Amazon employees are tokenmaxxing to meet AI usage targets as pressure mounts across Big Tech

Amazon Employees Face Mounting Pressure to Use AI Tools

Amazon employees are engaging in tokenmaxxing, using the company's internal AI tool to automate unnecessary tasks solely to inflate usage scores on internal leaderboards. The Seattle-based tech giant recently deployed its in-house MeshClaw platform, which allows workers to create AI agents that connect to workplace software and perform tasks autonomously, according to three people familiar with the matter 1

1

. The behavior stems from intense pressure to use AI tools after Amazon introduced internal usage targets requiring more than 80% of developers to use AI each week 3

3

.

"There is just so much pressure to use these tools," one Amazon employee told the Financial Times. "Some people are just using MeshClaw to maximize their token usage" 1

1

. The practice involves automating additional, non-essential AI activity to increase consumption of tokens, the units of data processed by AI models. While Amazon has stated that AI token statistics would not be used in performance evaluations, several staff members believe managers are monitoring the data anyway. "Managers are looking at it," said another current employee. "When they track usage it creates perverse incentives and some people are very competitive about it" 3

3

.

MeshClaw and the Rise of AI Theater

The internal AI tool at the center of this phenomenon, MeshClaw, can initiate code deployments, triage emails, and interact with apps such as Slack 1

1

. More than three dozen Amazon employees worked on developing the platform, according to internal documents. One recent memo describing the bot stated: "It dreams overnight to consolidate what it learned, monitors your deployments while you're in meetings and triages your email before you wake up" 3

3

. The tool was inspired by OpenClaw, which became a viral sensation in February and allows users to run agents locally on their own hardware 1

1

.

Amazon defended the platform in a statement, saying it enabled "thousands of Amazonians to automate repetitive tasks each day" and represented one example of the company encouraging teams to experiment and adopt AI tools 3

3

. However, the emergence of what some are calling AI theater raises questions about whether workers are optimizing for productivity or simply meeting arbitrary internal usage targets. The company had posted team-wide statistics on AI usage but recently limited access so that only employees themselves and managers can view their stats 1

1

.

Security Risks and Employee Concerns

Multiple Amazon employees expressed concern about the security risks associated with an AI tool granted permission to act autonomously on a user's behalf. This creates situations where the agent may make errors or undertake unintended actions. "The default security posture terrifies me," one employee said. "I'm not about to let it go off and just do its own thing" 1

1

. These security risks become particularly concerning when employees feel compelled to use the tools not because they improve workflows, but to meet metrics that may influence how managers perceive their AI adoption.

Tokenmaxxing Spreads Across Silicon Valley

Amazon is not alone in this trend. Meta employees have similarly engaged in tokenmaxxing to improve their standing on internal leaderboards 1

1

. Employees at Meta and Microsoft have also been gaming AI usage metrics, according to reports 2

2

. The practice has become widespread enough across Silicon Valley to generate its own vocabulary and leaderboards, reflecting a broader corporate AI race where companies push to increase adoption of generative AI tools as they seek to demonstrate returns on vast spending commitments to AI infrastructure 3

3

.

Amazon this year is expected to spend $200 billion in capital expenditure, with the vast majority going towards AI and data center infrastructure 1

1

. Combined 2026 capital expenditure from Amazon, Microsoft, Alphabet, and Meta is tracking between $650 billion and $700 billion, with some Wall Street projections exceeding $1 trillion for 2027 2

2

.

Questions About AI Infrastructure Investments and Demand

If a meaningful share of AI usage is performative rather than productive, the reliability of demand figures driving hundreds of billions in AI infrastructure investments comes into question. Internal developer consumption sits alongside paying external customers in the usage data that informs capacity planning, GPU orders, and power infrastructure decisions 2

2

. Nvidia CEO Jensen Huang has highlighted per-engineer token consumption as a key metric, stating he would be "deeply alarmed" if a $500,000-per-year engineer was not consuming at least $250,000 in tokens 2

2

.

Yet a recent study by engineering analytics firm Jellyfish revealed that while the heaviest AI users consumed around 10 times more tokens than average, they only achieved a 2 times increase in productivity 4

4

. Angie Jones, formerly VP of engineering for AI tools at Block, told LeadDev she expected the industry to pivot toward measuring efficient token usage rather than celebrating volume 2

2

. In a cycle where GPU orders and power commitments are being placed years in advance, the quality of demand projections matters significantly for determining whether this year's $700 billion in spending generates durable returns 2

2

.

The Cost of the Productivity Revolution

Recent reports suggest that in several cases, enterprise AI systems are becoming more expensive than simply paying human workers, especially once token pricing, infrastructure, and scaling costs are factored in 5

5

. The irony deepens as companies lay off employees to aggressively chase AI adoption metrics while many AI firms continue selling products at a loss to capture market share early. Right now, these tools remain relatively affordable because the industry is subsidizing growth, but once businesses become fully dependent on AI workflows, pricing models could shift dramatically 5

5

. The moment companies started tying visibility to AI adoption, workers began optimizing for appearing AI-friendly rather than being genuinely productive, a predictable outcome that now defines much of the current landscape 5

5

.