AI competence: Training is perfect for compliance – but otherwise insufficient

Contents

At Gartner's recent EMEA symposium, the topic of AI competence and the associated opportunities and responsibilities took center stage. The background was that since February 2, 2025, providers and operators of AI systems are obliged to ensure sufficient AI competence for all individuals involved in the development or operation of AI systems. This is not just a new training measure, but a milestone in AI utilization, as AI is officially no longer a playground for innovation labs and pilot projects, but is accepted as part of regular workflow.

General AI usage is still in its infancy. According to HP's current Work Relationship Index, only 23 percent of employees in Germany use AI, and only 14 percent consider themselves competent in handling AI. This gap between usage and competence is, on the one hand, the reason for the legal requirements, and on the other hand, it means significant business problems, because technology that is used incorrectly or inadequately is practically meaningless.

AI is poorly utilized due to lack of acceptance

Gartner analysts see it the same way. Their concise assessment: “AI is generally poorly utilized because it is not accepted by employees, not understood, or not meaningfully integrated into work.” Gartner thus contradicts the widespread management thesis: “If the use case is good enough, usage will come on its own.” Gartner calls this a myth and presented a lot of survey data to debunk it. For example, 87 percent of CIOs explicitly state that AI usage does not occur “on its own.” Analyst Jamie Kohn provided a particularly illustrative example from the recruiting area: 39 percent of applicants would already use AI for resumes or cover letters. At the same time, only 26 percent trust that employers will use AI fairly to evaluate applications. This creates a paradoxical race: applicants are increasingly automating their documents, while companies are increasingly automating their selection with AI. Yet both sides distrust the outcome.

To make AI manageable, training is considered a solid solution. While Gartner confirms the necessity of training, it is not enough on its own. “We often see a kind of 'two-week syndrome,' meaning new tools are briefly tried out and then people return to their daily routine,” says analyst Alicia Mullery. She refers to the Ebbinghaus curve, i.e., the forgetting curve when learned material is not sufficiently used. According to this, without repetition, 70 percent of newly acquired skills are lost within a day and 90 percent within a week. Instead, differentiated training tailored to user groups is needed: skeptics require different formats than power users or managers. And learning success must be linked to measurable performance indicators. There can be no shortcuts for managers, because according to Gartner, only 8 percent of HR managers believe that their managers have the necessary skills for effective AI deployment. This means: Analysts certainly see the need for AI training – but only if it is differentiated and linked to measurable results.

Training is mandatory: But it's just the beginning

For Gartner, however, another factor is much more important: AI is changing the world of work. Those who only train without reorganizing work are training their employees for jobs that will soon no longer exist. Analyst Daryl Plummer puts it unusually clearly: “Business units that fundamentally transform their work processes, instead of just introducing AI, achieve their revenue targets twice as often.” This also shifts responsibilities. AI competence is no longer purely an IT issue – but not an HR issue either. According to Gartner, it lies precisely in between, and that is another problem, because it leads to responsibility being shirked. The result is “accidental ownership”: the CIO becomes unintentionally responsible for acceptance, productivity, and culture. Gartner advocates for complementary governance between IT, HR, and business unit managers.

Implementing this is a difficult process, but the efforts are worthwhile, because the benefits of such an approach are immense: “Employees who use AI daily are at least 1.5 times more productive than others,” says analyst Harsh Kundulli. The crucial difference lies not in the technology, but in the work organization. “Companies with real work restructuring achieve their revenue targets twice as often as others,” Kundulli summarizes the explanations.

Conclusion

Three insights emerge from the Gartner presentations:

1. AI competence is not a learning objective, but an operational state. Those who “complete” it have already lost.
2. Training without work restructuring is expensive stagnation. While this may be compliance-compliant, it does not make one competitive.
3. Without shared responsibility from IT, HR, and Business, AI will remain a perpetual discussion, because good technology is only one side of the coin. Employees and managers are the other – and they can block everything.

In other words: AI rarely fails today due to the model or computing power. It usually fails due to the organization and the affected people who refuse the new technology.

Interested parties can find Gartner's outlook on AI deployment by 2030 here. The analysts have also taken a closer look at the role of managers and the changing work in companies.

(mack)