🔍 Read the full analysis: A New Way To Frame AI Productivity: Amazon’s Outcomes Focus on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Amazon published revised software-engineering guidance on October 27 telling teams not to treat token use or AI adoption as performance targets. The memo says those metrics can help show patterns but should be judged against customer, business and team outcomes.
Amazon published new guidance for its software engineers on October 27, warning that targets based on AI token use, tool adoption or code volume can reward activity rather than results. The memo says engineers and managers should judge AI by its impact on customers, the business and their teams—a distinction that matters as companies seek to measure productivity gains without encouraging wasteful use.
The guidance is part of a revised role guideline for Amazon software engineers, written by the company’s most senior engineers and introduced in a memo by Terk Terkowitz, a vice president and distinguished engineer in Amazon’s stores business. Amazon published the memo on its own news site. It does not prescribe a fixed list of AI duties, saying definitions could quickly become outdated as tools change.
The memo describes token counts, AI tool adoption, code changes and reviews as input measures that can be useful for spotting patterns, but “deeply flawed” when turned into targets. It warns that targeting them can increase costs and system complexity. Its examples link tokenmaxxing—using a tool for the sake of generating more tokens—to Amazon’s frugality principle, and measuring lines of code to a conflict with “invent and simplify.”
The guidance does not say teams should stop using AI. It says input measures currently “correlate well” with relevant outputs, while stressing that correlation does not make them good individual performance targets. It asks leaders to examine both low and high adoption: why some teams are not finding AI useful, and whether heavy users are actually delivering better results. Managers are also asked to use the tools in their own production work and share examples of tasks where they helped and where they did not.
Six months at Amazon
From token leaderboards to an outcomes memo
Amazon spent 2026 learning the same lesson twice: once with how its engineers use AI, once with its own AI models.
- May 2026Token leaderboards backfireFinancial Times (via The Decoder): staff give an internal agent tool pointless tasks to inflate token counts. Reported, anonymous sources.
- Late May 2026KiroRank “deprecated”Amazon spokesperson to Business Insider. SVP Dave Treadwell: “Please don’t use AI just for the sake of using AI.”
- 24 Jul 2026San Francisco AGI Lab site closesConfirmed by Amazon to GeekWire. Frontier model research continues under Pieter Abbeel.
- 28 Jul 2026Most flagship Nova models wound downBusiness Insider, people familiar: Premier, Omni, Reel, Canvas to “keep the lights on”. Amazon: models “remain one of the most important things we’re working on”.
- 14 / 30 Sep 2026Nova Sonic v1, Nova Reel, Nova Canvas reach end of lifeAWS Bedrock model lifecycle page. Requests to them now fail.
- 26 Oct 2026Anthropic files its S-1Amazon notes: $6.2 billion principal, 12% ownership cap that Amazon can waive. Anthropic commits at least $110 billion to Amazon through April 2036.
- 27 Oct 2026The memo: outcomes, not tokenmaxxingUsage and throughput metrics are “deeply flawed” as targets. “Tokenmaxxing is at odds with frugality.”
- 29 Oct 2026Amazon’s third-quarter resultsAnnounced by Amazon. Not yet published when this article went live.
Sources: Amazon engineering memo (27 Oct 2026); Business Insider (29 May and 28 Jul 2026); GeekWire (24 Jul 2026); The Decoder (12 May 2026, summarising the Financial Times); AWS Bedrock model lifecycle (read 28 Oct 2026); Anthropic S-1 (26 Oct 2026).
Powered by Thorsten Meyer AI
Why AI Usage Targets Can Backfire
The memo addresses a measurement problem facing employers adopting generative AI: usage is easier to count than productivity. Token totals and adoption rates can provide a rough view of whether tools are being tried, but a target tied to those numbers can encourage unnecessary work and spending. The distinction affects how companies set engineering expectations, allocate AI costs and evaluate employees.
For workers, the memo offers a route to report that a tool does not help a particular task without treating that conclusion as failure. It says engineers should document the reasons and share them with managers and peers. For leaders, the instruction is to investigate results at both ends of the adoption range rather than assume either low use or high use signals performance on its own.
