In France, the professional use of generative artificial intelligence is progressing faster than the training that accompanies it. The Cegos barometer highlights a clear gap between the tools adopted by employees and the actual level of mastery within organizations. At the same time, the European AI Act introduces a competence requirement in its Article 4 for individuals using AI systems. Training teams is no longer just a lever for competitiveness; it is also an emerging regulatory constraint.
Legal obligation for training: what the Labor Code really imposes
Article L.6321-1 of the Labor Code requires employers to ensure the adaptation of employees to their positions and to maintain their ability to hold a job. This obligation is not limited to onboarding training: it covers the entire duration of the contract.
Failure to comply with this obligation exposes the company to sanctions. In the event of dismissal, an employee who has not received any training for several years may obtain damages before the labor court.
A law firm specializing in labor law reminds us that the omission of the professional interview opens a recourse for the employee, including to contest a refusal for training. Specialized resources, such as those offered on academie-entreprise.com, help identify pathways that meet the obligations of each organization.
AI training in companies: France’s lag behind actual usage
The Cegos barometer cited by Le Nouvel Économiste points out a paradox. French employees use generative AI in their daily professional lives, but few organizations have formalized rules for using AI. When training exists, it often remains a general awareness disconnected from concrete tasks.
Specific training for functions (marketing, HR, sales) is becoming more relevant compared to cross-functional modules. A salesperson using an automatic writing tool does not have the same needs as an HR manager filtering algorithm-assisted applications.
Article 4 of the AI Act and its implications for training
The European AI Act stipulates that providers and deployers of AI systems must take measures to ensure a sufficient level of mastery among users. The provision takes into account prior knowledge, the context of use, and associated risks.
In practice, the AI Act does not prescribe a unique program or a specific number of hours. Each company must therefore define the expected level of competence based on the tools deployed. For organizations that have not yet structured their internal training offerings, this regulatory ambiguity creates a double challenge: identifying needs and documenting compliance.
Skills diagnosis before training plan: method and limitations
Competitors almost systematically recommend “defining clear objectives” and “setting SMART goals.” This approach assumes that the company already knows the skills present in its teams. In reality, many SMEs launch training without prior diagnosis.
An effective skills diagnosis relies on three elements:
- An updated job reference framework that describes the expected technical and behavioral skills for each position, including recent digital skills
- A cross-reference between this framework and individual assessments from professional interviews to identify gaps between the required level and the actual level
- A mapping of training already completed to avoid duplicates and prioritize critical gaps for the business
The main limitation lies in the reliability of the data. Professional interviews, when conducted administratively, produce poor information. A diagnosis based on rushed interviews leads to poorly targeted training, which fuels skepticism among teams regarding the training itself.
Most requested training in companies: verified trends
The rise of AI-related training should not overshadow another phenomenon: soft skills remain among the most requested competencies by employers. Cegos emphasizes that AI and human skills do not oppose each other but complement each other, with organizations investing in both areas simultaneously achieving better results in terms of tool adoption.
Internal training or external organization: decide based on need
Internal training (mentoring, co-development, peer sessions) is cheaper and better suited to the company’s specificities. It works well for transmitting job skills or internal processes.
External training becomes necessary when the subject requires expertise absent internally, or when regulatory issues impose certification. This is the case for safety authorizations, quality certifications, or compliance modules related to the AI Act.
- Internal training: suitable for specific job skills, sharing best practices, and integrating new employees
- External certified training: necessary for regulatory obligations, advanced technical skill development, or obtaining recognized labels
- Blended format: combines e-learning for theory and in-person sessions for practical application, reducing time away from the job
The choice between these formats depends less on the budget than on the nature of the need. A poorly chosen training costs more than a well-targeted training, even if the latter has a higher price tag.
The CPF remains a mobilizable funding tool, but recent discussions around its reform introduce uncertainty. Companies that anticipate a potential transfer of responsibility by structuring their skills development plan are better positioned than those relying on the individual mechanisms of their employees.
