AI researcher Luc Julia argues that generative artificial intelligence will transform occupations and tasks instead of erasing entire professions. He frames this change through Joseph Schumpeter’s notion of “creative destruction”: new products, services and production methods disrupt existing industries, causing some roles to disappear while new ones emerge. Those new roles typically require different skills, which can produce temporary unemployment and retraining challenges.
A historical parallel: spreadsheets and accountants
Julia often cites spreadsheets as an example: computers and spreadsheet software did not eliminate the accounting profession; they automated repetitive, low‑value tasks (manual calculations and checks) and freed accountants to focus on advisory and analytical work. The tools increased productivity without replacing the core professional activity.
Complementarity between AI and people
He warns against the misconception that AI is in direct competition with humans. People and AI excel at different kinds of work and are complementary. The value produced by many generative systems depends on the prompt or instruction—the system is ineffective without it. Even in applications where AI automates significant parts of a task (for example, copilots in programming), human oversight is still necessary because AI can err or hallucinate.
Two workplace levels of AI use
Julia distinguishes two levels of AI usage in the workplace:
-
The first level covers workers who use and oversee AI tools. These employees must learn to interact with the system, craft prompts and validate outputs—analogous to learning to operate new industrial machinery. With AI handling operational execution, humans focus on prompting and supervision.
-
The second level concerns the “developers” or specialists who configure, fine‑tune, maintain and repair AI systems. These roles require technical and training skills: they will also teach first‑level users how to use AI effectively.
In Schumpeterian terms, some roles within professions will be replaced or transformed (first category), while new occupations will appear to create, operate and maintain AI tools (second category).
Implications for companies and the need for training
Some firms are already using algorithms to organize employee schedules and pace of work, but Julia says this is not an inevitable path for all companies. AI can raise the performance of employees who previously were not outstanding, acting as an extended intelligence that unlocks capabilities. At the same time, progress requires active learning and training: workers who fail to learn the tools risk being left behind.
Training should focus on how to extract the most value from AI; this is primarily an individual responsibility, supported by organizational measures.
Empirical findings from the French generative AI committee
In September 2023 the Élysée Palace invited Luc Julia and other experts to participate in France’s generative AI committee, co‑chaired by economist Philippe Aghion. The committee reviewed a study of nine thousand French companies and found that firms using AI created more jobs than comparable firms that did not use AI. The explanation is that AI increases productivity and competitiveness: expanded market activity raises demand for products and services and, consequently, employment.
The committee estimated that only around 5% of jobs are genuinely at risk from AI—meaning the share exposed to real displacement is relatively small.
Conclusion: less destruction, more transformation
Julia concludes that generative AI will be far less destructive to employment than some fear. Losses will occur, but they are likely to be marginal; the main effect will be transformation and higher demand for new skills. Everyone must learn to use AI—tools are accessible and relatively easy to learn—and the greatest barriers are often fear of the unknown or inertia.
This article is based on edited excerpts from Luc Julia’s book Robot, not pilot.
