# Amazon memo calls AI usage metrics flawed but keeps internal tracking

Digest AI · Policy & Regulation · published 2026-10-31T11:57:00Z

Canonical: https://digestai.news/story/amazon-memo-calls-ai-usage-metrics-flawed-but-keeps-internal-tracking

## Summary

Amazon’s internal memo, authored by VP Terkowitz in September 2026, redefines how the company measures software engineers’ AI adoption. The document rejects token counts and raw code generation as valid productivity metrics, calling them ‘deeply flawed’ because they don’t reflect customer value. Instead, it instructs engineers to document cases where AI tools don’t help their work and frames skepticism as a skill—though it doesn’t ban AI use or remove usage data from managers’ view. The memo follows months of internal pressure to adopt Amazon’s Kiro developer platform and coding agents, including a now-deprecated leaderboard called KiroRank that employees gamed by inflating token usage to climb rankings.

The memo’s timing is notable amid Amazon’s broader push to prove AI’s business impact, alongside record spending on AI infrastructure. While it avoids outright bans, the guidance creates tension: managers are told to study adoption data for insights but not to rank employees by it. Critics, like commentator Corey Quinn, argue the memo’s mixed messaging—calling token counts flawed while keeping them in play—creates ambiguity, especially during a period of job cuts. The memo also highlights a cultural contradiction: AWS bills customers by token consumption while telling its own staff to treat tokens as a cost to minimize. Engineers are now asked to prove AI’s limitations in writing, adding administrative burden to an already strained workforce.

## Key points

- Amazon’s memo rejects token counts and raw code generation as flawed productivity metrics, per VP Terkowitz’s September 2026 guidance
- Engineers must document cases where AI tools fail to help work, framing skepticism as a skill, but usage data remains available to managers
- KiroRank, an internal leaderboard for token usage, was deprecated in May 2026 after employees gamed it to inflate metrics

## Why it matters

The memo signals a shift in how large tech companies measure AI adoption, prioritizing long-term outcomes over usage volume—a move that could influence vendor contracts and internal policies beyond Amazon. It also raises questions about cultural alignment between AI spending and frugality principles, particularly for AWS customers who pay by token.

## Sources

1. [Amazon AI Memo Mixes 2 Messages After KiroRank](https://shattered.io/amazon-ai-memo-mixed-messages-tokenmaxxing-kirorank-2026) (shattered.io, 2026-10-31)

## Cite

Digest AI, "Amazon memo calls AI usage metrics flawed but keeps internal tracking", 31 October 2026, https://digestai.news/story/amazon-memo-calls-ai-usage-metrics-flawed-but-keeps-internal-tracking

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/amazon-memo-calls-ai-usage-metrics-flawed-but-keeps-internal-tracking.json