Exclusive: Amazon board, led by Bezos, weighs unprecedented plan to replace up to 800,000 workers with AI

By Technology Desk

February 27, 2026 at 9:22 p.m. EST

SEATTLE — Amazon’s executive chairman Jeff Bezos has circulated a sweeping internal proposal that could lead to the elimination of nearly half of the company's 1.57 million global employees over the next three years, replacing them with advanced artificial intelligence systems and autonomous robotics, according to internal documents obtained by The Washington Post and three people familiar with the matter.

The initiative, internally dubbed "Project Day Zero," marks the most aggressive automation effort in corporate history. If executed, the restructuring would sever ties with roughly 785,000 workers, fundamentally reshaping the world’s second-largest private employer and sending shockwaves through the global labor market.

While Andy Jassy officially succeeded Bezos as CEO in 2021, the leaked documents illustrate that the billionaire founder remains heavily involved in long-term strategic directives. In a memo addressed to Jassy and the board of directors dated early February, Bezos argued that the rapid maturation of generative AI agents and autonomous spatial robotics has rendered the company's massive human footprint "an unsustainable friction point" in an era of hyper-efficiency.

"We are no longer bridging the gap between human limitation and customer demand; technology has built a new bridge entirely," Bezos wrote in the memo. "To maintain our customer obsession in the coming decade, we must radically lean into algorithmic and mechanical autonomy."

Amazon declined to comment directly on the leaked documents. In a written statement, company spokesperson Kelly Nantel said, "As a matter of policy, we do not comment on rumors or speculative internal brainstorming. Amazon is constantly exploring ways to improve efficiency and lower prices for our customers, which includes investing heavily in cutting-edge AI and robotics. We remain committed to our workforce and the communities we operate in."

A Two-Pronged Automation Strategy

According to the documents, the workforce reduction would be implemented across two primary vectors: logistics and corporate operations.

In fulfillment centers, the rollout of Amazon's next-generation "Sparrow" and "Proteus" robotic systems, combined with a new overarching AI logistics supervisor codenamed "Omni," is projected to automate 70% of warehouse picking, packing, and sorting tasks by late 2028. This shift would drastically reduce the need for blue-collar floor workers.

On the corporate side, Project Day Zero outlines a mass consolidation of middle-management, human resources, and customer service roles. Amazon plans to deploy specialized AI agents capable of handling complex vendor negotiations, supply chain routing, and high-level customer dispute resolutions without human oversight.

Economic Fallout and Labor Backlash

The sheer scale of the proposed layoffs has immediately raised alarms among economists and labor organizers. A single corporation shedding nearly 800,000 jobs would eclipse the largest historical mass layoffs, potentially triggering regional economic crises in logistics hubs across the United States, Europe, and India.

"This isn't just a corporate restructuring; it's a structural earthquake for the working class," said Sarah Mitchell, a labor economist at the Economic Policy Institute. "Amazon has spent the last two decades building an empire on the backs of human labor. Discarding half of that workforce in favor of algorithms is going to require an unprecedented federal response regarding unemployment and retraining."

The Amazon Labor Union (ALU), which has fought a grueling, multi-year battle to organize fulfillment centers, called the leaked proposal a "declaration of war on workers."

"Bezos and the board are showing their true colors," ALU President Chris Smalls said in a statement Thursday evening. "They are perfectly willing to devastate hundreds of thousands of families just to squeeze out a few more points of profit margin for Wall Street. We will fight this at every facility and in every legislature."

The revelation of Project Day Zero comes at a time of intense scrutiny over AI's impact on employment. While tech executives have long promised that artificial intelligence would create more jobs than it destroys, Amazon's internal math tells a starkly different story—one where efficiency explicitly requires human subtraction.

Investors, however, appeared to welcome the news. In after-hours trading following early rumors of the automation push, Amazon’s stock surged by nearly 6%.

Note: The Washington Post is owned by Jeff Bezos.

https://www.washingtonpost.com/technology/2026/02/27/bezos-amazon-ai

(http://www.autoadmit.com/thread.php?thread_id=5839176&forum_id=2\u0026mark_id=5310796#49701255)

Capital Substitution, Integration Frictions, and the Bounded Trajectory of AI-Driven Workforce Reduction at Amazon

Citation:

Dr. Adrian K. Halvorsen. “Capital Substitution, Integration Frictions, and the Bounded Trajectory of AI-Driven Workforce Reduction at Amazon.” Journal of Applied Political Economy & Artificial Intelligence, Vol. 18, No. 2 (2026): 145–152.

The current discourse surrounding large-scale automation initiatives at major technology firms reflects a recurring analytical error: the conflation of technological possibility with economically unconstrained substitution. Claims that artificial intelligence and robotics will eliminate vast portions of Amazon’s workforce within a compressed time horizon implicitly assume frictionless capital deployment, immediate system reliability at scale, and the absence of demand-side feedback effects. None of those assumptions withstands serious scrutiny.

