AI as the New Literacy: Rejoining the Battle for Light

By Roxan Roumégas

We are debating the wrong thing. Reducing artificial intelligence to yet another innovation is to miss the essential point. The issue is not innovation. The issue is progress — what truly raises the level of freedom, equality and dignity in a society.

AI touches a vital nerve: language. In an economy where almost everything passes through writing (procedures, forms, standards, platforms, instructions, reporting), language has become a frontier. AI can remove this frontier, or raise it higher. There will be no neutrality. There will be a direction.

Making the World Readable for Those It Excludes

France is not a country of illiterates. But it is a country where a massive share of adults struggles with basic skills. ANLCI estimates that 10.5% of 18-64-year-olds schooled in France are in severe difficulty with literacy and numeracy, including 4% in a situation of illiteracy.

The paradox is cruel: one can be excellent in a professional gesture, meticulous, reliable, and yet be put out of the game by the wall of writing. In many jobs, it is not only technical skills that count, but the capacity to read, interpret and apply sometimes complex, sometimes standardized instructions.

AI is not a robot that replaces. It can become a universal translator of the working world: automatic reading and clear reformulation of a notice, oral step-by-step guidance, identification of vigilance points, help in understanding pictograms and safety rules. What school did historically with literacy, AI can do with a new layer: practical access to language. Not by lowering the requirement, but by making the rule understandable. And when the rule becomes understandable, real skill can finally be recognized.

Skill Is No Longer Enough: One Must Know How to Read Oneself

The working world does not only suffer from a lack of candidates. It suffers from poor matching, tensions, rapid ruptures. France Travail noted that 57.4% of recruitment projects were judged difficult in 2024.

Behind these figures, a very concrete reality: individuals are asked not only to work, but to describe themselves, to justify themselves, to fit into grids, sorting algorithms, forms, structured interviews. Those who master these codes (often socially) appear better than they are. Those who do not master them (sometimes very competent) remain invisible.

A well-oriented AI can rebalance this point: help formulate know-how, prepare an interview, make job offers readable, explain expectations. It is less spectacular than speeches on disruption, but it is exactly there that equality of opportunity is created, when codes cease to be a toll.

Building an Infrastructure of Trust

Training is not a product like any other. It engages life time, public money, reconversions, sometimes repair trajectories. Recent reforms are not perfect, but they respond to a rational objective: making quality verifiable.

French law has set clear criteria, including one often underestimated: public information on services, access times and results obtained. The national quality framework and Qualiopi certification fall within this scope.

One can criticize bureaucracy. One must even criticize it when it becomes absurd. But one cannot forget the essential: without a trust standard, training becomes a market of promises. And in a market of promises, it is always the most fragile who lose.

The future is therefore not fewer rules or more rules. The future is intelligent rules, measurable, oriented toward real impact: integration, progression, satisfaction, transparency, readability of pathways.

School Becomes a Factory of Critical Spirit Again

AI changes a fundamental educational given: access to knowledge becomes more immediate, more customizable, more conversational. This does not make school useless. It changes its mission.

The French State explicitly affirms the development of critical thinking as an objective of instruction, and the educational institution equips this approach. The OECD, for its part, is already working on measuring media literacy and AI literacy via PISA 2029, a sign that the capacity to navigate information is becoming a structuring skill.

If knowledge becomes available, value shifts toward the capacity to verify, hierarchize, reason, debate, construct. School is no longer an encyclopedia. It becomes an agora again: not a factory for reciting, but a place where one learns to think rightly.

And this touches on a often misunderstood point: in primary school, the teacher cannot give written homework to be done outside class. Homework is oral (reading, lessons, research). This reminder is not a quirk. It states a profound idea of equality: when schoolwork depends too much on family context, school ceases to be an elevator and becomes an amplifier of gaps.

AI Can Be a Safety Net, Not an Alibi

Inclusive schooling is progressing, but it remains under tension. The Cour des comptes highlights the path traveled and persistent difficulties, and proposes improvement axes: steering, monitoring, evaluation, quality of schooling.

AI can play a useful role: reading assistance, reformulation, multimodal supports, help with organization, guided repetition, adaptation of language level. But on one condition: not transforming the tool into an excuse for not funding the human.

The good principle is simple: AI to increase autonomy when relevant. The human for relationship, fine adjustment, ethics, care and protection.

Innovation Is Not Progress

The two are too often confused. Social networks were a major innovation. Their impact on youth mental health and on the public space is the subject of alerts at the highest sanitary level in the United States. And on the terrain of truth, AI can also accelerate disinformation. UNESCO has warned about the risks of misleading content related to the Holocaust, precisely because generative tools can produce false narratives on a large scale.

This point is central: technology can illuminate or obscure. It does not naturally tend toward good. It follows economic and political incentives. If we want progress, governance is required.

Education Cannot Depend on an Opaque Market

Europe has equipped itself with a legal framework with the AI Act. But educational sovereignty is not summed up in a law: it implies infrastructure choices. If we entrust access to knowledge, exercises, assistance to closed tools, non-auditable, dependent on changing terms of use, then we outsource a regalian function: forming citizens.

The ambition to aim for is higher: pedagogical commons supported by the general interest, reference content, evaluated assistance tools, explicit rules of use, teacher training and transparency on data. UNESCO explicitly speaks of a human-centered vision and a capacity to plan long-term public policies on AI in education.

What We Stand For

At Mentivis, we defend a simple, demanding, humanist line.

AI must serve first those who are hindered, those for whom language, codes, writing, guidance or bureaucracy are walls.

• Quality must be provable: transparency of results, readability of pathways, honesty about career prospects.
• School must form critical minds, capable of investigating, verifying, discussing and resisting manipulation.
• Inclusion must remain a human project. The tool is a lever, not a justification for renunciation.
• Progress requires governance. Without it, innovation becomes a machine for inequalities.
• We are entering a century where knowledge can be everywhere. The question is no longer who has access to information. The question becomes: who has access to understanding, to method, to verifiable truth and to the capacity to act?

That is the battle for light. And that is where training, when it remains faithful to its vocation, is inevitably humanist: it does not only add skills, it adds power to be.

Roxan Roumégas is partner and president at Mentivis. He has directed technical schools and for more than fifteen years has accompanied educational institutions and public institutions in the design of pedagogical strategies, training engineering and educational transformation. He operates at the interface between academic vision, operational constraints and concrete implementation on the ground.

What Mentivis Actually Does

Mentivis is not a firm that applies off-the-shelf solutions.

It is an educational transformation operator that acts on the real mechanisms of change: behaviors, incentives, implicit rules.