🔍 Read the full analysis: Amazon’s AI Playbook Puts Engineering Outcomes On The Radar 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 engineering guidance on October 27 telling software teams to judge AI by its effect on customers, the business and colleagues—not by tokens used or tools adopted. The memo says usage metrics can still help identify patterns, but turning them into targets risks higher costs and unnecessary complexity.
Amazon published revised guidance for its software engineers on October 27, saying teams should judge artificial intelligence by its effect on customers, the business and colleagues—not by how many tokens they use or how often they adopt AI tools. The memo responds to concerns about usage measures becoming goals, as Amazon’s own AI investments face differing reported outcomes and the company prepares to report third-quarter results on October 29.
The memo, published on Amazon’s news site and attributed to Terk Terkowitz, a vice president and distinguished engineer in the company’s stores business, introduces a revised role guideline for software engineers. Written by senior engineers, it keeps AI usage out of the job definition, arguing that any formal definition of AI work could quickly become outdated. Instead, it anchors the role in impact on customers, the business and the team.
Amazon acknowledges that long-term results can be difficult and slow to measure. Some teams have used inputs such as token usage, AI tool adoption, code changes and code reviews as substitutes. The memo calls those measures “deeply flawed” when used as targets, warning that they can raise costs and add complexity. It also says the measures currently “correlate well” with important outputs, so it does not recommend discarding them as indicators.
The guidance asks leaders to examine both low and high adoption. They should investigate why teams near the 10th percentile use AI less, and whether teams near the 90th percentile are delivering better results. Engineers who find that current AI tools do not help their work are asked to record and share that finding; the memo describes it as “a finding, not a failing.” Leaders are also told to use the tools on production work and explain 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
From AI Activity to Engineering Results
The guidance sets a distinction with practical consequences for performance reviews, engineering costs and product quality: AI adoption may be useful to track, but the memo says it should not stand in for value delivered. If usage is rewarded for its own sake, employees may spend time generating activity that does not improve a product or service.
For engineers, the position gives official backing to the idea that declining to use a tool can be reasonable when it does not fit the task. For managers, it shifts attention from easy-to-count inputs toward harder questions about whether work became faster, simpler or more useful. The memo’s approach does not establish that AI has improved results across Amazon; it calls for leaders to assess that relationship rather than assume it.
The policy also offers a lens on Amazon’s wider AI strategy. The source account describes the company’s Nova models and Anthropic investment as contrasting bets, but those comparisons are not the subject of the engineering memo itself. They illustrate why outcomes, rather than activity or spending alone, are relevant measures for a company investing heavily in AI.
Products Worth Considering
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with Mic 2, 1 CMOS, 4K/120fps, 3-Axis Stabilization, 128GB Card
Apple iPad Air (2025) 13‑inch Wi‑Fi 256 GB – Space Gray (M3 chip)
As an affiliate, we earn on qualifying purchases.
Earlier Usage Targets and AI Bets
The memo follows earlier reports of internal attention to developer AI usage. The Financial Times reported in May that some Amazon employees gave an internal agent unnecessary tasks to raise token counts. Those accounts were anonymous and remain reported allegations, not facts established by the October memo. The source material also says a company spokesperson confirmed to Business Insider that an internal leaderboard called KiroRank had been “deprecated,” describing it as an informal employee tracker that was not intended to encourage AI use for its own sake.
The source account says senior vice president Dave Treadwell told staff, “Please don’t use AI just for the sake of using AI.” The October guidance formalizes a similar principle in the engineering role guideline. It does not say that teams should stop using AI; rather, it preserves usage data as a possible signal while rejecting it as a standalone measure of performance.
Amazon’s model portfolio and its relationship with Anthropic provide additional, separate context. The source cites a Business Insider report that most flagship Nova models were shifted to maintenance, alongside AWS lifecycle listings showing several Nova products retired. It also cites Amazon filing figures valuing its Anthropic holdings at about $190 billion on June 30, against roughly $18 billion invested. Those figures and reports do not establish the commercial performance of either bet, and the filing valuation is not the same as cash returned.
“Tokenmaxxing is at odds with frugality.”
— Amazon engineering memo
How Amazon Will Judge AI Impact
The memo does not set out a single replacement metric or specify how individual performance reviews will weigh AI-assisted work. It says outcomes matter, while acknowledging that they take time to measure; its proposed checks rely on managerial judgment and comparisons between adoption levels. How teams will define and verify improved outcomes remains unspecified in the source material.
It is also unclear whether Amazon will change any company-wide AI targets, budgets or internal tracking systems as a result. Earlier reporting about employees inflating usage was based on anonymous accounts, and the memo does not confirm every reported practice. The company’s third-quarter results, due October 29, may provide financial information but would not by themselves show whether AI tools improved engineering productivity.
The cited Nova product changes and Anthropic investment figures do not provide a like-for-like assessment of returns. The source material does not establish the long-term revenue, cost savings or customer impact of either AI bet.
Amazon Results Due October 29
Amazon is scheduled to publish its third-quarter results on October 29, one day after the memo. Those results may add financial context to the company’s AI spending and investments, but they should not be treated as a direct test of the new engineering guidance: the memo does not supply a timeline or outcome benchmark for measuring its effect.
Within engineering, the next stated steps are for managers to review low- and high-adoption teams, use AI tools on their own production tasks, and share what they learn about where the tools help. The source material does not identify a date for a follow-up policy review or disclose how Amazon will report the findings. For now, the confirmed change is the principle the company has put in writing: measure engineering impact, not AI usage by itself.
Key Questions
What did Amazon tell engineers about AI usage?
Amazon said engineers should be judged by customer, business and team outcomes, not by token counts or AI tool adoption alone. Usage data may still help teams spot patterns.
Is Amazon telling employees to stop using AI?
No. The memo does not reject AI use. It says teams should assess whether AI improves results and should not treat usage itself as a performance target.
What does the memo say about employees who use AI less?
Managers are asked to examine why some teams use AI less. The memo says that finding current tools unhelpful for a task can be “a finding, not a failing”, and asks employees to share their reasoning.
When will Amazon report its third-quarter results?
Amazon is scheduled to report its third-quarter results on October 29, 2026, two days after the October 27 memo.
Does the memo prove Amazon’s AI investments are paying off?
No. The memo concerns engineering guidance, not a financial assessment of Amazon’s AI portfolio. The source material cites reports and filing values related to Nova and Anthropic, but does not establish their long-term commercial returns.
Source: ThorstenMeyerAI.com
