🔍 Read the full analysis: Why Amazon’s AI Bets Shift The Conversation From Usage To Impact on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Amazon’s Oct. 27 engineering guidance says software engineers should be judged by customer, business and team outcomes, not token counts or other AI usage measures. It still describes adoption data as a useful signal, while Amazon’s Nova and Anthropic investments offer contrasting examples of the company’s broader AI bets.
Amazon published revised guidance for its software engineers on October 27, telling them to measure AI by its effects on customers, the business and teams rather than by token counts or tool adoption alone. The memo responds to concerns about usage metrics becoming goals and arrives as Amazon prepares to report third-quarter results on October 29, putting the focus on whether its AI spending is producing useful outcomes.
The guidance was written by Amazon’s most senior engineers and published on the company’s news site. It says measures such as AI-tool adoption, tokens used, code changes and reviews can help teams observe activity, but are “deeply flawed” when treated as targets. The memo warns that doing so can increase costs and system complexity without showing whether customers or the business benefit.
The document does not say teams should stop using AI. It says those input measures currently “correlate well” with desired outputs, while emphasizing that correlation is not a substitute for assessing results. It asks leaders to examine both the lowest and highest levels of AI adoption: why some teams are not finding the tools useful, and whether heavy users are actually delivering more.
It also says engineers should record cases where current AI tools do not help. “If your work does not benefit from current AI tooling, that is a finding, not a failing,” the memo says. Leaders are asked to use the tools in their own production work and describe tasks where they helped and where they did not. The guidance avoids a fixed definition of AI use in engineering roles, saying that one written now could quickly become outdated.
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 Work Outcomes
The memo matters beyond an internal performance policy because companies are trying to work out how to judge returns from AI tools. Usage is easy to count; improvements in quality, speed, customer experience or cost are harder to establish. Amazon’s guidance draws a line between using adoption data to identify patterns and using it to rank employees or teams.
That distinction has practical consequences. A target based on tokens or code volume can reward unnecessary activity, while a team with limited usage may have a sound reason: its work may not suit current tools, or the tools may not yet improve its process. Amazon’s approach, as described in the memo, asks managers to investigate those differences and connect adoption to outcomes rather than assume that more usage automatically means better work.
For readers tracking the company’s AI strategy, the timing is also relevant. Amazon is due to report quarterly results on October 29. The source material provides no forecast for those results, and the memo itself does not establish the financial impact of AI. It does, however, make clear what Amazon’s engineering leadership says should count internally: impact, not activity alone.
Products Worth Considering
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
Logitech MX Mechanical Wireless Illuminated Keyboard Tactile (Graphite)
HP Sprocket Photo Booth Machine Instant Color Photo Printer with 10.1 Touchscreen, LED Ring Light, 250+ Frames, 3×4 ZINK Paper, QR Code Shar
As an affiliate, we earn on qualifying purchases.
How Amazon Reached This Guidance
The memo follows earlier reporting about internal AI-use tracking. In May, the Financial Times reported that some Amazon employees assigned unnecessary tasks to an internal agent tool called MeshClaw to increase token counts. That account was based on anonymous employee reports and should be treated as reporting, not as a company-confirmed finding. The same reporting described weekly AI-adoption targets and internal usage leaderboards.
Amazon did confirm to Business Insider at the end of May 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 meant to encourage 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 takes that message further by setting out how engineers and managers should think about measurement.
The source article also compares Amazon’s own Nova models with its investment in Anthropic. Business Insider reported in July that most flagship Nova models had been moved to a “keep the lights on” status; Amazon’s lifecycle page lists Nova Canvas, Nova Reel and the first Nova Sonic as retired. The article says Amazon invested about $18 billion in Anthropic and that a June 30 filing valued its holdings at about $190 billion. Those figures are reported values, not proof that the investment has produced equivalent operating returns. They illustrate why company AI bets need to be judged on outcomes, not spending or activity alone.
““Tokenmaxxing is at odds with frugality.””
— Amazon engineering memo, published October 27
What the Memo Cannot Show
The guidance sets out how Amazon wants engineers to evaluate AI, but it does not quantify whether AI tools have improved productivity, lowered costs or changed customer outcomes across the company. The memo’s statement that adoption metrics correlate with desired outputs does not provide a numerical estimate or establish that AI use caused those results.
It is also unclear how consistently managers will apply the guidance or what evidence they will use to assess impact across different kinds of work. The earlier reports of token inflation relied on anonymous accounts; Amazon’s confirmation that KiroRank was discontinued did not confirm every detail in those accounts. The source material likewise does not establish the eventual commercial performance of Nova or the realized return on Amazon’s Anthropic stake.
Amazon’s Results Due October 29
Amazon is scheduled to report its third-quarter results on October 29, one day after the memo’s publication. Those results may offer further company-wide financial context, but the source material does not say they will break out the returns from AI tools, Nova or Anthropic. The memo’s immediate next step is internal: managers and engineers are expected to use evidence from their work to distinguish useful AI adoption from activity that does not improve results.
Any assessment of the policy’s effect will require more than a count of tokens or tool users. Readers should look for disclosed measures tied to outcomes, such as customer impact, operating costs or delivery quality, and for clear explanations of the period and comparison used. Until Amazon provides that evidence, the guidance is a statement of how the company intends to evaluate AI work—not a quantified report of what that work has achieved.
Key Questions
What did Amazon change on October 27?
Amazon published revised engineering guidance saying engineers should be evaluated by customer, business and team impact, not by AI usage measures such as token counts on their own.
Is Amazon telling employees to stop using AI?
No. The memo says adoption metrics can correlate with desired outputs and asks leaders to examine how teams use AI. Its point is that usage should not become a performance target detached from results.
What does “tokenmaxxing” mean in this report?
It describes assigning AI tools extra or unnecessary work to raise token counts. The source material says earlier reporting described this behavior among some Amazon employees; those accounts were attributed to anonymous sources.
What does Amazon’s memo say about employees who do not benefit from AI tools?
It says a lack of benefit from current tools should be documented as a finding, not automatically treated as a failure. The guidance asks teams to explain where AI helps and where it does not.
Will Amazon’s October 29 results show whether its AI bets are paying off?
The results are scheduled for October 29, but the source material does not say whether Amazon will disclose returns specifically for AI tools, Nova or Anthropic. The memo itself provides no quantified measure of AI’s financial impact.
Source: ThorstenMeyerAI.com