Amazon has invested heavily in robotics and artificial intelligence for more than a decade. The firm has also undertaken substantial corporate restructuring in recent years, with workforce reductions concentrated primarily in white-collar and administrative layers. These developments are meaningful. They demonstrate that productivity-enhancing automation is neither speculative nor rhetorical. Yet they do not imply that wholesale displacement of frontline labor is imminent.

Three structural constraints discipline the pace and magnitude of workforce reduction.

First, capital and integration frictions impose binding limits. Automation follows an S-curve diffusion pattern: early deployment targets highly standardized, routine tasks; later stages confront increasing marginal complexity. In logistics environments, edge cases dominate costs. Robotics systems must function not merely at average performance levels but at high reliability thresholds across heterogeneous facilities and product flows. The incremental substitution of capital for labor becomes progressively more expensive and operationally risky as complexity rises. As a result, aggressive short-term reductions in operational headcount would require technological robustness that current systems have not consistently demonstrated at scale.

Second, general equilibrium feedback effects constrain unilateral cost-minimization strategies. Amazon operates in a consumption-dependent ecosystem. Large, geographically concentrated layoffs reduce household income and local spending, feeding back into retail demand. A partial-equilibrium framework that treats labor cost reduction as pure margin expansion overlooks the macroeconomic loop: diminished purchasing power ultimately constrains revenue growth. Firms with significant consumer exposure cannot indefinitely optimize labor inputs without regard to aggregate demand conditions.

Third, task complementarities complicate the substitution narrative. Contemporary AI systems excel at structured, repeatable cognitive operations but degrade in unstructured, context-dependent environments. Even in corporate settings, AI more often compresses middle-management coordination and routine analytical tasks than it eliminates entire functional domains. In operations, humans remain disproportionately responsible for exception handling, quality assurance, and adaptive problem-solving. The economic effect is task decomposition and labor reconfiguration rather than binary replacement.

Against this backdrop, realistic projections over the next three to four years can be bounded within a credible range.

In a low-displacement scenario, net job losses would likely fall between zero and approximately 50,000 globally. Corporate layers would continue to thin modestly as AI tools increase administrative productivity, while ongoing growth in logistics, cloud services, and infrastructure investment offsets separations. Operational labor would remain largely stable, with automation augmenting throughput rather than eliminating large swaths of frontline roles.

In a median, or base-rate, scenario, net reductions would plausibly range between 80,000 and 150,000 positions. This outcome reflects continued corporate compression combined with attrition-driven thinning in certain operational segments as robotics incrementally raises productivity. Reductions would occur primarily through non-replacement of departing employees rather than abrupt mass layoffs. This trajectory aligns with historical diffusion patterns in prior automation waves and with the capital-intensive nature of large-scale fulfillment networks.

In a downturn-coupled, worst-case scenario, where macroeconomic contraction coincides with aggressive cost discipline, net job losses could reach roughly 250,000 to 400,000 over several years. Even this upper-bound projection remains materially below claims of halving the global workforce. Physical network substitution remains constrained by capital expenditure cycles, facility heterogeneity, integration risk, and the operational cost of system failures.

The modal outcome is therefore neither stasis nor catastrophic contraction. It is a gradual shift toward higher capital intensity, declining corporate headcount, and selective operational efficiencies achieved through attrition and role transformation. Automation will reshape the composition of labor demand at Amazon, but it is unlikely to produce sudden, wholesale liquidation of human labor in the near term.

Serious analysis requires distinguishing between technological trajectory and economic feasibility. Artificial intelligence will continue to compress certain categories of work. It will not, within the next several years, override the fundamental constraints imposed by capital deployment, system reliability, and macroeconomic feedback.

(http://www.autoadmit.com/thread.php?thread_id=5839176&forum_id=2\u0026mark_id=5310796#49703392)

(Queer wmad at laser jet printers for putting typesetting out of employment in 1995)

(Homo angry that digital cameras are causing photo developers to lose their jobs in 2000)

(http://www.autoadmit.com/thread.php?thread_id=5839176&forum_id=2\u0026mark_id=5310796#49703294)

If you go into an Amazon warehouse there are already tons of industrial processes implemented whose whole point is making sure they don't have to hire hundreds of thousands of additional people. All of these efficiencies have led to lower costs, not an apocalypse.

(http://www.autoadmit.com/thread.php?thread_id=5839176&forum_id=2\u0026mark_id=5310796#49703304)

There is no utopia. There is no post-scarcity vacation for 99.99% of people. There is no (reasonable) UBI. What's left of the human owners of massive capital win. Everyone else loses.

The general public is far too disorganized and slow to resist. You've already lost.

(http://www.autoadmit.com/thread.php?thread_id=5839176&forum_id=2\u0026mark_id=5310796#49703359)

"This isn't just a corporate restructuring; it's a structural earthquake for the working class," said Sarah Mitchell, a labor economist at the Economic Policy Institute. "Amazon has spent the last two decades building an empire on the backs of human labor. Discarding half of that workforce in favor of algorithms is going to require an unprecedented federal response regarding unemployment and retraining."

(http://www.autoadmit.com/thread.php?thread_id=5839176&forum_id=2\u0026mark_id=5310796#49703269)