Amazon has just crossed an unprecedented industrial threshold: a growing number of robots now operate in its worldwide warehouses, representing a significant multiplication compared to 2019. But contrary to apocalyptic predictions, this massive automation has not destroyed employment — it has transformed it. The Seattle company has created many new technical job categories while considerably reducing staffing per warehouse.
This transformation illustrates a reality that public debates still struggle to grasp. Automation does not mechanically eliminate jobs: it reconfigures them, displaces them, requalifies them. But this mutation of work is accompanied by a degradation of conditions for the majority of employees, who have become isolated supervisors of machines in increasingly precarious environments.
The Essentials
- Amazon operates several hundred thousand robots in its warehouses by late 2024, a sharp increase since 2019
- The company has created many new qualified job categories linked to robotics
- Staffing per warehouse has dropped considerably while productivity surges significantly
- Working conditions are deteriorating for a large majority of remaining employees, isolated and under constant algorithmic surveillance
From Automation to Replacement: The Numbers Settle the Debate
Research by Daron Acemoglu and Pascual Restrepo at MIT demonstrates that each industrial robot eliminates an average of 3.3 jobs within a 50-kilometer radius. Mechanically applied, Amazon’s robot deployment should have destroyed several million positions in the United States. The reality is more nuanced.
Amazon today employs 1.5 million people worldwide, several hundred thousand more than in 2019 despite the robotics explosion. This net growth masks, however, a radical reorganization. In a typical warehouse, staffing levels have decreased significantly between 2020 and 2024. But the remaining positions bear no resemblance to the previous ones.
The transformation is evident in job postings. Amazon now recruits massively for “robotics associates” — operators trained in six weeks to supervise automated systems. These jobs represent a growing share of recruitment. In parallel, the company hires numerous robot maintenance technicians and logistics optimization engineers.
New Professions for a Post-Industrial Economy
Amazon’s internal catalog today lists many job categories that did not exist five years ago. These creations span all qualification levels. “Robot whisperers” — diagnosticians specialized in machines’ behavioral failures — benefit from attractive compensation after brief training. “Flow architects” design optimized pathways for mobile robots.
This taxonomic creativity reveals a deliberate strategy. Amazon segments each function into micro-specializations to maximize productivity and minimize dependence on individual skills. A traditional “picking supervisor” is now divided into “quality inspector,” “speed optimizer,” and “error tracker” — three distinct positions, each expert in a fragment of the process.
The approach works economically. Productivity per employee has increased considerably between 2020 and 2024 according to the company’s internal data. Each Hercules robot — the orange autonomous shuttles that transport shelving — replaces several mobile employees but generates new supervision and maintenance jobs. The net balance remains positive for Amazon, deficit-producing for overall employment.
The Programmed Isolation of New Workers
Behind employment statistics lurks a major qualitative degradation. A large majority of Amazon employees who now work alongside robots operate in an environment of total algorithmic surveillance. Every gesture is tracked, timed, evaluated by the company’s systems.
Sarah Chen, a former “fulfillment associate” turned “robotics coordinator” at the Tracy, California warehouse, describes this transformation: “Before, I worked as part of a human team. We talked to each other, we helped each other out. Now, my colleagues are machines. My manager is an algorithm that sends me alerts on my badge every ten minutes.”
This transformation massively affects workers’ mental health. Employees in direct contact with robots report far more symptoms of social isolation than in non-automated zones. Turnover reaches very high levels in roboticized positions, forcing the company to recruit constantly.
American unions document this precarization. The Amazon Labor Union counts a sharp increase in sick leave for mental health disorders in the most automated warehouses. “They’re selling us new jobs, but they’re the same repetitive tasks with less humanity,” analyzes Chris Smalls, president of the organization.
The Unequal Geography of Automation
Amazon deploys its robots according to precise geoeconomic logic. European and North American warehouses concentrate the majority of robotics equipment while Asian centers remain largely manual. This distribution reflects the relative cost of labor and local regulations.
In Germany, where Amazon employs many people, automation accompanies collective negotiations. Management commits to retraining a large portion of employees affected by robotization toward better-paid technical positions. The German model inspires Europe, which is building its AI Factories and discovering that governance matters as much as GPUs, prioritizing skills development over workforce reduction.
In the United States, the logic differs. Amazon automates prioritarily in states where minimum wage increases. According to the company’s internal calculations, each dollar of hourly increase accelerates robot deployment. Some states concentrate a disproportionate share of robots relative to the number of warehouses.
This geography reveals the limits of automation as social policy. The technical jobs created require training that a majority of Amazon’s current workforce cannot pursue by its own criteria. The company invests massively in its retraining programs, but reaches fewer than one-third of employees affected by robotization.
Automation Changes Syndical Power Dynamics
Robotization radically transforms social relations at Amazon. Traditional strikes lose their effectiveness against warehouses that function largely without human intervention. Hercules robots continue transporting shelves even when supervisors go on strike.
This evolution forces unions to rethink their strategies. The Amazon Labor Union is experimenting with new forms of action: coordinated slowdowns of supervision rates, temporary blocking of software updates, occupation of robot maintenance zones. These tactics exploit specific vulnerabilities in automated systems.
Amazon’s response reveals the strategic stakes. The company is massively recruiting “labor relations specialists” trained in conflicts in roboticized environments. It is also developing early detection systems for social tensions based on analysis of internal communications and workplace behavior patterns.
This arms race in technological workplace relations prefigures the future of industrial unionism. As analyzed in American AI enriching capital before labor, automation reshapes the balance of power between employers and workers, generally to the detriment of the latter.
The Troubling Productivity of Ghost Warehouses
Since 2023, Amazon has been testing “dark warehouses” — entirely automated warehouses that operate without lighting or heating. These installations employ only a small number of technicians compared to several hundred in an equivalent traditional warehouse. Their productivity significantly exceeds that of mixed human-robot centers.
Several ghost warehouses currently operate in the United States, the United Kingdom, and the Netherlands. Amazon plans to build more by 2027. These centers exclusively process orders for standardized small items — a significant portion of the company’s total volume.
The extrapolation concerns labor economists. If Amazon generalizes this model, the company could handle its current volume with far fewer employees than today. This prospect is prompting governments to question the taxation of automation and mechanisms for redistributing productivity gains.
Amazon’s experience reveals an uncomfortable truth about the future of work. Automation does not destroy employment through brutal suppression but through silent transformation. It effectively creates new professions while degrading the conditions of the majority of workers. This mutation poses a central political question: how can the formidable productivity gains from automation be redistributed to those who bear its social costs?