When AI expands the job before it replaces it
AI lets workers perform tasks historically associated with other professions. This apparent autonomy may also shift workload, responsibility and verification work.
RESEARCH
Notes, research and observations on artificial intelligence, work organisation and occupational health.
How should we make AI-driven work safe for humans?
01
Selected work connecting technical capability, work organisation and prevention.
AI lets workers perform tasks historically associated with other professions. This apparent autonomy may also shift workload, responsibility and verification work.
When a system assigns tasks, measures activity or recommends a decision, it changes authority without necessarily making responsibility clearer.
A model can be technically reliable and still make work more intense, less autonomous or more uncertain. Safety must include the effects of the deployed system on people.
Legal classification creates obligations. It does not replace analysis of real work, discussion of use or monitoring after deployment.
02 / ALL RESEARCH
An evolving index of scientific publications, notes, field analysis and governance perspectives.
9 publications shown
AI lets workers perform tasks historically associated with other professions. This apparent autonomy may also shift workload, responsibility and verification work.
When a system assigns tasks, measures activity or recommends a decision, it changes authority without necessarily making responsibility clearer.
A model can be technically reliable and still make work more intense, less autonomous or more uncertain. Safety must include the effects of the deployed system on people.
Legal classification creates obligations. It does not replace analysis of real work, discussion of use or monitoring after deployment.
Transformation begins with meaning, recognition, autonomy and the value of expertise — long before a job disappears.
Checking, reframing, integrating, reporting and recovering from error: automation removes some operations while creating others that often go uncounted.
Providers, management, business teams, IT, worker representatives and prevention specialists each see part of the system. Safety depends on explicit coordination.
Only 7.8% of managers in the study correctly anticipate differing effects across tasks. Heterogeneous performance can then become a homogeneous expectation.
France’s 2026–2030 occupational-health plan places AI-related transformations within prevention and calls for a national observatory on AI at work.