Sources: Amazon has shut down an internal leaderboard that tracked employees' use of AI tools after workers tried to boost their scores with needless tasks
Senior executive Dave Treadwell tells staff 'don't use AI just for the sake of using AI' as costs rise
Financial TimesRafe Rosner-Uddin
Context & Ripple Effects
Amazon’s internal push to embed AI into engineering work has been accompanied by higher output expectations, weekly usage targets, and broader role changes for AWS developers. Employees have also reported that immature tools can add work rather than remove it.
The company had already tightened oversight of AI-assisted code after outages, requiring more senior review for some changes. Against that backdrop, ending a usage-based ranking system signals that raw adoption metrics conflicted with both cost discipline and quality control.
First-order effects
Employees lose a visible incentive to generate low-value AI activity simply to improve a leaderboard position, while managers receive a clearer directive to judge use by work outcomes rather than tool consumption.
Amazon can curb unnecessary internal AI-tool calls at a time when the related coverage points to meaningful GPU spending and heightened scrutiny of AI-assisted engineering work.
Second-order effects
Teams that had translated AI mandates into usage quotas or rankings will need alternative measures, such as delivery quality, reliability, or demonstrated time saved; those measures are harder to standardize than activity counts.
The reversal reinforces the need for stronger review and governance around internal AI workflows, especially where increased usage can create both infrastructure costs and additional remediation work.
Third-order effects
If similar programs are revised, enterprise AI deployment is likely to shift from adoption-scorecard management toward outcome-based accountability, with fewer incentives to treat token use or tool engagement as a proxy for productivity.
The episode illustrates a persistent constraint on AI-led workforce redesign: companies can mandate uptake quickly, but durable gains depend on tool reliability, controls, and incentive systems that do not reward performative usage.
The trend: This is one data point in the move from blanket internal AI-adoption mandates to more cost-, quality-, and outcome-conscious deployment governance.
A month and a half ago I shared how tokenmaxxing is spreading as a weird, new trend, and all it does is generate a massive company bill. Amazon learning what was obvious even back then. Encouraging tokenmaxxing is very expensive and honestly pretty stupid [image]
just throwing it out there, but maybe both the tokenmaxing is the new religion *and* the tokenmaxing is dead discourse are overblown and the truth is about optimizing usage the same way you would virtually all other things?
Eventually, the world will figure out that AI use should be all about quality, not quantity. I'm surprised it's not obvious to everyone already. https://www.ft.com/content/b1a62a7f- 6df5-4c90-94ce-64ce9c9961b6
I can now probably say this: Two months ago, inside Anthropic someone suggested building a token leaderboard. A heated internal debate followed and the decision was made to *never* ever do it... because several people inside Anthropic simply thought ahead of the consequences
Once again reminded of the time the Groupon sales floor figured out how to smash their talk time targets by calling the toll free Hall and Oates hotline at lunchtime. [embedded post]
Amazon scraps AI leaderboard to stop workers boosting usage scores — Senior executive tells staff 'don't use AI just for the sake of using AI' as computing costs rise
Im calling BS on this story. 1. That would be 100,000 employees spending $5k/mo each or 10,000 employees averaging $50k/mo each. No way. 2. Anthropic isn't going to float that much on an invoice without checking in 3. The source being a consultant talking about a client = lie [im…
None of this is satire. → A company spent $500,000,000 on Claude in one month because nobody set usage limits → Uber ran leaderboards ranking engineers by how much AI they used, not what they shipped → Uber burned their entire 2026 budget by April. Their COO said he can't
Amazon scraps AI leaderboard to stop workers chasing usage scores | Senior executive Dave Treadwell tells staff 'don't use AI just for the sake of using AI' as costs rise
That was a really expensive way to learn tokenmaxxing is dumb. Maximizing LLM usage just for the sake of flexing your use of AI is a recipe for disaster.
Classic “show me the incentive and I'll show you the outcome” When mgmt made this decision they should have just taken it one step further and thought: 1. what would I do if I had this rule imposed upon me 2. what would a smart person who is trying to take advantage of this
This is exactly why “AI usage” is a terrible KPI. The incentive shouldn't be how many prompts an employee runs. It should be whether the AI-powered workflow they create gets adopted by others and actually saves time, increases output, or improves quality. Otherwise you just get
Who would've thought encouraging employees to use an expensive product as much as possible with no oversight over than measuring how much they're using it could possibly backfire?
Every day it seems we're getting articles about the real costs of AI. If AI is to be reined in, it won't be because of the job insecurity its causing or environmental harm, but because it's proving too costly compared to human labor.