Tech • IA • Crypto
Will AI (really) replace us all?
7/10TL;DR
Artificial intelligence is set to transform millions of tasks rather than eliminate entire jobs, with significant but uneven impacts across sectors and workers.
KEY POINTS
300 Million Tasks, Not Jobs
The widely cited figure from Goldman Sachs refers to up to 300 million tasks that could be automated, not full job losses. The distinction is critical: most roles consist of multiple tasks, only some of which are automatable. This reframes the debate from mass unemployment to structural job transformation.
Measured Exposure Varies Widely
According to the OECD, about 27% of jobs in developed economies are highly exposed to automation, meaning over 70% of their tasks could be automated. However, only 14% of jobs are considered fully automatable with current technology, including data entry and routine administrative roles.
Net Job Creation المتوقع
The World Economic Forum projects 92 million jobs lost but 170 million created globally by 2030, resulting in a net gain. This shift reflects historical patterns where technological disruption destroys some roles while generating new industries and occupations.
Tasks Over Roles
Analysis from McKinsey indicates that 60% of jobs have at least 30% of tasks automatable, but this does not imply job elimination. Instead, workers are expected to spend less time on repetitive work and more on decision-making, creativity, and interpersonal tasks.
France: High Exposure, High Opportunity
In France, about 16.3% of workers—roughly 5 million people—face significant automation exposure. At the same time, AI-related roles command salaries 56% higher than comparable positions without such skills, highlighting a growing wage premium for AI competence.
White-Collar Jobs Most Affected
Unlike past industrial revolutions, AI disproportionately impacts white-collar professions, especially those involving repetitive cognitive tasks. Roles in programming and customer service are among the most exposed, while manual and relational jobs remain relatively protected.
Observed vs Theoretical Automation
A 2026 study by Anthropic shows a gap between what AI could automate and what it actually does. While 94% of tasks in computing could theoretically be automated, only 33% are currently automated in practice, underscoring slow real-world adoption.
Young Workers Hit First
Hiring for workers aged 22 to 25 in highly exposed fields has dropped by 14% since 2024. Companies increasingly rely on AI tools for entry-level tasks, reducing opportunities for junior employees to gain experience.
Resilient Professions
Around 30% of jobs show little to no exposure to automation. These include roles requiring physical skill, human interaction, or adaptability, such as healthcare workers, teachers, tradespeople, and service staff.
Limits of Current Forecasts
Most studies measure technical feasibility, not real-world adoption. Barriers such as cost, regulation, and organizational resistance slow deployment. Adoption rates remain relatively low despite rapid technological progress.
Hidden Impact: Skill Obsolescence
An estimated 40% of skills could become obsolete within five years. Rather than sudden layoffs, many workers may face gradual wage stagnation and reduced demand as AI absorbs parts of their responsibilities.
CONCLUSION
Artificial intelligence is unlikely to eliminate work altogether but will rapidly reshape how work is done, raising urgent economic and social questions about who benefits from productivity gains and who bears the cost of transition.