Artificial intelligence (AI) is now an essential part of every organization. In fact, it has become a foundational element. Consequently, we are no longer in a phase of exploration or marginal experimentation with AI: we have entered a phase of gradual integration and structuring of its uses, with companies seeking to organize and guide the use of these technologies, even though their adoption varies depending on employees’ profiles and contexts.
But behind this rapid adoption of AI lies a deeper transformation of professions and skills, as AI is no longer content to merely assist humans in their tasks and is increasingly acting autonomously. This evolution of AI is leading us to gradually redefine the line between human tasks and those that can be automated.
A Two-Speed Adoption of AI
Generative artificial intelligence has seen unprecedented adoption in France. In 2023, 20% of the French population reported using generative artificial intelligence (GAI). By 2025, this figure reached nearly half the population (48%), representing a 28-point increase over the space of two years. By comparison, it is estimated that home internet access took three additional years to reach a comparable level of penetration in French households (CREDOC, 2026).
However, this adoption remains deeply unequal, as it is marked by significant polarization based on age and socio-professional category. Young people are the heaviest users: 85% of 18- to 24-year-olds and 73% of 25- to 39-year-olds report using GAI. Similarly, the self-employed (77%), managers and professionals (76%), and college graduates (65%) are among the most frequent users of IAG (CREDOC, 2026). This predominance of certain segments of the population reflects a genuine digital divide linked to educational background and skill levels: the most highly educated individuals are those who are most proficient with these tools and who best leverage their potential.
Moreover, the use of AI remains largely confined to the personal sphere rather than the professional one. In fact, only 35% of working people say they use AI at least once a week for work, compared to 41% in their personal lives. When we look at AI usage in relation to company size, the gap is even wider. While 58% of large companies report using AI, this rate drops to 31% for small and medium-sized enterprises and to 15% for very small businesses (IPSOS, 2026), even though very small businesses and SMEs account for 99% of the French economic tissue (OECD, 2024). These differences in AI usage illustrate the difficulties organizations face in integrating AI tools, especially when they lack the necessary resources to implement them. But it also highlights that the use of AI is often an individual practice rather than a corporate strategy.
As a result, AI is still primarily used for support tasks such as writing, summarizing, or simple research. Nevertheless, a genuine shift in usage is taking place, and AI is increasingly being deployed to perform tasks with greater added value: in-depth research (27%), creativity (26%), technical or specialized tasks (21%), and training (19%). This trend is supported by business leaders, 70% of whom believe that AI has already helped improve their productivity, and 53% of whom view it as a driver of positive impact for their organization (IPSOS, 2026). Thus, while AI tools are not yet fully integrated within organizations, these figures demonstrate a clear intention to deploy them in the coming years.
The Rise of Agentic AI
But a new stage is being reached with the rise of so-called “agent-based” AI, which is characterized by tools capable of acting autonomously. Agent-based AI can be defined as a tool capable of making decisions, planning actions, and performing complex tasks with limited human input. Agent-based AI differs from IAG in that it no longer simply responds to requests; it is capable of taking action, breaking down problems to propose more advanced solutions and responses.
Its operation is mostly based on four “complementary layers”:
- An application layer, which enables interaction with the user and understanding of their needs
- An orchestration layer, which coordinates the various tasks and agents to determine the course of action
- A logic layer, which performs the work and structures the reasoning of the agents used
- An evaluation layer, which analyzes the results and improves the content produced
These layers are called complementary because they work in synergy to enable AI to understand, decide, act, and learn in a nearly autonomous manner.
The potential impact of agent-based AI is significant, as this new form of automation opens up the possibility of performing complex tasks that were previously very difficult to automate. This represents a new opportunity for organizational transformation, leading to a rethinking of employees’ roles and skills. With the advent of agent-based AI, individuals no longer need only interact with the machine but must coordinate a network of autonomous digital agents. The employee becomes the “manager” of the machine—a conductor who supervises, mediates, and directs the actions of these digital agents.
