Companies Review AI Usage Metrics

Summary

Tom Fishburne's cartoon commentary highlights how companies like Meta, Amazon, and Uber have abandoned AI usage leaderboards after employees manipulated metrics. According to Visier data, 48% of employees admit to exaggerating their AI usage. An internal Amazon memo acknowledges that these metrics are 'deeply flawed' and 'disconnected from what matters.'

Why it matters

The piece captures a real and costly failure mode in corporate AI adoption. When usage volume becomes a KPI, it incentivizes performative behavior rather than genuine productivity gains. Employees run agents on unnecessary tasks or inflate their expertise. Uber consuming its entire 2026 AI coding budget in just four months serves as a concrete example of this financial damage. The article also emphasizes that the measurement problem is not new; Goodhart's and Campbell's laws have warned about this for decades, yet tech giants have continued to repeat the mistake.

Highlights

• Visier: 48% of employees admit to occasionally exaggerating their AI usage or expertise
• Meta and Amazon removed internal AI usage leaderboards after employees ran agents on unnecessary tasks to climb the rankings
• Amazon internal memo: 'If your job doesn't benefit from current AI tools, that is a finding, not a failure'

Critical take

The article is a single-panel cartoon commentary and lacks the depth of investigative journalism. The Visier statistic is cited without specifying the methodology. The Amazon memo is summarized in a single sentence. No independent verification is provided for Uber's budget claim. Consequently, the text reads more as a sharp editorial opinion than a comprehensive analysis of how companies should measure AI return on investment.