Amazon tracks engineers' AI use, ties to performance

Business Insider reports an internal document from Amazon's retail organisation, referred to as Stores, that maps the companywide AI rollout and measures adoption among engineering teams. The Business Insider story says the document tracks how many engineers use AI each month, how frequently tools are embedded in workflows, and links those metrics to output goals. According to Business Insider, the document asks over 2,100 engineering teams to triple software code release velocity with "AI-native" practices, and designates at least 25 teams to boost output by tenfold this year; the report says the progress is reviewed by Amazon's senior leadership team, the "S-Team." KMJ, citing reporting from The Information, reports Amazon is using an internal tracking system called Clarity to record which models employees use and how often, and that AI adoption is being factored into performance-review and promotion eligibility. Both outlets report employee pushback, including complaints about mandates and restrictions on using external tools.

Companies rolling out developer-facing generative tools typically instrument usage at the team and individual level to measure adoption and ROI. Industry-pattern observations note that such instrumentation commonly includes telemetry on API calls, model choice, prompt reuse, and integration frequency. Observers also note a frequent tension between standardising on an internal tool such as Kiro and offering access to external models like Claude Code, because integrations, latency, cost, and data governance tradeoffs vary by model.

For practitioners, tying tool adoption to productivity metrics raises familiar governance questions. Industry observers highlight that using adoption as an input to performance evaluations can accelerate uptake but also risks gaming, Goodhart effects, and morale problems if tool quality or autonomy are limited. Comparable corporate programs have produced faster automation of routine tasks but also generated pushback when engineers perceive mandates as reducing autonomy or when tooling fails to deliver expected productivity gains.

Indicators to follow include whether Amazon or its teams publish follow-up metrics on release-velocity gains or quality regressions, whether internal tooling choices (for example favoring Kiro over third-party models) change, and if external reporting names specific product or governance changes tied to Clarity data. Industry observers will also watch for broader adoption patterns across retail and cloud teams and any shifts in promotion criteria documentation.