ECONOMIC AND SOCIAL RISK · PROSPECTIVE SCENARIO
When human labour gradually ceases to be indispensable
The risk is not limited to the number of jobs eliminated. It also concerns the place human labour retains in value creation, learning and economic power.
This pathway examines a possible systemic risk. It does not describe a demonstrated consequence of current AI.
STARTING POINT
Starting point
This pathway examines a possible systemic risk. It does not describe a demonstrated consequence of current AI.
As models perform more intellectual tasks, use software, make decisions and act autonomously, organisations can progressively replace part of human labour with capital, compute and AI models. The shift need not happen suddenly: it can occur one task, occupation and decision at a time.
01
From copilot to economic infrastructure
AI is first introduced as a productivity tool. Its role can then move from assisting to recommending, deciding by default and acting autonomously. As human intervention begins to look slower or more expensive, competitive pressure can turn a technology choice into an economic necessity.
02
An economy in which labour matters less
Work is both a source of income and a form of economic power. If a larger share of value came from capital-intensive AI infrastructure, consequences could include weaker bargaining power, fewer entry-level learning routes and greater concentration of economic power.
What happens to a society when the main economic contribution of most of its members gradually becomes less necessary?
03
The ‘intelligence curse’: when incentives change
The authors draw an analogy with the resource curse. If organisations and states derive progressively more value from automated systems than from employees, incentives to invest in human capital could change. This is a systemic-risk scenario, not a demonstrated consequence of current AI.
04
And before unemployment?
The first signals may appear in work organisation: fewer junior positions, tasks removed from roles, higher productivity expectations, reduced discretion, greater monitoring, altered staffing ratios, shrinking learning opportunities and uncertainty about career pathways.
05
Occupational health as an early-warning system
Occupational health cannot predict future model capabilities, but it can observe what deployment does to real work. Useful signals include staffing, task content, workload, learning, autonomy, reported difficulties, sickness absence, incidents, turnover, evaluation disputes and the ability to challenge automated decisions.
06
Prevent dependence before it takes hold
Preserve human capability, entry routes and professional learning; keep alternative processes operational; measure hidden human work; ensure people can contest, correct and stop automated processes; and track bargaining power, skill transfer and job quality as deployment scales.
This is a prospective risk analysis, not a forecast and not evidence that a particular macroeconomic outcome is inevitable.