That approach also makes the memo more than a directive to adopt AI. It frames adoption as a question to test against actual work and outcomes. Amazon does not provide a new universal productivity formula; instead, it calls for judgment and evidence about whether the technology is helping in a given setting.
Products Worth Considering
Apple iPad Air 11-inch (M2), 512GB, Wi-Fi, Space Gray
Apple AirPods Pro 3 Wireless Earbuds
EcoFlow Smart Home Panel 2 Whole-Home Backup System Hub
As an affiliate, we earn on qualifying purchases.
From Internal Leaderboards to Guidance
The memo follows reports of internal efforts to track AI use. In May, the Financial Times reported that some Amazon employees assigned unnecessary tasks to an internal agent called MeshClaw to increase token counts. The accounts were anonymous and have not been independently confirmed in the supplied material. The report also said Amazon had a weekly AI-use target for developers and tracked token activity on leaderboards.
At the end of May, Amazon confirmed to Business Insider that a leaderboard called KiroRank had been deprecated. The company described it as an informal tracker created by a group of employees and said it was never intended to promote AI use for its own sake. Senior vice president Dave Treadwell also told staff, “Please don’t use AI just for the sake of using AI.” The October guidance now puts the distinction between usage and impact into an official engineering role guideline.
The source article also compares Amazon’s approach with its investments in AI models. It reports, citing Business Insider, that most flagship Nova models were moved to maintenance status, while Amazon’s filings valued its Anthropic holdings at about $190 billion on June 30 after roughly $18 billion in investment. Those figures and the characterization of Nova’s status are reported claims, not details established by the October 27 memo. Amazon’s scheduled third-quarter earnings report on October 29 may offer further financial context, but its results were not yet available when the source article was written.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo
How Amazon Will Judge Results
The memo does not specify a single replacement metric or explain exactly how managers will compare outcomes across teams and tasks. It also does not state whether the guidance changes compensation, promotion or formal performance-review processes. Its call for judgment leaves those implementation details open.
The relationship between AI adoption and output remains a claim in the memo: it says adoption metrics currently correlate well with outcomes, but does not provide supporting figures in the material available here. Reports about employees inflating token counts were based on anonymous accounts; Amazon’s confirmation that KiroRank was deprecated does not independently confirm every account of how the tracker was used.
Amazon’s third-quarter results were due October 29, after the memo’s publication. Their eventual figures may inform the company’s broader AI spending and investment picture, but they would not by themselves establish whether AI improved engineering productivity.
Earnings and Team-Level Reviews
Amazon is scheduled to report its third-quarter results on October 29, 2026. Those results will provide an update on the company’s finances, while the engineering memo sets out how managers should assess AI use within software teams.
Under the guidance, leaders are expected to examine teams with lower and higher adoption, use AI on their own production work, and share concrete examples of where the tools help or fall short. The next evidence of how the policy works will come from its application: whether teams use adoption data as a diagnostic, and whether they can connect AI use to better customer, business or team outcomes.
Key Questions
What did Amazon tell its engineers?
Amazon told engineers and managers not to use AI token counts, adoption rates or code volume as performance targets. The memo says to judge tools by their impact on customers, the business and the team.
Is Amazon telling staff to stop using AI?
No. The memo says adoption measures can correlate with useful outputs and does not discourage AI use. Its warning is against treating usage for its own sake as a goal.
What does the memo say if AI does not help a task?
It says that this is “a finding, not a failing.” Engineers are asked to document why the tool was not useful and share that information with peers and their manager.
What was the KiroRank leaderboard?
Amazon confirmed to Business Insider in late May that KiroRank, an informal employee-created tracker, had been deprecated. The company said it was not intended to promote AI use for its own sake.
When will Amazon report its next results?
Amazon’s third-quarter results were scheduled for October 29, 2026, two days after the engineering guidance was published. The results were not available at the time covered by the source material.
Source: ThorstenMeyerAI.com
