Artificial intelligence is no longer content to assist: it is preparing to decide. With the advent of “agentic” AI, an unprecedented shock wave threatens to shake up the foundations of our economy. Unlike past industrial revolutions, it is no longer the arms that the machine replaces, but the brains.
An alarming study by Coface and the Observatory of Threatened Jobs reveals that 16% of French workers, mainly executives and intellectual professions, are now on the front line of this historic change. Deciphering a social earthquake which could well mark the end of the golden age of white-collar workers.
The Agentic era: when AI learns to decide
To understand the extent of the phenomenon, we must forget ChatGPT and its approximate poems. We are entering the phase of agentic AI.
What is an AI “agent”? It is a system capable of making autonomous decisions, learning from its interactions and correcting its own reasoning without human intervention. If today’s generative AI is an assistant who writes, agentic AI is a collaborator who manages a project from A to Z.
It is this technological leap that endangers 16.3% of private and public jobs in France. For comparison, France is doing better than the United Kingdom or the Netherlands, but the macroeconomic impact remains dizzying.
“White-collar workers” in the crosshairs
This is the paradox of this revolution: the more education you have, the higher your salary, and the more exposed you are. The study identifies a critical threshold: as soon as 30% of the tasks in a job can be automated, the job is considered “in danger”.
The most vulnerable sectors form the backbone of the French tertiary economy:
- Legal and accounting activities: Formalized expertise is the ideal prey for algorithms.
- Publishing and Press: Information processing and writing are no longer human sanctuaries.
- Programming and IT consulting: Paradoxically, those who create AI are also the first to see their tasks codified.
- Banking and Insurance: Risk analysis and massive data management are already the playground of machines.
In short, everything cognitive, digital and expert can be fragmented into algorithms.
The revenge of “doing” on “thinking”
While senior executives worry, another world seems to breathe. Manual professions, those that require physical interaction with the real world or immediate empathy, are the big survivors.
- The mason, the plumber, the nurse or the cook remain, for the moment, irreplaceable.
- The tourism, hotel and catering sector appears to be a human sanctuary.
This shift could reshape the social hierarchy. For decades, the degree was the ultimate life insurance. Tomorrow, knowing how to manipulate abstract concepts could be worth less than knowing how to repair a leak or manage the unexpected on a construction site.
Comparison of exposure by type of employment
| Type of employment | Risk level | Vulnerability factor |
| Financial framework | Very high | Repetitive cognitive tasks, data |
| Lawyer / Jurist | Pupil | Text analysis, case law |
| R&D engineer | Medium/High | Simulation automation |
| Building craftsman | Weak | Physical complexity, variable environment |
| Nursing staff | Very weak | Empathy, complex technical gestures |
The end of triumphant metropolises?
The other earthquake is geographical. For thirty years, globalization has benefited large metropolises, concentrating wealth and “talent”. AI could reverse the trend, or rather, standardize from below.
- In Paris, nearly one job in five could disappear.
- In Lyon or Toulouse, we are close to 18%.
- Conversely, towns like Avallon (13%) or Briançon (12.3%) are much less exposed, because their economic fabric is based on local commerce, crafts and tourism.
We could see a reduction in territorial inequalities, not through the rise of rural areas, but through the erosion of the dynamism of urban centers.
Destruction… without creation?
The great hope of the optimists rests on Schumpeter’s “creative destruction”. The train killed the stagecoaches but created station masters. The problem is that AI is not a tool like any other.
If the machine replaces the human for existing tasks AND for future tasks, the social model of the welfare state falters. In continental Europe, where pensions and health are financed by labor contributions, the disappearance of high salaries (the largest contributors) in favor of capital risks creating a budgetary chasm.
The challenges for the State:
- Fall in tax revenue linked to the disappearance of white-collar jobs.
- Explosion of training needs to retrain millions of workers.
- Value sharing: Profit risks being massively concentrated among the owners of the technology (capital) rather than among those who operate it (labor).
A new social contract?
We are not facing a simple technological evolution, but a redefinition of what constitutes the value of a human being in the economy. If knowledge and expertise are no longer enough to guarantee employment, we will have to rethink taxation – perhaps that of capital or robots – to maintain the cohesion of an aging society.
AI is a black box, a mystery that even its creators sometimes struggle to decipher. But one thing is certain: the office of tomorrow will look nothing like the office of today.
