The professional training market is going through a period of contradictory tensions. On one hand, learning technologies are advancing rapidly. On the other hand, individual funding mechanisms are losing ground, with measurable consequences for the least qualified audiences.
Decline of the CPF and social divide in access to training
The use of the CPF has decreased by 11% in 2025 compared to 2024, with 1.226 million training sessions started according to Dares. This decline is a structuring fact for the entire sector.
This decline is not uniform. The share of beneficiaries with a qualification level of high school diploma or lower has dropped from 62% in 2023 to 58% in 2025. The share of non-executives has decreased from 83% to 79%. A senatorial report interprets these figures as a socially regressive effect of the braking measures, particularly the introduction of out-of-pocket expenses.
The use of the CPF by active employees has fallen by 26% in the year following this measure, while it is increasing among job seekers, who are exempt from it. This gap raises a direct question: do the CPF regulation policies, designed to limit abuses, create a selection by income in access to training? Documented resources on pratiques-de-la-formation.fr allow tracking the evolution of these mechanisms and their concrete effects on the ground.
Workplace learning: the end of a growth cycle
Apprenticeship had experienced spectacular expansion in recent years, driven by generous hiring aids. This cycle is reversing. The number of new apprenticeship contracts has significantly decreased in 2025 following the reduction of employer subsidies.
The decline primarily affects training at the high school level and below. Contracts in higher education are holding up better, which exacerbates an already documented imbalance: apprenticeship benefits higher education graduates more than the audiences it initially targeted.
For training organizations, this budget contraction alters decision-making. Overly extensive catalogs become difficult to maintain when the number of trainees decreases. The pressure on costs pushes towards shorter and more targeted formats, but not necessarily more effective for audiences that need long-term support.
Competency-based training: promises and blind spots
The competency-based approach (often referred to as “skills-based learning”) is becoming established in the vocabulary of HR departments. The principle: identify the skills applicable in work situations rather than relying solely on diplomas.
Competency frameworks are becoming dynamic, connected to the evolution of jobs. Companies are looking to map the existing skills within their teams to anticipate future needs. On paper, the approach is coherent.
In practice, field feedback diverges on this point. Several limitations appear:
- The granularity of the frameworks varies greatly from one sector to another, making comparisons difficult between companies or professional branches
- The tools for assessing skills often remain declarative, without validation in real situations, which weakens the reliability of management
- The “skills-based” logic can marginalize transversal knowledge (general culture, critical thinking, systemic understanding) in favor of short-lived technical skills
A competency framework has value only if it is regularly updated and backed by reliable evaluations. Without this, the risk is to replace one formality (the diploma) with another (the checked list of skills).
Behavioral skills and measurement limits
Behavioral skills (communication, stress management, collaboration) are gaining visibility in training plans. Their evaluation remains an open issue. The available data do not allow conclusions about the long-term effectiveness of short training sessions dedicated to these topics.
Artificial intelligence and personalization of training paths
Generative AI is changing the design of educational content. Platforms now offer to generate exercises, case studies, or adaptive paths based on the learner’s profile. The promise: each employee receives a path tailored to their level and objectives.
The gains are real in content production. A trainer can create variations of exercises in minutes that would have taken hours to write. Educational chatbots allow learners to ask questions outside sessions with a human tutor.
However, algorithmic personalization raises questions that the sector has not yet resolved:
- The biases of AI models can steer training recommendations towards conservative choices, reproducing majority paths rather than opening atypical trajectories
- Dependence on proprietary tools creates a risk of technological lock-in for small training organizations
- The evaluation of the pedagogical quality of AI-generated content remains largely manual, limiting real productivity gains
AI does not replace instructional design. It accelerates certain tasks, but the framing of objectives and human support remain crucial for the effective skill development of learners.
Short formats and on-the-job learning
Microlearning (modules of a few minutes) and on-the-job learning are becoming widespread. These formats address a real constraint: employees have little time dedicated to training.
Companies that integrate learning sequences directly into work tools (contextual tutorials, targeted notifications) observe completion rates higher than traditional classroom training. The short format works well for technical updates or regulatory reminders.
For complex learning (retraining, acquiring a new job), short formats alone are not sufficient. They complement a structured path but do not replace it. The temptation to break everything down into micro-modules to reduce costs can lead to fragmented learning, lacking overall coherence.
The professional training sector finds itself caught between two movements: a technological innovation that expands the realm of possibilities, and a budget tightening that restricts actual access. New methods have significance only if the audiences that need them most can access them. The financial accessibility of these mechanisms remains the unresolved issue of this equation.