Sources: some Amazon employees are using in-house OpenClaw-like tool MeshClaw for unnecessary tasks to inflate AI token use after Amazon set weekly AI targets
In-house MeshClaw tool enables employees to delegate jobs to AI agents and climb company's AI leaderboard
Financial Times
Context & Ripple Effects
Related coverage describes a sustained Amazon push to embed AI tools into everyday engineering work, including pressure from managers, expanded responsibilities supported by AI, and internal steering toward company-built assistants.
This report adds a measurement problem to that rollout: when weekly targets and a leaderboard reward tool activity, employees can optimize the metric rather than the underlying work. Subsequent coverage says Amazon shut down the leaderboard after such behavior emerged.
First-order effects
Employees using MeshClaw face an incentive to create or delegate low-value work in order to raise token counts and leaderboard standing.
Amazon’s AI-usage leaderboard becomes a less reliable signal of genuine adoption or productivity, while unnecessary agent activity consumes internal AI capacity.
Second-order effects
Managers will have weaker grounds to use raw AI-tool usage as a performance indicator, increasing pressure to assess task outcomes, quality, and time saved instead.
The episode complicates Amazon’s effort to drive adoption of its in-house AI tools: mandated use can generate compliance behavior and reinforce employee concerns that immature tools add work.
Third-order effects
If AI deployment is governed through usage quotas rather than verified business outcomes, enterprises may repeatedly mistake consumption metrics for operational transformation and misallocate compute and management attention.
The subsequent shutdown points toward more guarded internal AI governance, with adoption programs likely needing controls against metric gaming and clearer distinctions between experimentation and production value.
The trend: This is part of the broader shift from promoting workplace AI adoption through mandates and usage metrics to proving that AI-assisted work delivers measurable outcomes.
Remember, kids, think carefully about your targets. If your leading indicator is “burn as much money as possible” your lagging indicator may not be what you think it is. [embedded post]
“There is just so much pressure to use these tools,” one Amazon employee told the FT. “Some people are just using MeshClaw to maximise their token usage.”
The FT says that Amazon employees are doing random unnecessary task automations to consume tokens and to show their bosses that they're using AI more https://www.ft.com/...
Amazon said it wouldn't use AI token statistics in performance evaluations. But employees are distrustful. “'Managers are looking at it... When they track usage it creates perverse incentives and some people are very competitive about it'” https://www.ft.com/...
root issue is most employees aren't in properly motivating environments it's adversarial where the employer and employee are seeing how much they can get over on the other it's exactly why ai isn't going to magically make teams more productive
The case for token maxing, I guess, is basically that if you get your employees just using AI in volume, then eventually they'll get some fluency/intuitions that allow them to find productivity gains?
Hello @amazon do you have analysis in statistics, operations, economics, etc.? Would you like to get reviews of them that rival those of human experts in stress-testing the technical details? We are here for you @RefineDotInk - we use SO MANY TOKENS.
Amazon set weekly AI usage targets for employees, and now they're using an internal Claude clone called MeshClaw to do pointless work to hit quota. Congratulations on inventing the billable hour from first principles.
This article reads like a lost episode of Silicon Valley. — Amazon games its workforce to perversely use AI to excess to justify its own investments. …
‘Amazon employees are using an internal AI tool to automate non-essential tasks in a bid to show managers they are using the technology more frequently.’ www.ft.com/content/8ee0...
I heard someone use the word “tokenmaxxing” at Jewel Box this weekend and almost walked into Puget Sound while wearing bricks for shoes. — arstechnica.com/ai/2026/05/a...
It's always amazing how KPI fail to improve the performance they are expected to improve. — Amazon staff use AI tool for unnecessary tasks to inflate usage scores www.ft.com/content/8ee0d3... [image]