Financial Times · TechnologyMay 12, 2026, 12:00 UTC+8

Amazon staff use AI tool for unnecessary tasks to inflate usage scores

Amazon’s in-house MeshClaw tool lets employees delegate work to AI agents and raise their position on the company’s AI leaderboard; the post does not disclose the number of staff involved, the scoring rules, or the specific unnecessary tasks.

Why it matters: FT gives a concrete Amazon AI-adoption gaming story, clearing HKR-H/K/R. Missing participant count, scoring rules, and task examples keep it at the featured threshold, not a must-write item.

Read the original ↗Export Markdown

Take

Amazon’s MeshClaw story hits the chronic enterprise AI failure: measuring “how much AI was used” instead of “how much work disappeared.” The title says employees handed unnecessary tasks to AI agents to climb an internal leaderboard. The article body is paywalled here, so staff count, scoring rules, and task categories are not disclosed. That limits the scale claim, but not the mechanism risk.

I’m not surprised. Microsoft Copilot, Google Workspace, and Salesforce Einstein rollouts have all run into the same split: activation metrics look clean, workflow change is messy. If Amazon used a leaderboard to push agent usage, it basically invited Goodhart’s Law into the org chart. Agent adoption is not hurt only by low usage. It also gets poisoned when workers create fake work to prove they are “AI-native,” then contaminate telemetry, ROI math, and model feedback loops.

Source: Financial Times · Technology · ft.com
Published: May 12, 2026, 12:00 UTC+8 · May 12, 2026