involving a series of very precise and complex steps that the worker must navigate. “Today,” says De Giovanni, “through agentic AI it is possible to process data and automatically provide the employee with tasks in the right order and at the right time to perform them, also based on their onboarding role. The worker thus completes their onboarding process as quickly and simply as possible. However, updated information collected continuously and systematized is needed, a data culture approach that in human resources is often not at the level to guarantee such an evolutionary leap: this is a barrier. Onboarding in companies is still a very routine process that does not take into account role, seniority, and organizational context: onboarding personalization averages only 2.35 out of 5, with 62% of respondents positioned at the lowest levels of the scale.”

Read more There is a shortage of 315,000 AI experts: here is where companies struggle the most to find them

On the other hand, if we take a chapter like salary transparency, “EU regulations somehow push companies to liberalize a series of information that was a secret garden until recently. AI in this area could be a great enabler, but the data in question are very delicate and potentially risky to entrust to AI management. That’s why guardrails are needed within controlled systems and data, with clear and stringent rules of the game, carefully designed because when an AI algorithm is used in an uncontrolled way, gaps can arise with impacts even on corporate reputation and industrial relations,” explains De Giovanni.

The survey showed that regarding agentic AI, 50% of managers, one in two, consider adapting policies, processes, and systems a priority, and 42% worry about managing employee expectations: the critical issue appears to be the quality and privacy of internal data, to be resolved even before any external application. Large investments are needed starting from data. Even in previous technological waves, factors that slowed down application to human resources were always linked to the sensitivity of the data to be collected, systematized, structured, and always managed in compliance with regulations. Lack of AI skills in the HR function and budget constraints are indicated by 46% of managers as the main structural barriers, followed by data fragmentation mentioned by 42%. The problem does not seem to be the willingness to change, but the tools and the will to do so.

Read more Software, data, brands: 10 trillion on intangible assets. Italy behind