Generative AI: Illusion of Knowledge or Tool for Cognitive Atrophy?

The Illusion of Mastery Through Generative AI

The real risk of generative AIs does not lie in their ability to replace us, but rather in the possibility that they may encourage us to give up on learning. Technologies such as ChatGPT, Copilot, Gemini, and Claude have become ubiquitous in tasks like accelerated writing, document synthesis, code generation, and automated customer responses. They promise significant productivity gains, but at what cost? This ease of use can create an illusion of knowledge, where the user achieves results without understanding the underlying reasoning. This situation raises concerns about how these tools influence our ability to learn and develop sustainable skills.

Two approaches to AI are at odds: on one hand, generative AI that produces on our behalf, and on the other, pedagogical AI that aims to structure learning and enhance long-term skills. This dichotomy raises strategic questions about the type of professionals that AI will shape in the future. What kind of human will AI shape in the coming years? A constantly assisted collaborator, or a more autonomous and insightful professional?

The Dangers of Cognitive Atrophy

Nicolas Bourgerie, CEO of Teach Up, warns against the risks of excessive dependence on generative AI, particularly in areas where human judgment is crucial. Sectors such as complex customer relations, management, regulatory compliance, high-value sales, and strategic decision-making require expertise that AI cannot replace without consequences. “Delegating these tasks to generative AI without having previously strengthened one’s own mastery exposes one to costly errors, legal risks, or a rapid loss of credibility,” emphasizes Nicolas.

Regulated sectors like banking, insurance, and healthcare are particularly vulnerable to these risks. This also includes the automotive industry for both commercial and technical aspects. In environments like banking and the automotive industry, the spontaneous use of generative AI has shown trends toward standardizing business responses and weakening personalized arguments. Training programs based on client simulations have been implemented to enhance team skills.

Teams were required to structure their responses themselves, justify their choices, and manage objections without resorting to the tool in real-time. “This led to a better mastery of offerings, a greater ability to handle complex objections without assistance, and a more differentiated discourse. In some cases, managers noted a measurable increase in the quality of exchanges and in employee confidence when faced with the unexpected,” shares Nicolas. In these B2B environments, value rests on the expertise and trust of the teams. Excessive standardization of responses or dependence on the tool can weaken the ability to handle atypical cases.

The Pedagogical Approach to AI

Pedagogical AI is distinguished by its ability to place the learner in an active situation. Rather than providing immediate solutions, it poses open-ended questions, requiring a reasoned decision before delivering any feedback. The model then analyzes the response to detect conceptual gaps, weak reasoning, or blind spots. If the learner masters a point, the level increases.

“Instead of giving the solution too quickly, it introduces questions and reformulations that compel the learner to clarify their reasoning. This approach fosters a gradual consolidation of skills, rather than a superficial performance assisted by the tool,” adds Nicolas. Unlike a superficial performance achieved through constant assistance, the learner develops a mastery that withstands unforeseen situations. The tool does not mask weaknesses but exposes them to better address them.

This logic reverses the usual dynamic of generative AI. Instead of accelerating the task at the expense of learning, pedagogical AI deliberately slows down to enhance deep understanding. “On a larger scale, the strategic issue is choosing between an AI that produces in place of humans and one that enhances their capacity to learn. Measuring only speed and productivity ignores the risk of cognitive atrophy. In the long run, the differentiation of organizations will not come from those that automate the most, but from those that sustainably maintain and develop the mastery level of their teams,” concludes the CEO of Teach Up.