The Evolution of Skills: From Know-How to Know-How to Manage
With agent-based AI, the scope of AI applications is expanding beyond writing and analysis tasks, which means it increases the volume of tasks that can potentially be automated. However, contrary to certain misconceptions, this evolution does not result in massive job losses, but rather in a transformation of job content and a realignment of employees’ skills. This is particularly true for the professions most exposed to the influence of AI: architecture and engineering, IT, finance, administration, law, education, and the arts, as well as support functions—for example, sectors in which more than 30% of tasks are identified as automatable (COFACE, 2026). These professions, which rely heavily on the processing and use of information, are particularly affected by the automation of their tasks. They are also professions with a high proportion of women, particularly in legal, administrative, and support positions.
In this context, employees are involved less in the pure production phases but increasingly earlier in the process to define needs and structure the requests made to AI tools, and later in the process to evaluate the quality of the output produced by AI and validate its results. We are thus witnessing the development of expertise in managing AI tools, which complements employees’ existing skills.
Consequently, we are seeing a significant rise in the importance of cross-functional skills, “soft skills,” and particularly critical thinking. Employees must be able to detect errors made by AI, question the consistency of its reasoning, and identify biases and blind spots in the tasks performed by AI. The ability to step back, interpret results, and make judgments based on them is becoming a key skill within organizations in the era of AI development.
The need to develop these types of skills highlights the importance of actively including senior employees in organizations. Companies may sometimes mistakenly assume that senior employees face the greatest challenges when it comes to adopting AI. However, the skills that are now considered essential—such as critical thinking, decision-making, and the ability to assess situations—rely heavily on experience and an understanding of sensitive situations. Senior employees may therefore be best positioned to assess the quality and relevance of AI-generated results and identify their limitations. In addition, the OECD estimates that the inclusion of older workers in the workforce alone could lead to a 19% increase in GDP per capita across its member countries by 2050.
Thus, these developments are part of employees’ development of meta-skills, sometimes referred to as “meta-AI” skills. It is no longer a matter of understanding how a tool works and mastering its use; one must now understand its limitations and be able to interact effectively with it, particularly through effective prompts capable of anticipating the biases that AI might exhibit without human reasoning. Here, the employee becomes the orchestrator of production.
KYU supports you in your AI initiatives !
While artificial intelligence represents a major opportunity for organizations, its implementation implies transformations that go beyond the mere use of AI tools: evolving skill sets, redefining job roles, and adapting work organization models.
It also has a broader impact on corporate life itself and therefore raises issues related to social dialogue, particularly regarding its integration into corporate initiatives for Job and Career Path Management (GEPP) and the improvement of Quality of Life and Working Conditions (QVCT). Our webinar on GEPP in the age of AI is also available at the following link: https://webikeo.fr/webinar/la-gepp-a-l-ere-de-l-ia-planifier-ou-s-adapter#webinar
KYU supports you in evaluating and analyzing your practices to structure your AI strategy and guide you through the evolution of skills and job roles.
References :
CIANum (2026), IA agentique : enjeux et perspectives : https://www.conseil-ia-numerique.fr/files/uploads/2026/IA%20agentique%20CIANum_all.pdf
COFACE (2026), Emplois, compétences, valeur : ce que l’IA est en train de bouleverser : https://www.coface.fr/actualites-economie-conseils/emplois-competences-valeur-ce-que-l-ia-est-en-train-de-bouleverser
CREDOC (2026), Baromètre du numérique 2026 : https://www.credoc.fr/publications/barometre-du-numerique-2026-rapport
IPSOS BVA (2026), IA en entreprise : état des lieux et leviers d’accélération : https://www.ipsos.com/fr-fr/ia-en-entreprise-etat-des-lieux-et-leviers-dacceleration
Les Echos, Compétences des séniors : un atout face aux enjeux de l’IA : https://www.lesechos.fr/idees-debats/leadership-management/competences-des-seniors-un-atout-face-aux-enjeux-de-lia-2204698
Levin, I., Marom, M., & Kojukhov, A. (2025). Rethinking AI in Education: Highlighting the Metacognitive Challenge. BRAIN. Broad Research in Artificial Intelligence and Neuroscience, p.250-263 : http://dx.doi.org/10.70594/brain/16.S1/21
OCDE (2023), Perspectives de l’emploi de l’OCDE 2023 : https://www.oecd.org/fr/publications/perspectives-de-l-emploi-de-l-ocde-2023_aae5dba0-fr.html
Vie Publique (2026), IA agentique : une technologie qui suscite des questions : https://www.vie-publique.fr/en-bref/302417-ia-agentique-une-technologie-qui-suscite-des-questions
