AI especially threatens intermediate jobs. In France, the high cost of labor is accelerating the replacement of certain positions with cheaper AI tools.
We are told repeatedly that artificial intelligence primarily threatens interns. Let her nibble at the menial tasks, the repetitive missions, the bottom of the ladder. This is false. The real target of AI is intermediate jobs. And the higher the labor cost, the more attractive this target becomes. The economic reasoning is implacable. In France, an executive with four to five years of experience costs his employer around 10,000 euros per month, including charges. Employer social contributions represent 44.3% of gross payroll according to Ayming. Faced with this amount, a ChatGPT Team subscription at $25 per month seems insignificant. The calculation quickly becomes obvious: can an intern equipped with AI accomplish what an experienced junior did yesterday? Companies ask themselves this question every day. And the answer, more and more often, is yes.
The Indian paradox
India offers an illuminating counter-example. OpenAI’s largest market outside the United States, the country has 850 million Internet users and shows growth in ChatGPT downloads of 587% in one year. However, Indian companies are hesitant to invest in AI for their employees. The reason? An average monthly salary of 21,000 rupees, or around 230 euros. When labor costs so little, why spend to make it more productive? OpenAI understood this well: in August 2025, the company launched ChatGPT Go in India at 399 rupees per month, or less than $5, compared to $20 for the standard Plus subscription. The Indian market is adopting AI, but on an individual basis, not yet as a lever for transforming organizations. This paradox reveals an economic truth that our leaders refuse to see: AI does not destroy jobs uniformly. It hits where the return on investment is maximum, that is to say where work is expensive.
France on the front line
The data confirms this intuition. According to Goldman Sachs, countries with advanced IT infrastructure and high labor costs, such as Germany, Japan and the United States, are experiencing the most aggressive job displacement linked to automation. The United Kingdom saw 11% of its full-time jobs affected by AI in 2024. In the United States, 1.9 million positions were affected at the start of 2025. France has multiple vulnerabilities. An hourly labor cost of 38.4 euros, among the highest in the European Union. Social charges that increase each hiring. And, bitter icing on the cake, declining skills. The PISA 2024 ranking places France 23rd in mathematics, 28th in reading, 26th in science. Since 2018, the country has lost 21 points in mathematics, the largest drop on record. We are training an expensive workforce whose skills are eroding. This is the perfect breeding ground for massive AI substitution.
The political blind spot
Our political class raves about the quality of the French AI ecosystem. We celebrate Mistral, we praise investments in language models, we inaugurate data centers. But this is not where the battle is being fought. The real issue is not whether France produces good AI models. It is how our companies will use this technology. And the answer, in a country where labor is so expensive, is predictable: if we do nothing and wait, they will use it to replace, not to augment.
Time is running out
Each day that passes without structural reform of the labor market worsens the diagnosis. French intermediate jobs, those with three to seven years of experience, are in the sights. Analysts, junior lawyers, developers, consultants: all these profiles whose cost far exceeds the value perceived
the employer facing technological alternatives. A sudden reversal is possible. If companies decide massively in favor of AI to the detriment of employment, the impact on consumption and growth will be violent. The World Economic Forum estimates that 85 million jobs could be displaced globally by AI and automation by the end of 2025. France, with its toxic cocktail of high cost and declining skills, risks suffering a disproportionate share of this shock. The signals are there, the economic mechanisms are known, the alternatives exist. Other countries have chosen more flexible labor markets,
less overwhelming loads, more demanding training. They will absorb the AI revolution differently, perhaps even create jobs where we destroy none. The question is no longer whether AI will transform French employment. It is whether we will have the courage to adapt our structures before this transformation becomes destruction.
