Great that you, and the French government, are looking at this problem and starting to take it seriously, but there are several flaws in your analysis that together lead you to far more optimistic conclusions than are in fact likely.
The biggest flaw is this statement:
If more than 2/3 of tasks are automated, then the job will disappear. If less than 1/3 are automated, then nothing will change. In all other cases, the job will be transformed.
You leave out the economic drive for efficiency. If 50% of tasks for a given job are automated, then half as many people will be needed to do the same amount of work… and that is what will sooner or later happen. As a result, your estimate of job losses is probably wrong by a factor of four… 40% of jobs lost, not 10%… which is what other studies (Oxford University, McKinsey) have concluded. And that in turn means that your assertion that countries with strong welfare systems will be able to handle the resulting unemployment is also wrong.
The second big flaw concerns your assumptions around retraining. To begin with, what will those who lose their jobs be retrained to do? Where are the new jobs for these people going to come from? Every domain will be impacted by AI… including any new domains that come into existence. It won’t matter if is easier to retrain people if there is nothing to retrain them for. Further, if the remaining workers are expected to be able to supervise the AI, then they are going to need to be just as skilled as they are now. If not, then they will be easy to replace… and salaries for people who do easy-to-learn jobs and who are easy to replace are always low… so those who retain their jobs are also going to be profoundly and negatively affected.
And this leads to the third big flaw: you ignore the social and democratic side-effects of the destruction of large numbers of jobs. The experience of losing your job is always bad, but it is much worse when your job is destroyed… i.e., when there is next to no hope of finding a comparable job again because the pool of such jobs is shrinking. When someone’s job is destroyed, they lose much more than just an ongoing income stream. They also lose their self-respect, status, confidence in the future, future earning potential, and so on. The result: they will be angry, and will vote for unscrupulous populists on the left or the right who promise to make everything good again… just as the ex-manufacturing workers of the Midwest did when they voted for Donald Trump. It is not hyperbole to say that the continued existence of liberal democracy is dependent upon how well the transition to AI is handled. I have written about this at some length here: The AI Transition, Job Destruction, and Democracy.